Recent Financial Crises

Introduction
Lawrence R. Klein and Tayyeb Shabbir
1 ABOUT THIS BOOK
This volume has been assembled to focus on specific analytical as well as

policy-related issues pertaining to recent financial crises, most notably the
Asian Financial Crisis of 1997–98 and the ripple effects that translated into
uncomfortably close calls in Russia, Brazil and, somewhat later, full-blown
crises in Turkey and Argentina. Of course, much has been said and written
about the various aspects of these recent crises. However, motivated by
the belief that the issues raised by and surrounding these recent crises (in
fact, financial crises in general) are of ongoing importance, and despite
a perspective gained by scholarly analyses of these crises, there are import-
ant lessons to be learned and analytical issues that can be explored even
further.
It may be noted that the papers included in this volume have been
specifically prepared for it and thus have the distinction of being original
and not reprints, affording an analysis that the respective authors are
sharing for the first time.
Thematically, we have divided the chapters into three parts. Besides
providing an introduction, Part I addresses the issues of predictability
of currency crises; Part II consists of chapters that focus on a set of
reforms or ‘cures’ for preventing and/or ameliorating the after-effects of a
crisis and, finally, Part III consists of a set of econometric studies that
address several issues of analytical interest pertaining to labor market
behavior, investment and productivity, exchange rate adjustments and
estimation of China’s core inflation rate, as well as the ‘true’ cost of living
index for China over the twenty-year period that spans the Asian
Financial Crisis.
A brief guided tour of the contents of the specific chapters is given in
section 3, while in section 2 below we note the important questions raised
by the recent financial or currency crises.
1
2 Recent financial crises
2 QUESTIONS OF IMPORTANCE RAISED BY THE

RECENT FINANCIAL CRISES
The last decade or so has been marked by a significant number of major

financial crises including the Mexican (‘Tequila’) Crisis (1994–95), the
Asian Financial Crisis (1997–98), the Russian Bond Market Default
(1998), the Turkish Banking Crisis (2000) and the Currency Crisis that
followed (2001) and the Argentinean Financial Crisis (2002). While most
of the affected countries, to a lesser or a greater degree, have apparently
recovered from the immediate negative consequences of these crises, the
outcome could have been much worse in terms of the time to recover and
sustainability of crisis-free development. In any event, there exist the rel-
atively longer-run effects of these recent crises, which are still in the
process of unwinding. These secular after-effects are relevant, both for
practical as well as analytical or academic reasons. In practical terms,
several of the affected countries are still grappling with the secular after-
math of these crises. In addition to the adoption of country-specific eco-
nomic and financial sector reforms, these countries are still facing social
sector issues, such as income distributional effects and setbacks to trends
towards poverty alleviation that so strongly characterized the 1980s and
the early 1990s. On the other hand, in analytical terms or broad-based
academic terms, we still need to make sure that that the appropriate
lessons from these crises have been learned, and more importantly are
being instituted, regarding the workings and reform of the global
financial system. While much scholarly work focused both on the analyt-
ical and policy-oriented issues has already been undertaken, we believe
that there are still very important issues that need to be addressed regard-
ing the nature of these crises. The following is a selective list of some of
these important issues that are of special interest to us.
2.1 Nature and Dynamics of the Crises
How much have we learned about the nature and the dynamics of these
crises? In particular, are these crises predictable to any appreciable or
practical extent – say, enough to be able to allow us to develop an early
warning system? What are the factors that make economies vulnerable
to financial crises? (Chapter 1 by Klein and Shabbir as well as Chapter 2
by Tinakorn.) Again, can the policy-makers recommend policies that
will soften the impact of, if not eliminate such crises? (Chapter 3 by
Eichengreen.)
Introduction 3
2.2 Lessons Learned
What kinds of specific lessons have we learned, if any, from these episodes?
What kinds of economic, fiscal, monetary and perhaps even political
reforms will be necessary in order to ameliorate the vulnerability of a
country to a financial crisis? In this respect, it is heartening to note that con-
siderable progress has already been made in instituting certain reforms of
the global financial system that have evidently been inspired by the recent
financial crises. Prodded by the IMF and the Bank for International
Settlements (BIS), increasingly countries now report the value of non-
performing loans (NPLs) as a way of monitoring the commercial banking
sector. In addition, there has been a consistent trend towards emphasizing
the importance of maintaining transparency and prudent risk management
by private financial institutions (such as the hedge funds) as well as the
central banks of the respective countries. The latter are now routinely
expected to furnish prompt and regular reports regarding the amount as
well as the disposition of the international reserves they manage. The
occurrences of these recent financial crises have taught us the importance
of continuing to strengthen these trends.
Since the affected countries have mostly recovered, at least economically,
from the recent crises, there is a real danger that, for practical purposes, a
significant complacency may set in. Obviously, we must guard against this
possibility – revisiting the question of the lessons learned and how best to
implement them to minimize recurrence and/or to mitigate the severity of
future crises. The description of one very important way to do just that can
be found in Chapter 3 by Eichengreen.
2.3 ‘Cures’ for the Crises
The next set of issues is often presented under the rubric of ‘cures’ for
financial crises – actually they are the generally recommended reforms
apparently motivated by the experience of financial crises. We will discuss
three such broad categories of the so-called ‘cures’: appropriate exchange
rate policy, capital market reforms and private business governance.
2.3.1 Appropriate exchange rate policy

Regarding the question of the appropriate exchange rate policy, to many,
the appropriate policy is simply to have no deliberate policy at all! ‘Let the
market take care of the issue completely’, they declare. While it is true that,
in retrospect, the 1997–98 Asian Crisis marked a watershed ‘event’ that has
convinced many of the desirability of a completely flexible exchange rate,
many subtleties still need attention and the question of the appropriate
choice of an exchange rate system is not an open and shut case, as it is

sometimes made out to be. There are questions regarding the desirability
and/or feasibility of a single world or wide-area currency, or in general, the
extent and the mechanism for managing exchange rates in the context of
the new global financial architecture. Increasingly, private capital flows
have come to dominate trade flows and have thus lent a historically high
level of volatility in price behavior in this important market. The capital
mix has changed – there is, for example, more foreign direct investment
(FDI) and less capital flow on official account.
Again, there are important unresolved issues concerning how best to deal
with chronic current account imbalances (surpluses as well as deficits),
management of international reserves by the central banks, possibilities of
currency contagion, current account-induced flows and debt flows. A par-
ticularly vexing question is the appropriate policy response to a trade
surplus in the case of a large, fast-growing creditor country such as China.
It turns out that an often-recommended solution – appreciation of the yuan
relative to the US dollar – may do little to rectify the problem if the under-
lying cause of the hugely favorable trade surplus for China is to be found
in international wage differentials estimated to be in the range of 20 or 30:1
in favor of China, even though the differential has been narrowing lately
(Chapter 8 by McKinnon).
2.3.2 Capital market reforms

Besides currency reforms, capital market reforms are also critical for the
emerging economies. In this regard, there are two dominant issues and we
will refer to them as (a) optimal degree of liberalization and (b) relative
completeness of the scope of capital market instruments or institutions.
Capital market liberalization is desirable because it increases efficiency
through better allocation of resources. However, such efficiency gains may
come at a price since liberalization, inter alia, makes the economy more vul-
nerable to contagion and a host of other external shocks. Obviously, these
two opposing effects need to be balanced, leading naturally to the notion
of the ‘optimal capital market liberalization’. There is a related question of
what is the ‘safe’ rate at which transition should take place? Such a deter-
mination of optimality would depend, in part, on having on hand analyzes
of the various policies that can be used to mitigate the downside of capital
market liberalization. One such policy, to contain the potential vulnerabil-
ity due to openness on account of capital market liberalization, can be the
adoption of capital controls. However, such policy needs to be designed
precisely and thoughtfully, with assessment of relevant side-effects. There
are relatively few studies that undertake such an assessment. However,
Klein, Mariano and Özmucur (Chapter 4) introduce an econometric case
Introduction 5
study of Malaysia, which assesses the effects of capital controls, to help

manage the financial crisis of 1997–98 in that country and finds that the
capital controls there had many positive and desired effects in terms of eco-
nomic stabilization.
Regarding the other issue that we have termed ‘relative completeness of
the scope of capital market’, it has been noted that the emerging economies
lack well-functioning bond markets, even when they have fairly well devel-
oped equity markets. This is as if one leg is missing, and it makes the
country less resilient in the face of a potential financial crisis. However, the
study of bond markets in emerging economies is starting to receive increas-
ingly greater attention. There is a burgeoning interest in the study of ques-
tions such as the role of a viable bond market in improving the efficacy of
financial intermediation in these countries as well as how a robust bond
market may be a useful buffer to prevent or soften a future crisis. (Chapter
5 by Herring and Chatusripitak is a case study of the Thai bond market in
relation to the 1997–98 Asian Financial Crisis; a related, yet independently
arrived at, analysis of the importance of a bond market may be found in
Chapter 3 by Eichengreen.)
2.3.3 Private business governance

Finally, regarding the third aspect of the capital market reforms, we want
to note that corporate governance is an important issue in this respect.
Indeed, it is one of the host of other emerging issues that pertain (albeit
indirectly at times) to the possibility of future financial crises in emerging
economies. The question of corporate governance, in part, encompasses
concerns about the role of minority shareholders, full disclosure and trans-
parency, prudent risk management and the role of the board of directors
as independent overseers. It also takes up behavioral issues such as conflict
of interest and use of inside information (Chapter 3 by Eichengreen).
2.4 Secular Aftermath of the Asian Financial Crisis 1997–98
While the various economic indicators have shown sustainable and solid
improvement in the majority of the affected countries, their respective
social indicators are definitely still lagging relative to the pre-crisis trends.
The financial crisis in Asia left a deep mark in terms of social upheaval in
these countries, many of which were unprepared for such an eventuality.
We need to evaluate the income distributional impact in the medium- to
the long-run sense as well as to assess the state of the present and future
adequacy of the social safety nets in emerging economies that may be espe-
cially prone to financial crises (Chapter 1 by Klein and Shabbir). One par-
ticularly important question pertains to the behavior of the real wage rate
in the affected countries pre- versus post-Asian Crisis. (Chapter 7 presents

a case study of Thailand regarding real wage rate behavior pre- versus post-
Asian Crisis.)
2.5 Focus on New Developments in the Asian Region
We feel that in any discussion of the recent financial crises, in addition to

reflecting on the past, we ought to be forward-looking as well. In this spirit,
we want to note the emerging developments in East Asia/or Greater Asia as
a region. Our primary interest in this region lies in particularly focusing on
countries that were directly affected by the Asian Crisis of 1997–98, such as
South Korea and Thailand, as well as China and Japan, which were only
indirectly involved in one or the other aspect of the above crisis because they
are important actors in the region. Finally, we also want to indulge in a
prospective look at India as an emergent regional economic powerhouse and
thus see how the future may look in the wake of post-crisis Asia, where China
and India have emerged as the relatively fastest-growing dynamic and open
economies (at least in the aspects of trade flows), which may be the domi-
nant part of the ‘New Asian Miracle’ going forward and thus may stand to
gain the most from any valuable lessons that the recent financial crises can
teach us. We will, thus, divide the relevant region into three groups: East
Asian directly affected countries China and Japan and India (South Asia).
2.5.1 East Asian directly affected countries

An important question is, how will the experience of the last crisis trans-
form the affected countries? It seems that each of the affected countries has
been responding in unique fashion when it comes to the speed of recovery,
depth of the commitment to reforms and adoption of longer-run reforms
to prevent or mitigate the effect of the ‘next crisis’. There are some initial
hints that these countries may no longer represent a monolithic group that
seemingly followed a nearly uniform mantra for growth. Also, these coun-
tries, albeit in varying degrees, each maintained a peg of their respective
currency to the US dollar. Instead, it seems that many of these countries
will be following fairly differentiated paths from now on. These idiosyn-
crasies in the chosen development strategies are important both from a
policy perspective as well as their theoretical perceptions of growth and
development. Is a new paradigm of growth emerging that may lessen the
future possibility of financial contagion?
2.5.2 China and Japan

China’s case is important from many points of view. China of course, has
been growing tremendously fast since reform in 1978. It is poised to be a
Introduction 7
dominant part of future economic prosperity of the region – a ‘New Asian

Miracle’, if you will. Its openness in terms of trade flows makes it vulner-
able to crises. However, it has a longstanding policy of minimal variations
in the exchange rate in the face of persistent current account surpluses. This
is just the opposite of what Thailand faced on the eve of the crisis in
1997–98, and decidedly is a relatively better situation to be in than Thailand
was when it faced shortages of international reserves rather than having to
worry about disposal of surplus. However, an imbalance is an imbalance,
all the same. Thus, China is also at the center of the debate about the appro-
priate exchange rate and the capital controls policies to adopt. These are
issues that have been an intimate part of the financial crisis experience.
China is also an obviously important player through its trade and direct
investment links to the region. Thus focusing on China is only natural.
However, we know that some people have expressed a degree of skepti-
cism about the veracity of China’s growth statistics. This is an aspect of the
Chinese economy that is of interest to us. Besides addressing the question
of China’s ‘true’ growth rate, we are also interested in exploring what role
China may have played – a stabilizing or a destabilizing one – during the
1997–98 Asian Crisis.
Of course, China is a major regional power and thus very strategic.
However, the debate about China’s exchange rate posture has captured
wide attention. Its policy of nearly fixed parity of yuan per dollar and
capital controls in the face of burgeoning current account surpluses has
exposed the currency to potentially destabilizing speculative and political
pressure because vested interests feel effects of large competition from
many countries. While these pressures are of a different nature in compar-
ison with those that Thailand had to face in the weeks and months preced-
ing the bhat devaluation of 2 July 1997, still, in a fundamental sense, both
kinds of such pressures represent essentially two sides of the same coin if
the fundamental goal is to avoid exchange rate instability. The major ques-
tion, of course, is whether we have learned enough from the previous crises
so that we can more confidently determine the optimal degree of openness
of a country’s capital and exchange rate markets.
The other important regional power, Japan, is currently showing some
signs of a nascent recovery from a long slump. However, the question still
remains whether Japan has recovered enough to lend a strong helping hand
in the event of the next crisis erupting.
2.5.3 India (South Asia)

There are historic changes afoot in South Asia as well. India, in particular,
has started to emerge as an Asian ‘giant’, side by side with China. Both
these countries have enjoyed strong growth in the real sector (GDP)
although they reached high growth in unique fashion – China’s has been a
case of growth led now by exports of manufactured goods, while India’s
growth has been fueled by growth of service sector activity (especially in the
information technology [IT], financial and health sectors). While each of
these countries have exhibited impressive rates of GDP growth in recent
years, there are significant financial sector issues such as stock and bond
market reform, greater transparency and better corporate governance that
still need attention. It is clear that going forward, financial sector and
capital market reforms will be necessary for the continued success of these
countries while keeping in mind some of the important ways – political and
structural – in which these countries differ from each other. However, the
important point is that, as the new players on the growth playing field so to
speak, these countries’ reliance on open trade and foreign direct investment
flows makes them vulnerable to financial crises. It is worth watching to see
whether lessons from the recent crises were learned well. During the crisis
period, neighbors and other competitors blamed China for stealing their
export markets. Now they enjoy China’s import appetite. As for India,
advanced countries blame ‘white collar’ unemployment problems on
India’s burgeoning service sector, but should be benefiting from reduced
operating costs in the near future.
3 SALIENT POINTS OF THE CONTRIBUTIONS TO

THE VOLUME
The following is a brief guided tour of the contents and the themes
explored in the various chapters in this volume.
Chapter 1: ‘Asia Before and After the Financial Crisis of 1997–98:

A Retrospective Essay’ by Lawrence R. Klein and Tayyeb Shabbir
Since financial crises can be costly both in economic, as well as, societal
terms, it is only natural to inquire whether such crises are predictable.
Besides providing a brief introduction, this chapter reviews the various
approaches to prediction of currency crises. In this respect, the authors
conclude that while econometric predictive models can be very useful in
identifying various indicators of ‘vulnerability’, such exercises are not
a cure-all. Therefore, exploring the various aspects of the aftermath of a
financial crisis is also quite important. In this regard, the authors focus on
the income distributional consequences of the Asian Financial Crisis of
1997–98 as well as its impact on the poverty alleviation trends in the
affected countries.
Introduction 9
Chapter 2: ‘Indicators and Analysis of Vulnerability to Currency Crisis:

Thailand’ by Pranee Tinakorn
This chapter by Professor Tinakorn is a case study of Thailand, a country

that has the dubious distinction of being the first country in Asia whose
currency succumbed to speculative pressures on 2 July 1997, setting off
well-known ripple effects, which impacted not only some of the neighbors,
but also Russia and Brazil. This series of currency crises (and, in many
instances, follow-up banking crises) prompted anew many efforts to
‘predict’ currency crises so as to pre-empt them or, more realistically,
decrease the vulnerability to such harmful episodes. In this spirit,
Tinakorn’s study is an important attempt to identify an early warning
system.
Tinakorn uses time-series monthly data from January 1992 to December
2000 for Thailand to estimate a probit model as well as a ‘signals’ model à
la Kaminsky et al.1 (1998) to analyze the indicators of a currency crisis in
Thailand. Tinakorn defines a currency crisis as when ‘there is an accumu-
lated three month depreciation in exchange rate of 15 percent or more;
or there is an accumulated three-month depletion in net international
reserves, of 15 percent or more’. It is important to consider net interna-
tional reserves, which are gross reserves adjusted for swap obligations of the
Bank of Thailand (BOT) – a decidedly better measure than just the equat-
ing of currency crises to (realized) currency depreciation, since the mone-
tary authority may successfully ward off speculative attacks by depleting
international reserves. (This is an interesting point since anyone focusing
only on gross reserves would have missed the fact that swap obligations had
skyrocketed for BOT in 1996–97 during this [ex post] ‘window of vulnera-
bility’ for the country.)
This study’s main finding is that, in the case of Thailand, there are several
early warning indicators of a currency crisis that are worth watching. The
following seven are supported simultaneously by the ‘signals’ analysis as
well as the estimated probit model for the country: export growth, ratio of
current account to GDP, real exchange rate misalignment, growth of
M2/international reserves, ratio of fiscal balance to GDP, real GDP growth
rate and change in stock prices. Further, the signals analysis identifies the
following four indicators in addition to the above seven: terms-of-trade
growth, ratio of short-term external debt to international reserves, growth
of domestic credit/GDP and inflation rate.
Thus, in hindsight, an awareness of the above set of early warning indi-
cators by the policy-makers in Thailand would have been helpful in either
handling the management of the international reserves better and/or it
might have enabled them to negotiate a less contractionary package with
the IMF, thus possibly ameliorating some of the extreme effects of an

unanticipated devaluation and the contraction of the economic activity
that necessarily followed. However, there is an important issue of trans-
parency. A former IMF official has indicated in a lecture at the University
of Pennsylvania that Thailand’s international reserve accounts were in
trouble prior to the crisis outbreak, but the international institution would
not have made that known to the public on its own authority in advance,
for an obvious fear of panic in the financial markets.
Chapter 3: ‘The Next Financial Crisis’ by Barry Eichengreen
In this chapter, Barry Eichengreen contends that, as of spring 2005, the

‘next’ financial crisis for emerging economies may already be brewing but
it may or may not materialize since, as a result of some of the lessons
learned from the East Asia Crisis of 1997–98, we have a more stable global
financial system as well as improved country-specific financial sectors
rendering emerging economies less vulnerable to any impending crisis.
However, lest we feel tempted to rest on our laurels, the author lists an
agenda for future reforms, in addition to providing an excellent assessment
of the reforms that have already been put in place.
Eichengreen maintains that, as of spring 2005, the following factors con-
stitute a ‘potentially fatal cocktail’ that may foretell a crisis in emerging
economies:
1. Rising US interest rates as a result of the FED’s (Federal Reserve’s)

reversal of its easy monetary policy it had adopted as of spring 2001.
In the past, US domestic interest rate increases have been harbingers
of financial problems in emerging markets.
2. The ever-increasing US ‘twin deficits’ – the current account deficit as
well as the dudget deficit.
3. The international oil price increases, which may lead to a significant
slowdown in China’s growth momentum, which will mean a negative
ripple effect throughout the emerging economies since China has lately
emerged as a significant engine of regional as well as global economic
growth.
While the above factors may appear to be clouding the horizon for the emerg-
ing economies, Eichengreen stresses that ‘no ill effects are evident yet’
perhaps due to the reforms instituted since the last crisis in 1997–98. He
singles out ten reforms already instituted as noteworthy, namely, lengthening
the maturity structure of emerging economies’ debt, their smaller current
account deficits, larger foreign reserve stockpiles, relatively greater flexibility
Introduction 11
of the exchange rate mechanisms, greater fiscal ‘responsibility’ as well as

reduced leverage in international financial systems, greater multilateral sur-
veillance of the financial systems, greater financial sector transparency,
incorporation of collective action clauses in sovereign debt instruments, and,
finally, comparatively greater transparency of the IMF itself.
Besides noting the already instituted reforms that have been motivated
by the lessons learned from the recent financial crises, Eichengreen also
stresses that we need to continue on this path. He stresses five such future
reform targets in emerging economies: continuity of earlier reforms,
improved credit and bond markets, enhanced exchange rate flexibility, fea-
sibility to borrow in own currency on the world capital markets and con-
tinued governance reforms at the Bretton Woods Institutions.
One unique feature of Eichengreen’s chapter is that while assessing the
reforms put in place in the last few years, he also sheds light on the ‘costs’
of such reforms – a hitherto neglected area of research and analysis. In this
regard, if one were to add a ‘sixth’ item to the above agenda for future
reforms, it could be the goal of maximizing the ‘net [of costs] benefits’ of
these reforms that are inspired by a desire to learn from the past crises.
Chapter 4: ‘Capital Controls, Financial Crises and Cures: Simulations

with an Econometric Model for Malaysia’ by Lawrence R. Klein,
Roberto S. Mariano and Süleyman Özmucur
Liberalization of capital account is generally favored as a desirable policy

as it can lead to greater availability of capital and increased efficiency. These
‘pro-growth’ effects, however, may be tempered by the fact that liberaliza-
tion of capital account can render a country more vulnerable to financial
crises due to an enhanced exposure to external shocks.
In the face of a financial crisis when the domestic policy-makers often
experience the feeling of a loss of control of their economy, an imposition
of capital controls (at least temporarily) is often advocated as a solution.
Such controls, it is argued, can help to bring under control the possibility of
panic-driven flight of capital and uncontrolled depletion of international
reserves in the wake of a crisis. However, such controls are generally disfa-
vored by those who hold the relatively orthodox view that considers any
interference whatsoever with the ‘market mechanism’ as counterproductive.
On the eve of the 1997–98 crisis, Malaysia imposed capital controls to
manage the financial crisis. This chapter by Klein–Mariano–Özmucur is a
description of use of a simultaneous equation macroeconometric model – a
method to study the possible impact of capital controls while taking
account of the various feedback effects. Such a model is estimated for
Malaysia. The model is a fairly detailed one with 438 equations and 607
variables. Such detail allows the authors to supplement a traditional macro-

economic model by incorporating a very thorough specification of the exter-
nal sector; in particular, the authors specify capital account, foreign direct
investment as well as portfolio investment, as endogenous to the model.
The major conclusion of this chapter is that capital controls in Malaysia
had a number of desirable effects on important macroeconomic indicators
such as the real GDP growth rate (the net effect was an increase of 0.07
percent over the post-crisis period, 1998–2001). The GDP deflator (whose
rate of change, of course, measures inflation rate) would also have been
higher in the absence of capital controls. Thus, generally speaking, in the
case of Malaysia, the capital controls helped the policy-makers manage the
crisis better and the economy enjoyed many positive stabilization benefits
with minimal negative effect on foreign direct investment inflows. It is,
however, important to recognize that the success of capital controls seems
to be contingent on the particular ‘context’ in which they are imposed,
quality of policy intervention and the initial conditions in the economy.
Chapter 5: ‘The Case of the Missing Market: The Bond Market and
Why It Matters for Financial Development’ by Richard J. Herring and
Nathporn Chatusripitak
Over the last decade, there has been an increased interest in analyzing the
role of financial institutions and financial markets in economic growth
and development. However, the main focus has been on equity markets,
and bond markets have been almost entirely overlooked. This chapter by
Herring–Chatusripitak, concentrating particularly on Asian economies,
tries to redress this situation by seeking to explain how the absence of a
well-functioning bond market may adversely affect other markets, savers,
investors and banks, and, in particular, how it may render the economy
more vulnerable to a financial crisis. It concludes with an analysis of recent
financial development in Thailand to illustrate both the problems associ-
ated with the absence of bond markets and the proposed solutions.
The authors assert that absence of well-functioning bond markets can
make an otherwise vibrant economy more vulnerable to a financial crisis (as
was the case in East Asia during the mid-1990s). One major implication of
the absence of a bond market is that the economy lacks a market-determined
term structure of interest rates that accurately reflects the opportunity cost
of funds. This deficiency can make firms under- or over-invest relative to the
societal efficient allocation on whether the firm’s internal rate of discount is
too high or too low (the latter was the case in the early to mid-1990s in East
Asia). Also, lack of ‘true’ term structure will impede accurate pricing of
equity in the stock market as well as pricing of credit risk. Again, in the
Introduction 13
absence of a well-functioning bond market, hedging in the derivatives

(including foreign exchange) market will be relatively more expensive, if it is
possible at all. As a result, market participants may end up assuming more
financial risk than they would choose if there were efficient derivatives
markets, just as the Asian Financial Crisis in 1997 showed that many market
participants had accepted excessive exposure to foreign exchange risk. The
greater risk obviously increases the vulnerability to a financial crisis.
Perhaps the most worrisome implication of an underdeveloped bond
market is that, in such economies, the banking sector is more vulnerable to
inefficiencies, liquidity shocks and falling prey to ‘moral hazards’ of the
kind often referred to as ‘crony capitalism’. Lack of competition from the
bond market makes banks ‘too big’, leads them to prefer short-term credit,
which in turn, leads to biases in firms’ investment policies in favor of short-
term assets and away from longer gestation ventures. Also, lack of oppor-
tunity for the banks to rely on the bond market to spread their own
portfolio risk makes these highly leveraged institutions more vulnerable to
a liquidity shock with obvious and often immediate repercussions for the
rest of the economy.
In Thailand, prior to the 1997 Financial Crisis, the bond market was
severely underdeveloped. As a result, Thai firms tended to over-invest, the
efficiency of investment was declining and the economy was heavily reliant
on bank lending. Consequently, when the banks suffered heavy losses, new
lending ceased and firms had to halt investment projects, resulting in pro-
longed and painful economic contraction. Since then, the Thai government
has begun to implement a number of reforms to stimulate development of
both primary and secondary bond markets. These include developing a
yield curve for government bonds, efforts to promote risk management and
market liquidity, centralizing the clearing and settlement of bonds, upgrad-
ing accounting and disclosure standards, and active participation in
regional initiatives to strengthen Asian bond markets. These changes have
markedly improved the liquidity of the government bond market, and the
Thai bond market is more than four times larger, relative to GDP, than
before the crisis. The Thai example shows that bond markets do matter for
financial development, and that an expanded role for the bond market may
be used to rebuild financial systems after the crisis.
Chapter 6: ‘Investment, Growth and Productivity during the East Asian

Financial Crisis’ by F. Gerard Adams and Tayyeb Shabbir
In this chapter, Adams–Shabbir examine the impact of the 1997 East Asian
Financial Crisis on real GDP growth and total factor productivity (TFP)
of the East Asian countries during and after the crisis. Rather than taking
the more typical approach to analyzing the crisis and its impact in terms of
such factors as financial flows, exchange rate misalignment and contagion,
the authors approach the crisis from a production input/factor productiv-
ity perspective.
In the first place, the authors look at the financial crisis and its impact on
the growth record of East Asian countries, by comparing a number of
growth characteristics pre- and post-crisis. The impact was uneven with
apparently minimal effect on China yet with serious recessionary effect on
the economies of Thailand, South Korea and Indonesia. In general, real
GDP growth rates fell sharply in 1997–98 due to the crisis, and although
the post-crisis period shows renewed growth, it is at substantially lower
growth rates than in the pre-crisis period. Also, post-crisis relative to pre-
crisis, there was a downward trend in labor productivity growth. In add-
ition, compared with the exceptionally high values for the investment/GDP
ratio, there was a sharp drop in this investment share, and post-crisis recov-
ery was of a relatively smaller magnitude. In fact, the downward swing in
investment was not matched by a similar swing in domestic saving, hence
foreign inflows turned to outflows. Finally, though exports recovered in the
post-crisis period, and somewhat offset the lower investment share, the
export growth for this period was also lower than its pre-crisis rate.
The authors then seek to disentangle growth of output into that attrib-
utable to increased inputs and the residual factor, or total factor produc-
tivity. This residual represents the difference between the growth of total
output and the weighted sum of labor and capital input and includes all ele-
ments not taken into account in the computation of growth inputs, includ-
ing technological change, economies of scale, the composition of output,
the role of exports and the cyclical position of the economy. The authors
define unexplained TFP as the change in TFP less the business cycle effect
less the industrial/export effect. Statistics show that total labor and capital
input growth remain lower post-crisis than pre-crisis.
They next undertake a statistical analysis of the factors associated with
the growth of TFP. To measure the effect of the 1997 Financial Crisis on
TFP, during which there were severe declines in production, they perform
regressions linking TFP to a series of dummy variables covering the
1998–2001 periods. They find that declines in production, particularly in
1997 and 1998, have clear impacts on TFP, that the loss in productivity
growth associated with this period was not made up in later years, and that
the coefficient of a time-trend variable was significantly negative. They also
perform regressions linking TFP to other variables such as increasing
exports, share of investment, and industrialization, and find that change in
industrial output and change in exports make significant contributions to
TFP change. Other measured factors, including foreign direct investment,
Introduction 15
did not show up with statistical significance. Unexplained TFP growth is

relatively small. Thus the ultimate equation explaining TFP growth indi-
cates that productivity growth in East Asia varies considerably with cycles
in business activity and depends greatly on the expansion of industrial pro-
duction and exports.
Chapter 7: ‘What Really Happened to Thai Wage Rates during the East
Asian Financial Crisis?’ by Jere R. Behrman, Anil B. Deolalikar and
Pranee Tinakorn
In their chapter, Behrman, Deolalikar and Tinakorn (BDT) explore what

happened to real wage rates during the Asian Financial Crisis of the late
1990s. The most important earning asset of most members of developing
countries is their labor, therefore what happens to real wage rates is import-
ant because of (a) the implications for the purchasing power of workers’
income and the design of effective anti-poverty and social safety net poli-
cies, (b) the impact on time allocations of workers, and (c) the implications
for the extent to which labor market adjustments occur through price rather
than quantity effects.
Conventional wisdom and most of the claims in previous literature indi-
cate that real wage rates fell considerably in Thailand due to the crisis. This
chapter uses as its benchmark a World Bank (2000) study of the behavior
of the aggregate Thai wage rates as a result of the crisis. This World Bank
study estimates that the Thai real wage rate fell by 4.6 percent during the
crisis. The methodology used was to subtract the percentage change in the
number of employed workers between the pre- and post-crisis period from
the percentage change in the aggregated real wage earnings between the pre-
and post-crisis period. However, BDT argue that the methodology used in
the World Bank and other studies is subject to at least four possibly import-
ant limitations that may make a considerable difference in understanding
what really happened to real wage rates and what the implications are for
policy. The chapter then examines these four limitations in detail.
First, the World Bank assumes that the number of wage recipients is a
fixed share of total employment. But the composition of employment
shifted from wage to non-wage workers, and this by itself creates a bias of
2.5 percentage points in World Bank estimates (implying ‘true’ change of
⫺4.6% ⫹ 2.5% ⫽ ⫺2.1%). Second, it assumes that hours worked are con-
stant, but, in fact, hours worked did change, which, again by itself, creates
a bias of 5.5 percentage points (implying a ‘true’ change of ⫺4.6% ⫹ 5.5%
⫽ 0.9%). Third, the World Bank weights wage recipients proportionately
to their hours worked to obtain the mean wage rates across wage recipients;
whereas BDT feel that it is more appropriate to weight individuals equally
and estimate that this aspect of the World Bank methodology by itself
causes a bias of 1.4 percentage points (implying a ‘true’ change of ⫺4.6%
⫺1.4% ⫽ ⫺6.0%). If all of these first three biases were corrected, the World
Bank estimate would change to an increase of 2 percent in Thai real wage
rates during the crisis.
However, more importantly, BDT contend a fourth issue related to the
above World Bank methodology. This issue concerns the World Bank
study’s assumption that no change took place in the composition of wage
recipients between the pre- and post-crisis periods. Instead the BDT study
looks at data for all workers and for subcategories defined by the three
observed characteristics of gender, age and schooling and finds that as a
result of the crisis, wage employment shifted relatively from females to
males, from younger to older workers, and from lower-schooled to higher-
schooled individuals – all shifts from lower to higher real wage categories.
The failure to account for these important compositional changes in World
Bank estimates means that the estimated overall average real wage rate
change is biased upwards. The authors’ best estimate of how much real
hourly wage rates declined due to the crisis is 7.8 percent (due to the
methodology used by BDT, their estimate is free from the first three biases
that the World Bank study had to contend with).
The major conclusion of the study was that the methodology used by the
World Bank presents a misleading picture of what happened to Thai real
wage rates during the Asian Financial Crisis. Although the biases in the
World Bank study are partially offsetting, the severity of the impact of the
crisis on declines in the real wage rate is underestimated by 3.2 percentage
points or about 40 percent (comparing a World Bank estimate of a 4.6
percent decline with this study’s preferred estimate of a 7.8 percent decline).
This study notes that even the decline of 7.8 percent is probably an under-
estimate of the true decline because of the probable compositional changes
that occur because of unobserved characteristics such as ability and moti-
vation. The authors feel that the best solution would be for longitudinal
labor force data to be collected as a matter of routine so that comparisons
could be made for the wage rates for the same individuals over time. If these
longitudinal surveys are not available, then a second-best solution is to
follow the methods in this chapter, in particular controlling for composi-
tional changes with respect to observed characteristics such as gender, age
and schooling.
Besides the above question of computing the ‘true’ magnitude of the
decline in the Thai real wage during the 1997 crisis, the BDT study also
examined the claim that the poorer and more vulnerable suffered most in
the crisis and found some, but limited, support for that claim. Regression
estimates showed that youths fared worse than prime-age adults, but that
Introduction 17
some other groups typically characterized as more vulnerable fared rela-

tively well: those with primary or less schooling, females and older adults.
Chapter 8: ‘Exchange Rate or Wage Changes in International Adjustment?

Japan and China versus the United States’ by Ronald I. McKinnon
In this chapter, McKinnon puts forth an analysis that concludes that, under
the world dollar standard, a discrete appreciation by a dollar creditor
country of the United States, such as China or Japan, has no predictable
effect on its trade surplus. Currency appreciation by the creditor country
will slow its economic growth and eventually cause deflation but cannot
compensate for a saving–investment imbalance in the United States. Under
a fixed exchange rate, however, differential adjustment in the rate of growth
of money wages will more accurately reflect international differences in
productivity growth. International competitiveness will be better balanced
between high-growth and low-growth economies, as between Japan and the
US from 1950 to 1971 and China and the US from 1994 to 2005, when the
peripheral country’s dollar exchange rate is fixed so that its wage growth
better reflects its higher productivity growth. Also discussed is the qualified
case for China moving toward greater flexibility in the form of a very
narrow band for the yuan/dollar exchange rate, as a way of decentralizing
foreign exchange transacting.
One important implication of the McKinnon hypothesis as articulated
in this chapter is that even if we made some desirable changes in the
yuan/dollar exchange rate, the hope of narrowing the China–US trade
surplus may not materialize since the more important determinant of the
Chinese advantage is its relatively lower wage rate. According to available
data, anecdotal evidence indicates that it may be as high as 20 to 30:1 in
China’s favor compared with about 5 to 10:1 in India’s favor when we
compare India versus China. However, it is interesting to note that China’s
relative advantage in wage competitiveness may already be narrowing and
the dynamic implications of this trend towards the exchange rate and trade
policy vis-à-vis China should be very instructive and important to watch.
Chapter 9: ‘Adjustment to China’s CPI-based Inflation Rate to Account for

the “True” Cost of Living, 1993–2004’ by Lawrence R. Klein, Huiqing Gao
and Liping Tao
In this chapter, the authors Klein, Gao and Tao (KGT), in a pioneering
study, adjust China’s Consumer price index (CPI) to account for the ‘true’
cost of living for 1993–2004, a period that covers the Asian Financial Crisis
period 1997–98.
Chinese economic growth since 1978 has certainly been spectacular in

magnitude, and, as a result, many scholars have drawn the conclusions that
the numerical size of Chinese economic growth is overstated. However, that
is not the opinion of the authors of this chapter. In order to examine
China’s growth rate, they have focused on its rate of price level change as
measured by a consumer price index – the inflation rate that can be used to
convert nominal GDP to real GDP. They argue that the magnitude of
inflation has been overstated, because it does not take into account quality
change or lifestyle change. Hence, they estimate an adjustment to China’s
CPI by estimating the ‘true’ economic cost of living index, using the linear
expenditure system (LES). The true ‘cost’ of living in China can indicate
how much the inflation might be lowered, and the real growth rate corres-
pondingly increased, for the economy as a whole.
The LES equation can be written as:
冢冣兺
n n
(pitqit ⫺ pit␥i ) ⫽ ␤i rt ⫺ 兺
j⫽1
pjt␥j ;
i⫽1
␤i ⫽ 1
↕ ↕ ↕ ↕
expenditure on minimum total total minimum ␤i ⫽ marginal
the i-th category subsistence income subsistence propensity to
expenditure expenditure consume
on the i-th on all categories out of
category supernumerary
income.
This equation is expressed in current value prices (nominal). Goods and

services are grouped into eight classes: food; clothing; household facilities;
medicine and medical services; transportation, post and communication
services; educational, cultural and recreation services; residential; and mis-
cellaneous commodities and services. Each of these aggregate groups can
be treated as a non-inferior good. Engel curves separately relate expendi-
ture on each group to total income, and the part of the Engel curves that
the authors of this chapter use for parameter estimation is in a linear range.
They use cross-section data from family budgets (for both urban and rural
populations) to estimate the parameters ␥i, which represent necessary or
minimum levels of consumption for each grouping of goods and services,
and they use time-series aggregates to estimate the parameters ␤i, which
represent marginal propensities to consume.
Using these parameter estimates, they evaluate the ‘true’ cost of
living index. They find a lower inflation rate across both urban and rural
Introduction 19
populations on the basis of ‘true’ cost of living, compared with standard

CPI calculations, which translates into higher growth rates for real con-
sumption. If nominal consumption rates were adjusted using this lower
inflation rate, KGT estimate a higher real consumption growth of 9.32
percent on a per capita basis and 9.57 percent on a household basis (versus
7.65 percent and 6.65 percent, respectively without such an adjustment).
Because the authors see persistent overestimates of the price deflator for
consumption, they conclude that one or two percentage points should be
added to the growth rate. Although they determine this for consumption,
they argue that close to the same order of magnitude should be considered
for GDP as well.
In addition, the authors offer another approach to adjusting the official
published price indexes, which takes into account a variety of indicators of
life quality or lifestyle in a wider sense. Using ten such indicators, they
compute an index as a ten-element weighted average, which grows from 100
in 1980 to 115.47 in 2002. This indicates an estimated growth rate of r ⫽
0.65596 percent, nearly double that for the United States in the period
1980–2000. Thus they conclude that the adjustment to China’s growth rate
could be as high as 1 percent or more.
The adjustment of the US inflation rate for quality of life and other
changes has been broadly accepted, so it seems reasonable to adjust China’s
price indexes for the same reasons. In fact, lifestyle changes in China have
been much greater than for the United States.
Chapter 10: ‘Estimating China’s Core Inflation Rate’ by Deming Wu
The premise of this chapter by Deming Wu is to ascertain the ‘core’

inflation rate for China in the same manner as is routinely adopted for other
advanced economies such as the United States. This constitutes a very
sound and extremely interesting methodological exercise with important
policy implications as well. The empirical estimates obtained as a result of
this effort are a very important statistic that allows us to characterize prop-
erly the stabilization policy in China. This is an important element both
domestically as well as in the context of the debate pertaining to China’s
perceived role in the Asian Crisis of 1997–98 when, at times, it was criticized
for being too ‘deflationary’. One of the main findings of the chapter is that
China’s core inflation rate – an idea familiar to the US macro-empiricists –
was positive unlike the observed official overall inflation rate, which was
negative. This implies that, judged by the same standards as adopted, say,
in the US for determining the inherent or underlying true inflation rate by
looking at the ‘core’ inflation rate, China pursued a rather expansionary
monetary policy during 1997 and 1998. Thus, if anything, China was a
stabilizing influence in East Asia during the crisis period. It is evident that
this chapter makes very important contributions methodologically as well
from a policy perspective in the context of the debate on the Asian
Financial Crisis of 1997–98.
NOTE
1. Kaminsky et al. (1998).
REFERENCES
Kaminsky, G.L., S. Lizondo and C.M. Reinhart (1998), ‘Leading Indicators of

Currency Crisis’, IMF Staff Papers, 45 (1), March.
World Bank (2000), The: Workers and the Crisis, Bangkok: Social Monitor 4.
PART I
Analysis of Currency Crises

1. Asia before and after the financial
crisis of 1997–98: a retrospective
essay
Lawrence R. Klein and Tayyeb Shabbir
1 INTRODUCTION
It is common knowledge that the Asian Financial Crisis refers to the onset
and the aftermath of the currency crashes with attendant sharp declines in
output growth and plummeting stock markets in many of the previously
fast-growing countries in East Asia during 1997–98. Amongst the countries
were particularly hit hard were Thailand, Republic of Korea, Indonesia,
Malaysia and the Philippines (often dubbed the ‘Affected Five’). For this
group of countries, the average decline in the 1997 real GDP growth was
about 10 percent from the trend value in 1996, the currency devaluations
ranged from 30 percent to 80 percent and stock markets declined by as
much as 70 percent, which was the case in Indonesia.
Occurring on the heels of past economic achievements of historic pro-
portions, these severe economic shocks were particularly painful as expect-
ations of an ever-increasing living standard had to be sharply and suddenly
pared down in these countries. After all, during the three decades preced-
ing the crisis, these East Asia countries had enjoyed a period of remarkable
economic growth as well as social sector achievements. The decades leading
up to the crisis had witnessed an average real GDP annual growth rate of
more than 7 percent, a decline in poverty from an incidence of six in ten
to two in ten, income per capita increases of up to ten-fold in Korea and
four-fold in Indonesia, Malaysia and Thailand, nearly 100 percent primary
school enrollment, remarkable reductions in infant mortality and increases
in life expectancy. Unfortunately, the Asian Crisis of 1997–98 had suddenly
but surely put all of these past achievements in serious jeopardy. It was
rightly feared that the Asian Crisis will be costly both in economic and soci-
etal terms. However, the majority of the analyses of the costs of this crisis
have focused on its economic or financial aspects while the disruptive social
or societal impact of this episode has been relatively ignored; in fact, one
23
24 Analysis of currency crises
of the three major goals of the present study is to contribute towards rec-
tifying the understanding of this imbalance. The other two goals are to
present a retrospective overview of the nature and efficacy of the econo-
metric predictive models of currency crisis as well as the identification of
the pre- versus post-crisis pattern of macroeconomic indicators of ‘vulner-
ability’ for the ‘Affected Five’ countries in order to note the appropriate
lessons that may be learned.
Three major goals of this chapter are to:
1. Review the different theories that motivate the specification of econo-

metric predictive models of currency crises. Given the relatively high
cost of these financial crises, being able to assess their relative pre-
dictability can be a very worthwhile exercise.
2. Identify the macroeconomic indicators that mark the ‘vulnerability’ of
a particular economy to a financial crisis. This can be very useful in
order to learn the appropriate lessons from the historical patterns and
thus, hopefully, allow the policy-makers to devise appropriate policies
to minimize the chances of a crisis even when precise predictions may
not be forthcoming.
3. Focus on the impact of the 1997–98 Asian Crisis on the social
sectors and poverty reduction efforts of the ‘Affected Five’ countries,
in order to provide an illustration of the negative societal impact that
such financial crises often inflict in addition to the customary analysis
of economic loss.
The chapter is arranged in the following manner. Section 2 presents an

overview of the currency crisis theories and econometric predictive models
that have been inspired by them; section 3 presents a discussion based
on the pre- versus post-crisis patterns of certain macroeconomic indica-
tors in the case of the East Asian countries affected by the 1997–98 crisis;
section 4 examines the statistical relationship between financial crisis,
poverty alleviation and income distribution, again, using East Asia
as the basis of this case study and, finally, section 5 contains concluding
comments.
2 CURRENCY CRISIS THEORIES1
The theoretical literature on currency crises can be categorized into three

‘generations’, each delineated by the underlying mechanism that brings
about the crisis. The first generation focuses on ‘fundamentals’, and empha-
sizes the role of unsustainable government policies that are incompatible
Asia before and after the financial crisis of 1997–98 25
with a pegged exchange rate, and which eventually lead to its collapse. The
canonical model here was presented in Krugman (1979), based on a gov-
ernment that is financing persistent budget deficits through monetization.
This simplest of models suggests that in the period leading up to a specula-
tive attack, one should notice a gradual decline in reserves. It also suggests
that budget deficits and excessive growth in domestic credit may be poten-
tial early warning indicators for speculative attacks. Extensions of the basic
model look at other factors that may force the government to abandon the
peg. For example, expansionary policy may lead directly to a worsening of
the current account through a rise in import demand; the same result may
occur indirectly through a rise in the price of non-tradables and the subse-
quent overvaluation of the real exchange rate.
Thus, the behavior of external sector variables such as the trade balance,
the current account deficit and the real exchange rate may provide some
warning regarding the vulnerability of a country to a speculative attack.
Further, other potentially useful indicators in this respect are possible meas-
ures of the ‘quality of existing debt’, say, the proportion of non-performing
loans of the banking system.
Second-generation models were motivated by episodes such as the
exchange rate mechanism (ERM) crisis in 1992–93, where some of the
countries did not seem to possess the characteristics described in first-
generation models. This led researchers such as Obstfeld (1994) to enrich
existing theories of currency crises. The key element in second-generation
models is recognition that there are both benefits and costs to maintaining
a peg, and that market participants’ beliefs over whether a peg will not hold
or can affect a government’s cost of defending it. The circularity inherent
in second-generation models – that government policy is affected by expect-
ations, and expectations are affected by government policy – leads to the
possibility of multiple equilibria and self-fulfilling crises. These second-
generation models suggest that anything that affects a government’s
decision whether to maintain a peg or not – because of unemployment,
inflation, the amount and composition of debt, financial sector stability
and so on – might contain information on the likelihood of a crisis occur-
ring. Also, evidently, the inflexibility of the ERM system prevented poten-
tially helpful policy actions. This crisis brought to the forefront the
potential for conflict between a member country’s goals and the restrictions
of the system. On the other hand, we saw later in 1997–98, Hong Kong,
with its very strict parity with US dollar, let interest rates go extremely high,
but they had big reserves when attacked by the speculative hedge funds and
could withstand these shocks.
The third generation of currency crisis models focuses on the issue of
contagion, or why the occurrence of a crisis in one country seems to affect
the likelihood of a crisis occurring in other countries. Masson (1998)

suggests three possible reasons why crises seem to come in clusters. First,
there may actually be common external shocks (for example, fluctuations
in world interest rates) that affect all the countries involved. Second, there
may be spillover effects from one country to another due to trade competi-
tiveness effects or portfolio rebalancing effects. Lastly, speculative attacks
might spread from country to country merely on the basis of market senti-
ment or herding behavior. This class of models suggests that early warning
systems should include external variables such as LIBOR (London
Interbank Offered Rate) dynamics or the economic growth rate of trading
partners that indicates a critical trouble in a trading partner economy or in
‘similar countries’ that should be taken into account.
2.1 Early Warning Systems
Recent efforts such as those of Demirgüç-Kunt and Detragiache (1998);

Eichengreen and Rose (1998); Kaminsky and Reinhart (1998, 1999);
Kaminsky et al. (1998); Kumar et al. (2003); and Pérez (2005) towards
devising an early warning system for an impending financial crisis have
taken two related forms.
The first approach estimates a probit or logit model of the occurrence of
a crisis with lagged values of early warning indicators as explanatory vari-
ables. This approach requires the construction of a crisis dummy variable
that serves as the endogenous variable in the probit or logit regression.
Classification of each sample time-point as being in crisis or not depends
on whether or not a specific index of vulnerability or speculative pressure
exceeds a chosen threshold. For currency crises, the index of vulnerability
is often based on a weighted average of the following three variables:
● percentage changes in nominal exchange rates;

● percentage change in gross international reserves;
● difference in local and key foreign short-term interest rates.
Other possible determinants of relevance include current account balance,

inflation rate and domestic fiscal balance. Explanatory variables typically
come from the real, financial, external and fiscal sectors of the economy,
but, in one way or another, often involve international economic relations.
The second method uses a signaling or leading-indicator approach to get
a more direct measure of the importance of each potential explanatory vari-
able. Apart from constructing a crisis dummy variable, the approach also
constructs binary variables from each explanatory variable – thus imputing
a value of 1 (a ‘signal’) or 0 (no signal) for each explanatory variable at each
point in time in the sample – based on whether each variable exceeds a

chosen threshold or not. These signals are classified according to their
ability to call a crisis: a signal is a ‘good signal’ if a crisis does occur within
a specified period (usually 24 months), and is a ‘false signal’ otherwise.
Thresholds are chosen so as to strike a balance between the risk of having
many false signals and the risk of missing many crises. More precisely, a
signal-to-noise ratio is computed for each explanatory variable over the
sample period – as a quantitative assessment of the value of the variable as
a crisis indicator. This is done by classifying each observation of the binary
signaling variable into one of the following four categories:
Crisis occurs within No crisis occurs within

24 months 24 months
Signal A B
No signal C D
Thresholds are then chosen to maximize the signal-to-noise ratio [A/

(A⫹C)]/[B/(B⫹D)]. Based on this metric, Kaminsky et al. (1998) find that the
best early warning indicators for currency crises include exports, deviations
of the real exchange rate from trend, M2/reserves, output and equity prices.
Whether this approach or a probit/logit approach is used, to cover a wider
sample for purposes of estimating incidence probabilities for a crisis, many
studies have been typically made of a panel of countries – and usually with
some homogeneity assumptions about crisis behavior across countries.
2.1.1 Drawbacks in the above approaches

Despite the current popularity of these approaches,2 they have some draw-
backs, which include the following (see Abiad, 2002 and Mariano et al.,
2002 for a more detailed discussion):
● There may be an inadequate treatment of serial correlations inherent

in the dynamics of a crisis. Neither the probit/logit approach nor the
signaling approach give us information on dynamics, showing how
long crisis periods tend to last – nor do they give information about
what variables affect the likelihood of the end of a crisis period.
● Artificial serial correlations may even be introduced inadvertently
through the explicit manner in which the crisis dummy variable is
constructed. Many previous studies focus primarily on predicting the
onset of a crisis, that is, the first period when speculative attacks
occur. To achieve this in their binary crisis variable, they make use of
so-called ‘exclusion windows’, which remove any crisis signals that
closely follow previous crisis signals. This procedure can introduce

artificial serial correlations; see Abiad (2002).
● Classification errors may result when constructing the crisis dummy
variable (either as a false signal of a crisis, or a missed reading of a
crisis). Because the threshold used to delineate crisis periods from
tranquil periods is arbitrarily chosen, misclassification of crisis
episodes can occur. If the threshold is too high, for example, some
periods of vulnerability may not be detected.
● There needs to be a better framework for significance testing of the
influence of explanatory or indicator variables (in the signaling
approach). The signaling approach could be improved if it were
based on an explicit stochastic model. As with many indicator
models, it needs formal testing in order to assess its performance
vis-à-vis other approaches.
● There may be possible inconsistencies in the estimation of crisis inci-
dence probabilities because of heterogeneity across countries. It is
possible that the variables that are important in determining crisis
likelihoods for one country are unimportant for another country.
And even if the same variables affected crisis likelihoods for all coun-
tries, the degree of likelihood of a crisis occurring may differ from
one country to the next.
2.2 Markov-switching Approach3
Recently, a newer approach to the early warning methodology tries to

address these drawbacks. This approach constructs for each individ-
ual country a Markov-switching, autoregressive model that allows inter-
cepts, lag coefficients and error variances to switch stochastically over
time according to the value taken by a latent Markov chain, describing the
vulnerability of the country’s currency to speculative attacks (Mariano
et al., 2003). In a related work, Martinez-Peria (1999), applies a Markov-
switching model to speculative attacks, but the primary purpose in that
paper is to evaluate the ability of the model to date crisis periods and to see
whether market expectations affect crisis probabilities. Here the focus is on
the use of the model as an early warning system. The Mariano et al. (2003)
model is a predictive model of a currency crisis and consists of two parts.
First, a Markov chain model of the unobservable financial vulnerability
of the country, say St. It argues that what we observe are indicators of this
latent attribute of the country. It assumes two states:
● normal (St ⫽0);

● vulnerable (St ⫽1).
This model further assumes that the Markov chain is of order 1, with tran-
sition probabilities that are time-varying through dependence on observ-
able indicator variables. In Mariano et al. (2002, 2003) there are reports of
results that are based on the following indicator variables:
● deviations of real effective exchange rate from trend (REERDEV);

● month-to-month percentage changes in the ratio of M2 to inter-
national reserves (M2RATIO);
● month-to-month percentage changes in real domestic credit (DCR).
Second, a Markov regime-switching time-series model of percentage

changes in nominal exchange rates. This model differs from standard cases
in the sense that it includes the unobservable state variable St as an add-
itional endogenous variable. With the inclusion of St, we introduce the
notion that the exchange rate dynamics behave in a different fashion,
depending on whether financial conditions are normal (St ⫽0) or vulner-
able to currency pressures (St ⫽1). The Mariano et al. (2002, 2003) model
reflects this by allowing the parameters in the relevant time-series model
to change in value over time, as financial conditions become normal or
vulnerable.
2.2.1 Markov-switching model in some detail

Let St be a two-state Markov chain of order 1 with transition probabilities
pt and qt, so that at any given time t, St can take on two values, 0 or 1,
according to the following probability law:
Pr(St ⫽ 0 | St⫺1 ⫽ 0) ⫽ pt and Pr(St ⫽ 1| St⫺1 ⫽ 1) ⫽ qt.
Further, let
yt ⫽month-to-month percentage changes in nominal exchange rates,

xt ⫽the vector of exogenous variables at time t to be used to explain yt,
zt ⫽the vector of exogenous variables at time t to be used as indicators of
currency vulnerability, which may overlap with xt.
We assume also that the transition probabilities vary over time according
to values of indicator variables in the following manner:
pt ⫽ F(z⬘t ␥) and qt ⫽ F(z⬘t␦),
where F(•) is the standard unit Gaussian cumulative distribution

function.
The second part of the model consists of a univariate linear model for yt:
yt ⫽ ␣st ⫹ xt⬘␤st ⫹ ␴st␧t.
In this model, the model parameters (␣, ␤, ␴) are subscripted by St – indi-

cating that their true (unknown) values are shifting between two sets of
possible parameter values: (␣0, ␤0, ␴0) and (␣1, ␤1, ␴1) depending on
whether financial conditions are normal or vulnerable.
The estimation procedure we use is direct maximization of the likeli-
hood, where the likelihood function is calculated by using an iterative
process, described in detail in Hamilton (1994). Collect all the parameters
of the model into a single vector ␪⫽(␣0, ␤0, ␴0, ␣1, ␤1, ␴1, ␥⬘, ␦⬘). Using
information available up to time t, ⍀t, we can calculate for each time t
(using the iteration below) the value of Pr(st ⫽ j | ⍀t; ␪) , the conditional
probability that the t-th observation was generated by regime j, for j⫽1, 2,
. . . N, where N is the number of states (in this chapter, N⫽2). We will stack
these conditional probabilities into an (N ⫻ 1) vector ␰ˆ t|t.
Using the same iteration, we can also form forecasts regarding the con-
ditional probability of being in regime j at time t ⫹1, given information
up to time t: Pr(st⫹1 ⫽ j | ⍀t; ␪) , for j ⫽1, 2, . . . N. Collect these forecast
probabilities in an (N ⫻ 1) vector ␰ˆtⴙ1|t. Lastly, let ␩t denote the (N ⫻ 1)
vector whose j-th element is the density of yt conditional on st.
The optimal inference and forecast for each date t can then be found by
iterating on the following equations:
(␰t | t⫺1 °␩t )

␰ˆ t|t ⫽
1⬘(␰t | t⫺1 °␩t )
␰ˆtⴙ1|t ⫽ Pt⫹1 · ␰t | t
where Pt is the (N ⫻ N) transition probability matrix going from period

t ⫺ 1 to period t (for the two-state model in this chapter; Pt is the 2 ⫻ 2
matrix [pt 1 – pt; 1 – qt qt] ), and ⬚ denotes element-by-element multiplica-
tion. Given an assumed value for the parameters, ␪, and an assumed start-
ing value for ␰ˆ 1|0 (the unconditional probability of st at t⫽1), we can then
iterate on the above equations to obtain values of ␰ˆ t|t and ␰ˆtⴙ1|t for t ⫽1, 2,
. . . T. The log likelihood function L(␪) can be computed from these as
T
L(␪) ⫽ 兺log f(yt |Xt,Yt⫺1; ␪)
t⫽1
where
f(yt |Xt,Yt⫺1; ␪) ⫽ 1⬘ (␰ˆ t|t °␩t ).
One can then evaluate this at different values of ␪ to find the maximum like-
lihood estimate.
2.2.2 Empirical results for Markov-switching models of currency crises

The Markov-switching predictive models of currency crises of the type out-
lined above have recently been estimated for Argentina (Alvarez-Plata and
Schrooten, 2003), Turkey (Mariano et al., 2004) as well as countries in
Southeast Asia4 (Brunetti et al., 2003 and Mariano et al., 2003).
In the case of the Argentinean crisis of 2002, Alvarez-Plata and Schrooten
(2003) estimate a two-regime Markov-switching model with constant (exoge-
nous) transition probabilities. They find evidence of regime switch prior to
the crisis; however, they merely assume the cause of this shift to be changes
in ‘expectations of private investors’. This ignores the (testable) alternative
possibility of structural determinants of transition probabilities. On the
other hand, Mariano et al. (2004) for Turkey, and Mariano et al. (2003) for
Southeast Asian countries, estimate a Markov-switching model with con-
stant as well as time-varying transition probabilities.
For the Southeast Asian countries of Indonesia, Malaysia, the
Philippines and Thailand, Mariano et al. (2003) report that the transition
probabilities are affected by the real effective exchange rate, money supply
relative to international reserves and real domestic credit. However, the esti-
mated model does not perform uniformly well across the sample countries.
Whereas for some currency depreciation episodes like those in 1981, 1984
and 1997 for Thailand and in 1997 for Malaysia, the model provides strong
early warning signals, but for other episodes, such as in 1997 for Indonesia
and the Philippines, the model manages at best to generate only weak
warning signals. Incidentally, a related study, Brunetti et al. (2003), extended
the Mariano et al. (2003) model to account for conditional heteroskedastic-
ity of the exchange rate depreciations – the volatility is hypothesized not
only to switch between two regimes but to follow a GARCH process within
each regime. The preliminary results reported in Brunetti et al. (2003) claim
improved forecast performance. Further, this study also notes that stock
market returns are an additional determinant that may affect transition
probabilities in addition to those reported in Mariano et al. (2003).
The formal econometric predictive models like the Markov-switching
model discussed above are very useful tools. While their formal structure is
an extremely desirable feature, due to technical limitations, they can also
restrict the scope of the discussion necessary to get a broad sense of the
possible determinants of the vulnerabilities of economies to a currency

crisis. Thus, in the following section, we present an analysis based on
certain observable important indicators of vulnerability of an economy to
a currency or a financial crisis in general.5
3 VULNERABILITIES OF THE ‘AFFECTED FIVE’
3.1 Competing Hypotheses as Explanations of the Crisis6
The literature that has been inspired by the East Asian Crisis of 1997–98
proffers two competing hypotheses as the possible causes of currency
crisis – the ‘contagion’ hypothesis versus the ‘vulnerability’ hypothesis.7
According to the ‘contagion’ view, the capital flight in Thailand induced
by expectations of an impending devaluation of the bhat relative to the US
dollar (to which it was pegged), had negative informational ‘spillover’
effects that doomed some of the neighbors by casting doubt in the minds
of the investors about otherwise ‘healthy’ economies.
The ‘vulnerability’ hypothesis, on the other hand, maintains that some
economies were inherently vulnerable to a crisis because of relatively long-
term deterioration in economic fundamentals. This predisposed them to
crisis when faced with shocks that may lead to expectation of exchange rate
devaluation.
The above set of rival explanations can be distinguished such that the
‘contagion’ factor is analogous to a ‘trigger’ and ‘vulnerability’ means sus-
ceptibility to a country crisis on fundamental grounds.
Of course, in their respective extreme forms, the contagion and vulner-
ability hypothesis afford significantly different interpretations as well as
policy implications, whereas the vulnerability hypothesis implies that the rel-
ative susceptibility of an economy can be ascertained with the aid of observ-
able indicators. Also, high levels of vulnerability make a crisis inevitable and
any random event could actually provoke it. On the other hand, extreme ver-
sions of the contagion hypothesis deny or certainly minimize the significance
of observable indicators of vulnerability. Most of the literature dealing with
the East Asian Crisis of 1997–98 has taken an intermediate stance.
After all, the pure contagion explanation has had to contend with the
question that if Thailand spread the virus, how did Thailand become
affected in the first place? Evidently, it is hard to escape having to rely on
an explanation that is based on deteriorating fundamentals in Thailand
before 1997. Again contagion theory fails to explain why some countries
such as Singapore, Taiwan and China proved to be resilient to the spread-
ing ‘virus’.
We feel that the intermediate position is the most tenable one and thus
examining the determinants of vulnerability in terms of critical real as well
as financial sector indicators is a very useful exercise. In the analysis that
follows, we examine the pre- as well as post-crisis pattern of certain critical
macroeconomic fundamentals to provide an assessment of the dynamics of
the economies of the ‘Affected Five’ before and after the ‘breakpoint’
marked by the 1997–98 crisis.8
3.2 Examining the Pre-crisis and Post-crisis Patterns of ‘Fundamentals’
As evident from the data trends in current account deficit, export growth
rate, government budget fiscal balance and other pertinent macroeconomic
variables such as the real exchange rate overvaluation, the ‘Affected Five’
countries (Thailand, Republic of Korea, Indonesia, Malaysia and the
Philippines) were vulnerable before the outbreak of the currency crisis of
the baht on 2 July 1997. The problem is particularly telling if one looks at
the increase in the current account deficit and the slowdown in the export
growth rate. This, accompanied by the fragility of the financial sector, in
particular, the upsurge in the proportion of non-performing loans, con-
tributed greatly to the currency and, more broadly, the financial crisis that
followed. Tables 1.1–1.5 (and the corresponding Figures 1.1–1.5) below
present the pertinent data for the pre- as well as post-crisis period
(1993–2004) and additional comments for each of the ‘Affected Five’ coun-
tries separately.9 On the other hand, Figure 1.6 through Figure 1.11 present
the comparative data for all of the ‘Affected Five’ countries for several
important indicators considered one at a time.
As can be seen from Table 1.1 and Figure 1.1, before 1997, Thailand’s
CA (current account) balance (as a percentage of GDP) had deteriorated
sharply from –5.41 percent in 1994 to –7.89 percent in 1996, while its export
growth had declined precipitously from an average growth of 20.11 percent
during 1993–95 to a decline of –1.91 percent in 1996, and the fiscal balance
(as a percentage of GDP) dropped significantly from 3.01 percent in 1995
to 0.94 percent in 1996. These indicators clearly pointed to Thailand’s
serious vulnerability to a crisis, which, of course, unfolded on 2 July 1997.
In the years following the onset of the crisis in 1997, while Thailand had
started showing definite signs of turnaround in its external sector indicators
as early as 1998 (the CA deficit had turned positive by then), this recovery
was not across the board – in fact, the GDP growth rate registered a steep
decline in 1998. Indeed, it was not until 1999, that a firmer and more broad-
based recovery milestone was reached when, in addition to a sustained CA
surplus, the GDP growth rate turned significantly positive, and both
the growth rates of exports as well as imports were positive. Thus, for all
Table 1.1 Macro indicators: Thailand
Year 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
CA (% of GDP) ⫺4.90 ⫺5.41 ⫺7.88 ⫺7.89 ⫺1.97 12.66 10.17 7.61 5.41 6.06 5.57 4.39
Export Growth* (%) 13.36 22.15 24.82 ⫺1.91 3.76 ⫺6.78 7.42 19.52 ⫺6.92 5.71 18.18 21.56
Import Growth* (%) 12.35 18.44 31.85 0.62 ⫺13.37 ⫺33.75 16.94 31.34 ⫺2.82 4.45 17.35 26.04
Fiscal Balance 1.89 2.69 3.01 0.94 ⫺1.50 ⫺2.79 ⫺3.33 ⫺2.23 ⫺2.40 ⫺1.41 0.40 0.13
(% of GDP)
34
Real GDP Growth (%) 8.99 9.24 5.90 ⫺1.37 ⫺10.51 4.45 4.76 2.14 5.41 7.03 6.17
REER (1996 ⫽100) 90.78 92.28 91.54 100.00 93.71 82.39 85.61 79.56 78.11 82.16 80.49 79.12
Inflation Rate (%) 3.31 5.08 5.79 5.83 5.60 8.08 0.30 1.55 1.66 0.60 1.81 2.77
Notes:
Def.: REER⫽real effective exchange rate index (1996⫽100).
* Based on [{Index(t)/Index (t⫺1)}⫺1]⫻100 where Index value for 1996⫽100.
Source: http://aric.adb.org.
30.00
25.00
20.00
15.00
10.00
Percent
5.00
0.00
35
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
–5.00
–10.00
–15.00
CA (% of GDP) Export Growth (%) Fiscal Balance (% of GDP)

Real GDP Growth (%) Inflation Rate (%)
Figure 1.1 Macro indicators: Thailand

practical purposes, 1999 is the year that marks the onset of the post-crisis
recovery for Thailand even though fiscal balance took a little longer to
recover and started to improve only in 2000.
The Republic of Korea (Table 1.2 and Figure 1.2), in a scenario similar
to Thailand’s, exhibited plenty of pre-crisis signs of vulnerability – its CA
(as a percentage of GDP) turned to a deep deficit in 1996 (–4.41 percent)
from a slight surplus of 0.29 percent in 1993, while its export growth slowed
down significantly from an average of 18.10 percent p. a. during 1993–95
to a relative trickle in 1996 (3.72 percent) and its real GDP growth that
averaged 8.59 percent p. a. for 1994–95, registered at only 6.75 percent in
1996.
While the crisis meant a sharp decline for Korea’s real GDP growth rate
(–6.69 percent in 1998), as a whole, it was relatively the least hampered by
the crisis amongst the group of ‘Affected Five’ countries. In 1999, Korea’s
real GDP growth rate recovered vigorously (10.89 percent), its fiscal deficit
started declining as well, and, in fact, turned to a surplus by 2000, which
has proven sustainable since. Again, following a decline of 2.83 percent in
1998, the export growth rate resumed a positive trend with the exception of
a drop in 2001 on account of the recession in the US, a major importer of
Korean manufactured goods (average export growth of 17.35 percent p. a.
during 1999–2004, excluding 2001, which compares favorably with the
average of 18.10 percent p. a. for the pre-crisis ‘miracle’ years of 1993–95.
The country’s current account balance also turned positive starting in 1998
though this happened primarily because of a relatively much steeper drop
in its trend import growth, which had dropped to –35.50 percent in 1998;
import growth did in fact turn positive by the following year and still the
CA balance stayed positive.
In general, ex post, the trajectory of Korea’s macroeconomic recovery
from the crisis was a ‘V-shaped’ rather than a ‘U-shaped’ one – that is, quick
and sharp rather than dull and drawn out. (See Figure 1.6 in this chapter
for comparative changes in real GDP growth rate for the ‘Affected Five’
countries.)
In addition, in the aftermath of the crisis, Korea has had relatively the
most to show for its efforts to clean up its financial sector both in terms of
the reform of the corporate ownership structure and decreased frequency
of the non-performing loans as a proportion of the total debt of the
banking system (though no statistics have been noted in the tables in this
regard).
Table 1.3 and the corresponding Figure 1.3 look at the case of Indonesia
whose economy, like that of Thailand and Korea, exhibited pre-crisis
increase in external sector imbalance (CA deficit increased from –1.45
percent of GDP to –3.41 percent of GDP). However, unlike the Thai or
Table 1.2 Macro indicators: Republic of Korea
Year 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
CA (% of GDP) 0.29 ⫺0.96 ⫺1.73 ⫺4.41 ⫺1.70 12.58 6.02 2.65 1.93 1.27 1.96 4.05
Export Growth* (%) 7.31 16.75 30.25 3.72 4.97 ⫺2.83 8.59 19.89 ⫺12.67 8.00 19.29 30.97
Import Growth* (%) 2.48 22.13 32.02 11.26 ⫺3.81 ⫺35.50 28.38 34.01 ⫺12.08 7.82 17.55 25.52
Fiscal Balance 0.33 0.26 ⫺1.54 ⫺4.22 ⫺2.71 1.25 1.32 3.80 1.12 0.72
(% of GDP)
37
Real GDP Growth (%) 5.49 8.25 8.92 6.75 5.01 ⫺6.69 10.89 9.33 3.10 6.35 3.10 4.64
REER (1996 ⫽100) 91.93 94.13 95.67 100.00 95.98 79.50 86.89 90.03 87.56 93.71 95.52 97.13
Inflation Rate (%) 4.79 6.27 4.45 4.92 4.44 7.52 0.81 2.26 4.31 2.77 3.51 3.59
Notes:
35.00
30.00
25.00
20.00
15.00
Percent
10.00
5.00
38
0.00
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
–5.00
–10.00
–15.00
CA (% of GDP) Export Growth (%)

Fiscal Balance (% of GDP) Real GDP Growth (%)
Inflation Rate (%)
Figure 1.2 Macro indicators: Republic of Korea

Table 1.3 Macro indicators: Indonesia
Year 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
CA (% of GDP) ⫺1.45 ⫺1.67 ⫺3.34 ⫺3.41 ⫺2.22 4.27 4.11 5.3 4.87 4.3 3.40 1.21
Export Growth* (%) 7.88 9.31 13.39 9.68 7.28 ⫺8.6 ⫺0.37 27.66 ⫺9.34 1.49 6.82 17.24
Import Growth* (%) 3.71 12.95 27.03 5.66 ⫺2.91 ⫺34.41 ⫺11.2 38.06 ⫺7.62 1.06 4.03 42.93
Fiscal Balance 0.61 0.94 2.22 1.16 ⫺0.67 ⫺2.95 ⫺1.15 ⫺1.19 ⫺3.77 ⫺1.76 ⫺1.65 ⫺1.24
(% of GDP)
39
Real GDP Growth (%) 7.54 8.22 7.82 4.7 ⫺13.13 0.79 4.92 3.44 3.66 4.88 5.13
REER (1996⫽100) 93.02 93.63 91.26 100.00 95.46 48.07 68.06 64.19 63.44 77.92 83.39 78.48
Inflation Rate (%) 9.69 8.52 9.43 7.97 6.14 56.15 19.96 3.93 11.54 11.84 6.75 6.08
Notes:
60
50
40
30
Percent
20
10
40
0
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
–10
–20

Inflation Rate (%)
Figure 1.3 Macro indicators: Indonesia

Korean cases, Indonesia showed a far less pronounced drop in export

growth, partly perhaps because Indonesia has ‘less to lose’. However, while
not being in the vanguard of the ‘New Asian Miracle’, Indonesia had
started to show real promise when the Asian Crisis struck in 1997. Thus
Indonesia was affected significantly as evidenced from the fact that it
suffered a negative real GDP growth rate of 13.13 percent for 1997 as com-
pared with an average growth rate of 7–8 percent p. a. for the early to mid-
1990s period.
After the crisis, Indonesia recovered to essentially a relatively low tra-
jectory of GDP growth rate in the 3–4 percent p. a. range. Notice that
unlike others in the group of ‘Five Affected’ countries, import growth in
Indonesia was negative as far into the future as the year 2001 and its post-
crisis fiscal deficit continues to persist through to 2004, the end of our
sample period. Also, Indonesia’s post-crisis economic recovery was accom-
panied by relatively higher inflation rates as compared with the other coun-
tries in the same cohort. Indonesia was also a victim of relatively greater
amount of political and social turmoil in the aftermath of the Asian
Financial Crisis.
Table 1.4 and the corresponding Figure 1.4 present the facts for the
Malaysian case. In terms of the possible indicators of vulnerability,
Malaysia’s export growth dropped to a relatively anemic 5.93 percent in
1996 from an average of 20.10 percent p. a. during the 1993–95 period. Its
import growth slowed down considerably as well, essentially stagnating in
1996 compared with an average growth rate of 25.46 percent p. a. during
1993–95. Regarding the country’s CA balance, a somewhat classic indicator
of crisis vulnerability, Malaysia’s case indicated an overall trend of rela-
tively worsening CA deficits that increased during 1994 and 1995 compared
with 1993. (However, the year 1996 showed a substantial CA deficit reduc-
tion compared with the preceding year.)
After the 1997–98 crisis, Malaysia’s GDP growth rate was fairly quick to
recover – the country experienced a negative GDP growth rate in only one
year, 1997, and recovered smartly the following year. Thus, Malaysia resem-
bles Korea in this respect. However, what has been unique about Malaysia’s
policy reaction to the Asian Crisis was its institution of capital controls –
a policy that was at odds with the standard IMF prescription for other
countries in East Asia in similar straits at the time. The Malaysian capital
control program was looked at with a fair amount of skepticism when it
was put in place. After all, Malaysia did not enjoy the huge foreign
exchange reserves that China and Hong Kong possessed, to back up their
retention of fixed dollar pegs. However, in the end, the steady and sustained
Malaysian macroeconomic recovery won over several of the skeptics. Thus,
in the aftermath of the Asian Crisis, Malaysian GDP recovered fairly
Table 1.4 Macro indicators: Malaysia
Year 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
CA (% of GDP) ⫺4.47 ⫺6.05 ⫺9.78 ⫺4.8 ⫺5.18 13.53 15.92 9.14 8.27 7.82 12.82 12.57
Export Growth* (%) 15.72 24.74 25.85 5.93 0.41 ⫺6.75 15.34 16.08 ⫺10.45 5.92 11.60 20.52
Import Growth* (%) 14.32 30.55 31.52 0.1 0.68 ⫺26.14 12.08 25.35 ⫺10.03 7.98 4.84 25.92
Fiscal Balance 0.21 2.26 0.84 0.72 2.35 ⫺1.77 ⫺3.15 ⫺5.75 ⫺5.51 ⫺5.62 ⫺5.30 ⫺4.32
(% of GDP)
42
Real GDP Growth (%) 9.9 9.21 9.83 10.00 7.32 ⫺7.36 6.14 8.55 0.32 4.12 5.42 7.14
REER (1996 ⫽100) 93.42 92.53 93.93 100.00 99.08 81.53 82.33 80.94 87.28 89.49 84.79 78.48
Inflation Rate (%) 3.56 4.94 4.06 3.49 2.67 5.27 2.74 1.52 1.41 1.83 1.09 1.42
Notes:
30
25
20
15
10
Percent
0
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
43
–5
–10
–15

Inflation Rate (%)
Figure 1.4 Macro indicators: Malaysia

quickly while maintaining a low and stable inflation rate, though fiscal bal-
ances were continuing to worsen well into the 1990s and even early in the
new decade.10
Finally, referring to Table 1.5 and the corresponding Figure 1.5, we turn
to a discussion about the Philippines – a country that had just started to
show promise when the Asian Crisis hit in 1997. In fact, relative to its
ASEAN (Association of Southeast Asian Nations) partners, the Philippines
were relatively more vulnerable – its GDP growth rate had averaged only
about half of what Korea had been able to achieve in the years leading up
to the Asian Crisis; further, for the period, early to mid-1990s, the
Philippines suffered from a current account deficit in the range of 4–5
percent of GDP. The country also had its share of political instabilities and
crises and, collectively, these pre-crisis vulnerabilities made the recovery
period somewhat anemic. Here it is important to note that the ‘Marcos
Regime’, which lasted for almost two decades from 1965–85, had been
replaced by a relatively democratic institution only for about ten years when
the Asian Crisis occurred. Thus, the Philippines were somewhat ‘late
bloomers’ whose ascent to a sustained higher economic stage was signifi-
cantly interrupted by the crisis in 1997. Still the country appears to have
made a sustained, albeit modest-sized, recovery from the crisis. However,
the future prospects of the country would depend on continued political sta-
bility that the country has enjoyed of late, getting the increasing fiscal deficit
under control and continuing to make progress in terms of the financial
sector and governance-related reforms.
Incidentally, many of the above observations regarding the pre- versus
post-crisis relative performance of the ‘Affected Five’ countries can also be
visualized in an alternative format as presented in the following set of
Figures 1.6–1.11. Here each figure displays information about a given eco-
nomic indicator across countries rather than across economic indicators for
one country at a time as above. In particular, note that Republic of Korea’s
GDP growth rate was relatively the fastest to recover after the crisis (Figure
1.6) and, relatively speaking, Indonesia had to suffer from the highest rate
of inflation in the aftermath of the crisis (Figure 1.11).
3.3 Did We See the Asian Financial Crisis of 1997 Coming?
The above heading poses a particularly interesting question of obvious

importance. There may not have been a siren song of pre-crises warning,
but several researchers made important observations that noted the devel-
opment of some unusual patterns that presaged the crises.
Table 1.5 Macro indicators: The Philippines
Year 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
CA (% of GDP) ⫺5.54 ⫺4.56 ⫺4.49 ⫺4.77 ⫺5.27 2.36 9.18 8.36 1.84 5.39 1.76 2.37
Export Growth* (%) 15.78 18.53 29.36 17.78 22.8 16.93 18.78 8.69 ⫺15.57 9.51 2.91 9.52
Import Growth* (%) 21.2 21.23 24.38 22.22 10.81 ⫺17.46 3.59 12.26 ⫺4.15 7.17 3.14 8.82
Fiscal Balance ⫺1.48 1.07 0.59 0.29 0.06 ⫺1.88 ⫺3.75 ⫺4.06 ⫺4.0 ⫺5.24 ⫺4.66 ⫺3.86
(% of GDP)
45
Real GDP Growth (%) 2.12 4.39 4.68 5.85 5.19 ⫺0.58 3.4 5.97 2.96 4.43 4.62 6.03
REER (1996⫽100) 81.67 87.55 90.58 100.00 100.6 85.14 91.29 82.47 79.9 82.01 75.47 71.48
Inflation Rate (%) 6.88 8.36 8.03 9.01 5.97 9.7 6.68 4.31 6.12 3.1 3.48 5.98
Notes:
35
30
25
20
15
Percent
10
5
0
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
46
–5
–10
–15
–20

Inflation Rate (%)
Figure 1.5 Macro indicators: The Philippines

15.00
10.00
5.00
Percent
0.00
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
47
–5.00
–10.00
–15.00
Thailand Korea Indonesia Malaysia Philippines
Figure 1.6 Relative real GDP growth rates (%) for the Asian ‘Affected Five’
20.00
15.00
10.00
5.00
Percent
0.00
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
48
–5.00
–10.00
–15.00
Figure 1.7 Relative current account (as % GDP)

6.00
4.00
2.00
0.00
Percent
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
–2.00
49
–4.00
–6.00
–8.00
Figure 1.8 Relative fiscal balance (as a % GDP)

40.00
30.00
20.00
Percent
Percent
10.00
50
0.00
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002
–10.00
–20.00
Figure 1.9 Relative export growth rates (%)

50.00
40.00
30.00
20.00
10.00
Percent
0.00
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
51
–10.00
–20.00
–30.00
–40.00
Figure 1.10 Relative import growth rate (%)

60
50
40
Percent
30
52
20
10
0
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
Figure 1.11 Relative inflation rates (%)

For one, Krugman (1994) asserted that since the high GDP growth
rate of the Asian ‘tigers’ was fueled primarily by ‘Soviet-style’ accumu-
lations of machinery, infrastructure, and other factor inputs rather than
‘efficiency’ gains, these growth rates will ‘slow down’. Also, there were
many other discussions, albeit often of a generic nature, about the
possibility of the appearance of the ‘Mexican Crisis’ of December 1994
and early 1995 in other countries. A somewhat specific case of such analy-
sis is Lau (1995) who noted, in an ex ante sense, that Thailand and
Indonesia were the most likely venues for a Mexican-type financial crisis
to occur.
On the other hand, Inada et al. (2000) is a retrospective simulation exer-
cise that concludes that one factor at the heart of the difficulty in East Asia
was the fact that many of these countries had, to a significant extent,
pegged their currencies to the US dollar, which grew relatively very strong,
and many of these countries, such as Thailand, did not act early enough (as
China did in 1994) to devalue their currencies, which would have averted or
significantly ameliorated the impact of the currency crises. Of course, the
Chinese exchange rate action was not in response to a pending crisis, but
was a move to unify their exchange rate policies in January 1994 and hold
the renminbi-dollar rate at constant par from that time forward. However,
the other East Asian countries failed to react to that move by China and
thus lost export trade share accordingly.11 It is interesting to note that in
this new century, China’s growth has been so rapid and intense that they
have required imports from many partner countries, including East Asian
neighbors, and have contributed to favorable macroeconomic performance
over a wide area.
3.4 Asia in the 21st Century – China and India as the New Players for the
‘New Asian Miracle’
Looking ahead to the future developments in Asia, China and India are
emerging as two increasingly important regional and eventually global eco-
nomic powers. They are relatively the fastest-growing (with respective GDP
growth rates of 8–10 percent for China and 6–7 percent p. a. for India in
the last ten years) with dynamic and open economies, at least with respect
to trade flows. These two countries appear fully poised to be the new players
in the ‘New Asian Miracle’, if we may modify a much-worn label. The
important question from our perspective is: how much are these countries
going to benefit from the lessons learned from the recent financial crises
around the world? No doubt, the open economies of these countries will
make them vulnerable to external shocks. They need to be ready with robust
financial institutions, willingness to adopt sound macroeconomic policies

and of course, have a viable social safety net, just in case.
Specifically, China’s case is important from many points of view. China,
of course, has been growing tremendously fast since reform in 1978 and is
already a dominant part of future economic prosperity of the region.
However, its openness in terms of trade flows makes it vulnerable to crises.
In the face of prospective openness of capital account, China has a long-
standing policy of minimal variations in the exchange rate despite inter-
national reserves that are projected to reach $1 trillion in 2007. Thus, China
is at the center of the debate about the appropriate exchange rate and the
capital controls policies to adopt. These are issues that have been an inti-
mate part of the financial crisis experience. China is also an obviously
important player through its trade and direct investment links to the region.
Thus, focusing on China is only natural.
However, we know that some people have expressed a degree of skept-
icism about the veracity of China’s growth statistics. This is an aspect of the
Chinese economy that is of interest to us. Besides addressing the question
of China’s ‘true’ growth rate, we are also interested in exploring what role
China may have played – a stabilizing or a destabilizing one – during the
1997–98 Asian Crisis. The major question, of course, is whether we have
learned enough from the previous crises so that we can more confidently
determine the optimal degree of openness of a country’s capital and
exchange rate markets.
In terms of the specific situation that India finds itself in, there is certainly
a historic change afoot in this country. India is another emerging Asian
‘giant’, side by side with China. Both these countries have enjoyed strong
growth in the real sector (GDP) although they reached high growth in unique
fashion – China’s has been a case of growth led now by exports of manufac-
tured goods, while India’s growth has been fueled by growth of service sector
activity (especially in the information technology [IT], financial and health
sectors). While each of these countries have exhibited impressive rates of
GDP growth in recent years, there are significant financial sector issues such
as stock and bond market reform, greater transparency and better corporate
governance that still need attention. Even though India appears ahead in
terms of financial market efficiency and institution-building for now, much
work remains. Also, China is apparently making financial sector reforms as
its high priority too. In any event, it is clear that going forward, financial
sector and capital market reforms will be necessary for the continued success
of these countries while keeping in mind some of the important ways – polit-
ical and structural – in which these countries differ from each other.
However, the important point is that, as evidently the new players on the
growth playing field, so to speak, these countries’ reliance on open trade and
foreign direct investment flows make them vulnerable to financial crises. It is

useful to watch to see whether lessons from the recent crises were learned
well. During the crisis period, neighbors and other competitors blamed
China for stealing their export markets. Now they enjoy China’s import
appetite. As for India, advanced countries blame ‘white collar’ unemploy-
ment problems on India’s burgeoning service sector, but should be benefiting
from reduced operating costs in the near future.
4 FINANCIAL CRISIS, POVERTY AND INCOME

DISTRIBUTION
What happens to the poor as a result of economic downturns induced

by financial crises? This is an important question, in part, because it is often
the developing countries with their disproportionate share of the poor that
are vulnerable to financial crises. However, relative to studies of the impact
of financial crises in terms of output or financial loss, there are few studies
of the societal cost of such crises in terms of effects on poverty and income
inequality.12 Some of the important questions in this regard are:
● What are the mechanisms through which financial crises affect

poverty and income distribution?
● What are the magnitudes of the above effects?
● What relevant policy implications arise in this context?
Theoretically, financial crises are expected to worsen poverty (or slow down
the rate of poverty alleviation) and worsen inequality for a number of
reasons:
1. Worsening labor market conditions. The financial crises bring in their

wake high-level job loss, lower wage rates, forced transition into infor-
mal sector activities with poorer labor market prospects and increased
job insecurity.
2. Relative price change. The relative price of tradables increases due to
currency depreciation causing decrease in relative wage rates in the
non-tradable sector, which may additionally suffer because deprecia-
tion-induced export growth bolsters employment prospects relatively
in the export sector.
Large currency depreciations also lead to an increase in domestic
inflation rates that hurt the relatively poor who have few assets to fall
back on. To the extent food may be imported this also hurts the relatively
poor who spend proportionately more of their budget on food.
3. Post-crisis tight fiscal and monetary policy. As a result of the crisis,

the governments invariably feel pressure in terms of fiscal budgets.
Often this means that public expenditure on health, education and
the various transfer and social sector programs will be cut back in the
countries beset by a financial crisis, thus hurting the poor. The situation
is worsened as, generally speaking, tight fiscal policies are accompanied
by tight monetary policies undertaken to stem capital flight or other
stabilization goals.
4. Regional and ethnic distress. Financial crises and resulting economic
downturns invariably exacerbate the various political, regional and
ethnic tensions in a country. Regrettably, tolerance is an easy victim
during such periods of turmoil and it is usually the ‘have-nots’ who end
up paying the price of such societal disruption.
Baldacci et al. (2002) examine the relationship between currency crises and
poverty across a number of developing countries during 1960–98.
Compared with a standardized control group that did not suffer any such
crises, the study reports that besides increases in inflation rates ‘by nearly
62 percent relative to the pre-crisis year’ and an increase in unemployment
by 1.1 percent in crisis years relative to pre-crisis years, the inequality
increases as well as poverty worsens in the group of countries that suffer
currency crises. A decline in average income of a country is accompanied
by a more than proportionate decline in income of the lowest and second
lowest quintiles, which worsens income inequality as well as increases inci-
dence of poverty, as households in the lower quintiles are more likely to
have income below the poverty line. Further, government cutbacks in
expenditure on social programs, health and education are associated with
falling incomes of the poorest. In the case of Mexico, the study finds that
the ‘poverty rate spiked to nearly 17 percent of the population in 1996, from
nearly 11 percent in 1994, reversing the gains made in 1992–94’.
Thus, as discussed above, occurrence of financial crisis can be expected
generally to have an impact on the incidence of poverty as well as the
income distribution in the country. In fact, such crises will also have
broader effects on the living standards of the affected countries. We will
discuss these issues in turn mainly for the case of the ‘Affected Five’ coun-
tries of East Asia in the context of the 1997–98 Financial Crisis.
4.1 Impact on the Incidence of Poverty
Generally speaking, poverty worsens at the onset of a financial or cur-

rency crisis and, in many instances, such an ill effect is relatively long-lived.
Table 1.6 presents the data for selected countries that have recently suffered
Table 1.6 Incidence of poverty in selected countries
Country and Type of Crisis Incidence of Poverty as Percentage

(year in parentheses)
Before crisis Year of crisis After crisis
Argentina, hyperinflation 25.2 (1987) 47.3 (1989) 33.7 (1990)
and currency
Argentina, contagion 16.8 (1993) 24.8 (1995) 26.0 (1997)
Indonesia, contagion 11.3 (1996) 18.9 (1998) 11.7 (1999)
and financial
Mexico, currency and financial 36.0 (1994) nd 43.0 (1996)
Russian Federation, financial 21.9 (1996) 32.7 (1998) nd
Thailand, currency and financial 11.4 (1996) 12.9 (1998) nd
Note: Based on national poverty lines and per capita household income except for
Indonesia (per capita expenditure), Mexico (household income), and Russia (household
expenditure per equivalent adult). Data for Argentina refer to Greater Buenos Aires. nd⫽
no data available.
Source: World Bank (2000), p. 163.
a financial crisis. For instance, poverty rose by almost 50 percent in

Indonesia as a result of the 1997 Asian Crisis and, in the case of Russia,
the incidence of poverty rose from 21.9 percent to 32.7 percent between
1996 and 1998. While in the East Asian case, poverty fell as the economies
recovered; in Russia as well as Latin American countries that experienced
a financial crisis, the incidence of poverty remained relatively elevated com-
pared with its pre-crisis levels even several years later.
4.2 Impact on Income Inequality
The available empirical evidence suggests that, in general, inequality rises or,
at best, stays constant during a crisis. As reported in the World Development
Report of 2000–01 (World Bank, 2000), for Latin American countries,
inequality (as measured by the Gini coefficient) increased in 15 of 20 crisis
episodes for which there are data available. For the East Asian countries, on
the other hand, the pattern was somewhat more complex. While during the
crisis year or very soon after it (that is, 1997 or 1998), the Gini coefficient was
virtually unchanged, it often increased (signaling worsening inequality) in
the years following the onset of the crisis. However, in some cases, the Gini
coefficient has dropped as economies have recovered (for instance, Thailand).
Table 1.7 has pre- and post-crisis data for the Gini coefficient (column 1)
for several of the East Asian countries affected by the 1997 crisis – Malaysia,
Table 1.7 Gini coefficients, headcount index and mean consumption/month

for selected East Asian countries
Gini Coefficient Headcount Index Mean Consumption

(Scale: 0–100) (%) $2/day (1993 PPP $/month)
Malaysia
1990 46.2 18.5 195.32
1992 47.7 17.6 219.48
1995 48.5 14.0 253.64
1996 48.5 13.1 261.87
1997 49.1 8.8 315.95
1998 49.1 12.9 269.00
1999 49.1 12.6 271.52
2000 49.1 9.8 303.31
2001 49.1 9.8 303.46
2002 49.1 9.6 305.14
2003 49.1 8.7 316.58
2004 49.1 7.9 328.80
Korea
1990 42.9 ⬍0.5 301.09
1993 43.3 ⬍0.5 330.38
1994 43.7 ⬍0.5 362.09
1995 44.1 ⬍0.5 383.03
1996 44.5 ⬍0.5 411.09
1997 45.0 ⬍0.5 440.03
1998 45.5 ⬍0.5 480.46
1999 46.0 ⬍0.5 483.84
2000 46.4 ⬍0.5 400.86
2001 46.9 ⬍0.5 450.06
2002 47.3 ⬍0.5 478.67
2003 47.6 ⬍0.5 492.54
2004 48.0 ⬍0.5 535.38
Thailand
1990 43.8 47.0 102.88
1992 46.2 37.5 129.75
1996 43.4 28.2 143.92
1998 40.6 34.1 121.73
1999 40.7 33.6 123.50
2000 43.2 35.6 125.42
2001 42.4 32.0 131.21
2002 42.2 27.7 139.40
2003 41.4 23.7 147.03
2004 41.4 21.4 153.97
Table 1.7 (continued)
Gini Coefficient Headcount Index Mean Consumption

(Scale: 0–100) (%) $2/day (1993 PPP $/month)
Indonesia
1990 28.9 71.1 61.58
1993 31.7 61.6 68.54
1996 36.5 50.5 86.62
1999 31.0 65.1 66.80
2000 32.2 59.5 72.53
2001 32.1 58.7 73.44
2002 34.3 53.5 81.72
2003 34.8 50.1 85.88
2004 35.4 48.1 88.82
Philippines
1990 40.7 55.6 82.77
1991 43.8 53.5 90.32
1994 43.8 55.0 87.75
1996 42.9 53.1 89.10
1997 46.2 46.5 107.15
1998 46.0 45.2 110.21
1999 46.7 46.6 108.77
2000 46.2 46.9 107.2
2001 46.2 47.1 107.03
2002 46.0 45.3 110.61
2003 45.9 44.1 113.32
2004 45.9 41.8 118.10
Note:
The poverty lines in Table 1.7 are set at $2.15 per person per day (in 1993 PPP $) for all
countries. For most countries, 1993 World Bank PPP (purchasing power parity) estimates are
used. The PPP for the Philippines is from the Penn World Tables. Estimates for all countries
except Malaysia are based on surveys of household consumption. The estimates for Malaysia
use income surveys. These poverty estimates differ from those commonly found in national
poverty assessments for two main reasons. First, country assessments use national poverty
lines that differ from the uniform international poverty lines used here. Second, national
poverty lines also typically allow for spatial cost-of-living differentials within countries, but
such adjustments are omitted here to maintain a consistent methodology across countries. For
instance, in the case of Thailand, these differences explain why the above estimates indicate a
small increase in poverty between 1998 and 2000 (in spite of adjusting the CPI by the change
in the national poverty lines over this period), while national poverty line-based estimates
indicate a decline. Also for Thailand, the 2002 estimate is based on a longer consumption
module, which could lead to a small overestimation of consumption relative to 2000.
Korea, Thailand, Indonesia and the Philippines. It is worthwhile looking

at the pattern of the changes in Gini coefficient in each of these selected
countries.
In Malaysia’s case, the Gini coefficient increased from 48.5 in 1996 to
49.1 in 1997, yet after this initial increase in inequality during the crisis year,
the Gini coefficient has virtually stayed constant (at 49.1) since then.
In the case of South Korea, the Gini coefficient registered a slight
increase from 44.5 in 1996 to 45.0 in 1997. However, the Gini coefficient has
steadily increased since then signifying a steady trend towards deteriorat-
ing income inequality.
In Thailand’s case, the Gini coefficient started at 43.4 in 1996. While the
relevant statistic for 1997 is not available, the Gini coefficient’s value for
1998 was 40.6; which surprisingly showed an improvement in income
equality during the year of crisis. For the longer run, however, the Gini
coefficient has shown a mixed pattern – worsening in 2000, 2001 and 2002
relative to 1998 or 1999 before starting to improve of late.
In Indonesia, the Gini coefficient stood at 31.7 in 1993 – four years prior
to the 1997 crisis, while its value was equal to 31.0 in 1999 two years after
the crisis – thus, it would seem that income inequality was virtually
unchanged during the crisis. However, the 1996 Gini coefficient value of
36.5 would appear to tell a different story. If we are willing to set aside 1996
as an ‘outlier’, the trend appears to be one of increasing inequality since
1990, which certainly accelerated after 1999.
In the case of the Philippines, the Gini coefficient jumped in 1997 to 46.2
from 42.9 in the previous year – a significant increase in income inequality
during the year of the crisis. This elevated level of income inequality has
pretty much persisted over the longer run as evidenced by the values of the
Gini coefficient hovering quite close to 46.
When crises are accompanied by increasing inequality, economic down-
turns can impede or even reverse previous gains in poverty reductions and
also adversely affect standards of living. Let us look at this issue for the
East Asian countries affected by the 1997 crisis.
Also, in Table 1.6, we will measure poverty reduction by declines in
‘Headcount Index (%) $2/day’, which measures the percentage of the popu-
lation living below $2/day (column 2) and standard of living by ‘Mean
Consumption (1993 PPP $/month)’ which is given in column 3.
In the case of Malaysia, the headcount index had been steadily declining
before the 1997 crisis. However, in 1998, this index jumped almost 25
percent from 8.8 in 1997 to 12.9 in 1998 and even though significant
improvements were evident after 1999, it took several more years (until
2003) for poverty reduction to attain the level of 8.7, which approximated
the situation in 1997. Thus, once again, we see that a crisis accompanied by
even a very slight worsening income inequality between 1996 and 1997
meant reversals in poverty alleviation in Malaysia.
Similarly, the mean consumption in Malaysia that had steadily risen
from 195.32 in 1990 to 315.95 in 1997, dropped to 269.00 in 1998 although
it picked up rather smartly soon after that, perhaps, in part, due to the fact
that inequality measures were nearly constant after 1997, the year of the
crisis. If inequality would have worsened significantly after the crisis, it is
very likely that Malaysia’s mean consumption would have stayed depressed
for a considerably longer period of time.
In the Korean case, while income inequality increased after the crisis,
such a trend had been in place even before the crisis. Thus it would appear
that there were structural factors unique to the Korean economy that were
prompting an increasing trend in Gini coefficient values independent of the
1997 crisis. Nevertheless, the data indicate that any such pre-crisis trend
may have been exacerbated as a result of the crisis. Further, in terms of the
status of poverty reduction in Korea, the headcount index has stayed con-
stant at a relatively low level (⬍0.5 percent) since 1990 and the mean con-
sumption series also seems to have been little affected by the incidence of
the 1997 crisis. The most likely reason for this pattern is the fact that in the
Korean case, relatively speaking, the recovery from the crisis was the fastest
of all the affected countries (the ‘V’-shaped pattern of recovery), thus the
previous trends were little disturbed. Korea, in other words, has fared the
best by not suffering any major ill effects of the 1997 crisis in terms of pos-
sible reversals in poverty reduction.
Thailand was the ‘classic’ case where the 1997 crisis led to immediate
reversals in poverty reduction trends. Prior to the crisis, Thailand had been
making remarkable progress in poverty reduction as can be noted from the
fact that, by 1996, the headcount index had dropped to 28.2 from a high of
47.0 in 1990. However, one year after the crisis, the index had jumped
significantly to 34.1 (in 1998) relative to 28.2 in 1996. However, economic
recovery by 2002 meant that the pace of poverty reduction had been
restored. The headcount index was 27.7 in 2002 and, it has, in fact, contin-
ued to drop since then.
The mean consumption pattern in Thailand as of 1990 has followed a
pattern similar to that of the headcount index for the country. The mean
consumption increased from a level of 102.88 in 1990 to a peak of 143.92
in 1996 only to drop precipitously to 121.73 in 1998 on account of the
financial and economic crisis. The previous peak of 1996 was regained only
in 2003 when the mean consumption level rose to 147.03.
Similarly, for Indonesia, the mean consumption had been steadily
increasing before the 1997 crisis (61.58 in 1990 to 86.62 in 1996). However,
this figure tumbled to 66.80 in 1999 compared with its value of 86.62 in
1999, and it took several years of slow recovery to come close to the 1999
level, when in 2003, the mean consumption climbed back to 85.88. Thus,
standard of living, as measured by mean consumption was significantly
adversely affected due to the crisis and it recovered only after almost six
years.
The situation in Indonesia was similar to that of Malaysia. Indonesia
had been making significant gains by way of poverty reduction as head-
count index had dropped from 71.1 percent in 1990 to 50.5 percent in 1996.
However, it jumped back up to 65.1 percent in 1999. It was only several
years later in 2003 that the poverty reduction level attained in 1996 was
regained.
In the case of the Philippines, the trend in poverty reduction witnessed in
terms of a secular trend in lower headcount index values continued unabated
through the 1997 crisis year and after. Similarly, for the Philippines, the mean
consumption continued to show a steady upward trend through the 1990s
and into the early 2000s. In some sense, this result for the Philippines may
appear at odds with the results for similar measures for say, Indonesia or
Malaysia. However, for one, this may be due to the fact that the Philippines
had been a ‘latecomer’ to the economic growth era dubbed ‘The Asian
Miracle’ and thus had ‘less to lose’. For another, IMF- and later, World
Bank-inspired safety net programs and redistributive policies of the
Philippines’ government, may have cushioned the blow even in the face of
worsening income inequality.
4.3 Impact on Living Standards
The full impact of financial or economic crises on the living standards in

the affected countries goes beyond their impact on measures of poverty and
inequality. As measured by social indicators of health and schooling enroll-
ments, the living standards take a turn for the worse. As a result of the crisis,
typically, such social indicators either deteriorate or improve only at a
slower pace with their long-term implications for reducing earnings poten-
tial of the poor (Table 1.8).
As can be seen from the Table 1.8, in Mexico, on account of the debt
crisis, the infant and pre-school mortality due to nutritional deficiency rose
in the 1980s (reversing a beneficial trend) and such infant mortality rose
again in 1995 on the heels of the peso crisis. In Indonesia, the percentage
of women with body mass index in a range vulnerable for illnesses and
death increased by a quarter in 1998. Also, the average weight of babies
under age three years declined.
Besides the specifics noted in the above table, it may be mentioned here
that in Latin America, overall, the infant mortality rates increased only at
Table 1.8 Societal impact of economic crises in selected countries
Main Crisis Health Education

Indicators Indicators Indicators
Mexico 1995 • Per capita • Among children • Gross primary

GDP fell 7.8% under age one, enrollment
mortality from increased 0.44% in
anemia increased 1994, but fell 0.09%
from 6.3 deaths in 1995
per 100 000 live
births in 1993
to 7.9 in 1995
• Per capita • Among children
private aged one to four, the
consumption mortality rate from
fell 11.1% anemia rose from 1.7
to 2.2 per 100 000
Indonesia • Per capita • The share of • The dropout rate
1998 GDP fell 14.6% women whose body for children in the
mass index is poorest fourth of the
below the level population rose from
at which risks of 1.3% in 1997 to 7.5%
illness and death in 1998 for those aged
increase rose 25% 7–12 and from 14.2%
to 25.5% for those
aged 13–19. In both
cohorts the poorest
fifth experienced the
largest increase
• Per capita • Most indicators • The share of
private of child nutritional children in the
consumption status remained poorest fourth of the
fell 5.1% stable. The exception population not
may be the weight enrolled in school
(conditional on rose from 4.9% in
height) of children 1997 to 10.7% in 1998
under age three, for those aged 7–12
suggesting that and from 42.5% to
families may be 58.4% for those aged
investing in some 13–19. In both
members at the cohorts the poorest
expense of others fourth experienced
the largest increase
Source: World Bank (2000), p. 164.

a relatively lower rate in the 1980s compared with the earlier decade. Thus,
we can see that financial crisis adversely affect the state of the social indi-
cators of health either by slowing down their progress or even causing an
absolute decline in them.
Similarly, broad indicators of human capital investments such as school
attendance, enrollment and literacy were also adversely affected during
crises. It may also be noted that, in Mexico, the proportion of each graduat-
ing high school class that went on to the next level declined during the 1980s
debt crisis. Also, primary school enrollment dropped too in the 1980s. There
was a similar drop (of almost 1 percent) in 1995 on the eve of the peso crisis.
The adverse impact of financial and macroeconomic crises on human
capital accumulation and health improvements are the channels through
which such crises have secular impact on poverty and potential earnings of
the affected populace, especially the poor who are disproportionately
affected. According to the World Development Report 2000–2001 World
Bank (2000), for Indonesia, the dropout rate in the lowest fourth of income
distributions rose from 1.3 percent in 1997 to 7.5 percent in 1998 among
children aged 7–12 and from 14.2 percent to 25.5 percent among ages
13–19. At the same time, the proportion of poor children not enrolled in
school more than doubled from 4.9 percent to 10.7 percent.
In conclusion, it needs to be emphasized that the social costs of the Asian
Crisis were substantial, and a disproportionate burden fell on the poor or
the ‘near poor’. In each of the ‘Affected Five’ countries, the 1997–98 crisis
led to higher incidence of poverty, and the living standards in general were
adversely affected in these countries. Also, by and large, the income inequi-
ties were often exacerbated on account of the crisis.
4.4 Policy Implications
The above analysis, in sections 4.1–4.3, clearly implies the kind of public poli-
cies that are needed to ameliorate the situation precipitated by financial crises.
There is an obvious need to provide a better social security net that should be
‘targeted’ to the support groups most affected by the crisis, robust and put in
place well in advance of the crisis. Policy-makers need to take concrete steps
to protect pro-poor public spending in the wake of a financial crisis.
Finally, in an overarching sense, policy-makers should strive to balance
the purely financial aspects of the policy response to a financial crisis
against a meaningful concern for the societal costs of such measures. Thus,
somewhat akin to the now commonplace ‘environmental impact’ study of
any land development or economic growth project, there should be a social
cost impact study of the feasible financial policies. The final choice should
be based on the optimal policy net of its social costs.
5 CONCLUDING REMARKS
Since financial crises can be costly both in economic, as well as, societal
terms, it is only natural to inquire whether such crises are predictable.
Besides providing an overview of the various theories that motivate the
specification of possible early warning systems, this chapter reviewed the
performance of logit/probit as well as more recent Markov-switching
models of currency crisis prediction. In this respect, we would like to con-
clude that while econometric predictive models can be very useful in
identifying various indicators of ‘vulnerability’, such exercises are not a
cure-all.
While formal econometric predictive models of currency crises are very
useful tools, due to limitations related to convergence requirements and
other technical matters, they can be too structured and thus restrict the
scope of the discussion necessary to get a broad sense of the possible deter-
minants of the vulnerabilities of economies to a currency crisis. In this
chapter, we also examined the broad determinants of vulnerability or sus-
ceptibility to crisis of an economy. This enables us to think about the mech-
anisms that may be needed to monitor the economy as well as institute
policies to reduce its susceptibility to a financial crisis.
Finally, it is also very important to explore the aftermath of a financial
crisis. Since the after-effects of such crises are societal and not only eco-
nomic, this chapter focuses on the income distributional consequences of
the Asian Financial Crisis of 1997–98 as well as its impact on the poverty
alleviation trends in the ‘Affected Five’ countries. Amongst the suggested
policy implications is the need to put in place an early warning system and,
at the same time, institute structural reforms to reduce the chances of recur-
rence of financial crises or soften their impact.13 To deal with the societal
aftermath of a crisis, we recommend robust safety nets targeted towards
those in the economy that are most likely to be adversely impacted by a
financial crisis.
NOTES
1. This discussion is based, in a significant fashion, on Shabbir’s earlier joint work reported
in Mariano et al. (2002).
2. See Berg and Pattillo (1999) for a skeptical viewpoint.
3. This discussion is drawn from Shabbir’s earlier joint work reported in Mariano et al.
(2003).
4. Also see, Chauvet and Dong (2004).
5. The significance of these indicators can be motivated analytically, however, in this
chapter, we do not engage in such an exercise.
6. The discussion in this section has drawn on the analysis presented in Warr (2002).
7. For instance, Bhagwati (1998, p. 12) notes ‘[T]he only explanation that accounts for the
massive net capital outflows is panic and herd behaviour: whether of domestic or foreign
nationals’ and Radelet and Sachs (1998, p. 43) infer that the ‘crisis was triggered by dra-
matic swings in creditors’ expectations about the behavior of other creditors, thereby cre-
ating a self-fulfilling – although possibly individually rational – financial panic’. On the
other hand, as Dornbusch (1997, p. 21) notes, ‘vulnerability means that if something
goes wrong, then suddenly a lot goes wrong’. The ‘something’ that causes things to go
wrong can be considered as a trigger, whereas the possibility of ‘suddenly a lot going
wrong’ is conditional on existence of ‘vulnerability’.
8. The list of indicators we have examined is certainly not an exhaustive one. Warr (2002),
in a similar analysis of pre-1997 crisis determinants of vulnerability for Thailand,
Indonesia and Korea, concludes that in these countries there had been a significant
increase in fragility of the banking sector as measured by exposure to interest rate risk
and deteriorating quality of loans (in the absence of data on non-performing loans,
quality was indirectly proxied by an increasing ratio of total loans outstanding for the
banking sector to GDP) as well as currency depreciation risk (measured by the ratio of
foreign liabilities to total loans). Based on evidence provided by indicators such as
‘reserves’/ ‘Volatile’ Capital Inflows, ‘CA/FDI’ and significant real exchange rate appre-
ciation (as proxy of erosion in foreign competitiveness), Warr (2002) also demonstrates
that, in general, these countries had become vulnerable to a currency crisis much before
1997. Ghosh and Ghosh (2003) come to a similar conclusion for ‘deep’ currency crises,
however, see Berg and Pattillo (1999) for a somewhat skeptical viewpoint favoring con-
tagion as the primary explanation.
9. See Barro (2001) for an early attempt at examining the pre- versus post-crisis situa-
tion in the same countries. Also, see Adams and Shabbir, Chapter 6 in this volume, for a
more recent exploration of some analytical aspects of the pre- versus post-crisis com-
parisons.
10. Incidentally, Klein et al., Chapter 4 in the present volume, analyzes the macroeconomic
consequences of the Malaysian capital controls policy more formally by means of a
macroeconometric model and conclude that in the Malaysian case, the capital controls
policy had, on average, beneficial effects in terms of post-crisis stabilization.
11. Interestingly, Warr (2002) maintains that the real exchange rates were overvalued for
Indonesia, Thailand and Korea for reasons independent of China’s exchange rate policy.
12. Some exceptions are Hyun and Lim (2003); Baldacci et al. (2002); Birdsall and Haggard
(2000); Lee (1998) and Shabbir (1998).
13. The authors have discussed several reforms in the Introduction to this volume; add-
itional details about relevant reforms may be noted in Klein (2004).
REFERENCES
Abiad, A.G. (2002), ‘Early Warning Systems for Currency Crises: A Markov
Switching Approach with Applications to Southeast Asia’, PhD dissertation,
Department of Economics, University of Pennsylvania, Philadelphia.
Alvarez-Plata, Patricia and Mechthild Schrooten (2003), ‘The Argentinean
Currency Crisis: A Markov-Switching Model Estimation’, Discussion Paper,
DIW Berlin.
Baldacci, Emanuele, Luiz de Mello and G. Inchauste (2002), ‘Financial Crises,
Poverty, and Income Distribution’, IMF Working Paper 02/4.
Barro, Robert J. (2001), ‘Economic Growth before and after the Financial Crisis’,
Working Paper 8330, Cambridge, MA: National Bureau of Economic Research.
Berg, A. and C. Pattillo (1999), ‘Predicting Currency Crises: The Indicators Approach
and an Alternative’, Journal of International Money and Finance, 18, 561–86.
Bhagwati, Jagdish (1998), ‘Asian Financial Crisis Debate: Why? How Severe?’,
Paper presented at Managing the Asian Financial Crisis: Lessons and
Challenges Conference, Asian Strategic Leadership Institute and Rating Agency,
Kuala Lumpur, 2–3 November.
Birdsall, Nancy and S. Haggard (2000), After the Crisis: The Social Contract and
the Middle Class in East Asia, Carnegie Endowment for International Peace.
Brunetti, C., R.S. Mariano, C. Scotti and A.H.H. Tan (2003), ‘Markov Switching
GARCH Models of Currency Crises in Southeast Asia’, Working Paper 03-008,
Philadelphia: Penn Institute of Economic Research, University of Pennsylvania.
http://pier.econ.upenn.edu/Archive/03-008.pdf.
Chauvet, M. and F. Dong (2004), ‘Leading Indicators of Country Risk and
Currency Crises: The Asian Experience’, Economic Review, Federal Reserve
Bank of Atlanta,Q1, 25–37.
Demirgüç-Kunt, A. and E. Detragiache (1998), ‘The Determinants of Banking
Crises in Developing and Developed Countries’, IMF Staff Papers, 45 (1), 81–109.
Dornbusch, R. (1997), ‘A Thai–Mexico Primer: Lessons for Outmaneuvering a
Financial Meltdown’, The International Economy, September/October.
Eichengreen, B. and A. Rose (1998), ‘Staying Afloat When the Wind Shifts:
External Factors and Emerging-market Banking Crises’, NBER Working Paper
No. 6370, Cambridge, MA.
Ghosh, S.R. and A.R. Ghosh (2003), ‘Structural Vulnerabilities and Currency
Crises’, IMF Staff Papers, 50 (3).
Hamilton, James B. (1994), Time Series Analysis, Princeton, NJ. Princeton
University Press.
Hyan, Jin K. and B. Lim (2003), ‘Financial Crises and Income Distribution in
Korea’, Working Paper, Korea Development Institute.
Inada, Yoshihisa, Lawrence R. Klein and J. Makino (2000), ‘A Retrospective View
of the Asian Financial Crisis: Special Reference to Exchange Rate Policy’, in L.
Klein and S. Ichimura (eds), Econometric Modeling of China, Singapore: World
Scientific.
Kaminsky, G. and C. Reinhart (1998), ‘Financial Crises in Asia and Latin America:
Then and Now’, American Economic Review, 88, 444–8.
Kaminsky, G. and C. Reinhart (1999), ‘The Twin Crises: The Causes of Banking
and Balance of Payments Problems’, American Economic Review, 89, 473–500.
Kaminsky, G., S. Lizondo and C. Reinhart (1998), ‘Leading Indicators of Currency
Crises’, IMF Staff Papers, 45 (1), 1–48.
Klein, Lawrence R. (2004), ‘Financial Crises in Asia and Development Issues’,
mimeo, Department of Economics, University of Pennsylvania, Philadelphia.
Krugman, Paul (1979), ‘A Model of Balance of Payments Crises’, Journal of
Money, Credit and Banking, 11, 311–25.
Krugman, Paul (1994), ‘The Myth of Asia’s Miracle’ , Foreign Affairs, November/
December.
Kumar, M., U. Moorthy and W. Perraudin (2003), ‘Predicting Emerging Market
Currency Crashes’, Journal of Empirical Finance, 10, 427–54.
Lau, Lawrence J. (1995), ‘LINK meeting presentation’, Pretoria, South Africa.
Lee, Eddy (1998), The Asian Financial Crisis: The Challenge for Social Policy,
Geneva: International Labour Office.
Mariano, R.S., A.G. Abiad, S. Özmucur, T. Shabbir and A.H.H. Tan (2003),
‘Markov Chains in Predictive Models of Currency Crises – With Applications to
Southeast Asia’, Taiwan Economic Review, 31 (4), December.
Mariano, R.S., B.N. Gultekin, S. Özmucur, T. Shabbir and C.E. Alper (2002),
‘Predictive Models of Economic Financial Crises (Final Report)’, Report sub-
mitted to the Economic Research Forum (ERF) for the Arab countries, Iran and
Turkey for ERF Research Project ERF99-US-4004, mimeo, Department of
Economics, University of Pennsylvania, December.
‘Prediction of Currency Crises: Case of Turkey’, Review of Middle East
Economics and Finance, 2 (2).
Martinez-Peria, M.S. (1999), ‘A Regime Switching Approach to Studying
Speculative Attacks: A Focus on European Monetary System Crises’, World
Bank Development Research Group Working Paper 2132, June.
Masson, P. (1998), ‘Contagion: Monsoonal Effects, Spillovers and Jumps between
Multiple Equilibria’, IMF Working Paper 98/142.
Obstfeld, M. (1994), ‘The Logic of Currency Crises’, Cahiers Economiques et
Monetaires, 43, 189–213.
Pérez, J. (2005), ‘Empirical Identification of Currency Crises: Differences and
Similarities between Indicators’, Applied Financial Economics Letters, 1, 41–6.
Radelet, S. and J. Sachs (1998), ‘The East Asian Financial Crisis: Diagnosis,
Remedies, Prospects’, Brookings Papers on Economic Activity, 2, 1–89.
Shabbir, T. (1998), ‘The Social Costs of the Asian Financial Crisis’, mimeo,
Department of Economics, University of Pennsylvania, Philadelphia.
Warr, Peter (2002), ‘Crisis Vulnerability’, Asian-Pacific Economic Literature, 16 (1),
36–47, May.
World Bank (2000), World Development Report (2000/2001), Attacking Poverty,
New York: Oxford University Press.
2. Indicators and analysis of
vulnerability to currency crisis:
Thailand*
Pranee Tinakorn
1 INTRODUCTION
Thailand was the first of the Asian developing countries to experience the
financial crisis in 1997. The crisis started with a series of currency attacks
during late 1996 and mid-1997, which eventually led to the collapse of
Thailand’s fixed exchange rate system on 2 July 1997. The consequent
severe fall of around 40–108 percent in the baht value (from June 1997
average of 25.75 baht/$ to January 1998 average of 53.71 baht/$ before hov-
ering around 36–45 baht/$ later) led to a tremendous increase in the liabil-
ity side of the balance sheet for a significant number of enterprises that had
borrowed heavily from international markets. Many financial institutions
were faced with both liquidity and insolvency problems. The situation was
aggravated by the contagion effect of currency depreciation that hit other
Asian countries and had a further feedback effect on the Thai economy. In
addition, since the Bank of Thailand (BOT) had used up almost all of the
international reserves through the swap operation to defend the baht value
in vain, resulting in a net reserve of only about 1 billion dollars in July 1997
(see Figure 2.1),1 the country had to turn to the International Monetary
Fund (IMF) for a stand-by credit of 17.2 billion dollars.
The IMF package required the country to cut fiscal expenditure and raise
interest rates mainly to restore international confidence. This in effect dried
up whatever little domestic liquidity there was and the economy inevitably
headed into a deep recession. Thailand’s real GDP in 1998 contracted
by –10.4 percent and the unemployment rate increased more than three
times the rate of 1997.
In fact, even without the IMF austerity package, Thailand had been
heading for an economic downturn anyway as predicted by the composite
leading indicator. Tinakorn (1998) constructs a composite leading economic
indicator for Thailand and finds that the composite leading indicator had
69
Jan-97 Dec-97
45 000.00
40 000.00
35 000.00
30 000.00
(million bank)
25 000.00
20 000.00
15 000.00
10 000.00 Gross Reserves

Net Reserves
5000.00
0.00
95 l-9
5 96 l-9
6 97 l-9
7 98 l-9
8 99 l-9
9 00 l-0
0 01
n- n- n- n- n- n- n-
Ja Ju Ja Ju Ja Ju Ja Ju Ja Ju Ja Ju Ja
Figure 2.1 Thailand’s international reserves: gross and net
been following a downward trend since early 1996, with a leading time of six
months on average. Although Thailand needed the IMF stand-by credit
badly, the Thai authorities might have negotiated for a less contractionary
package if they had had prior knowledge about the leading indicator of the
economy and where it was heading. Similarly, the IMF might have yielded
to a less contractionary package if they had known that the economic down-
turn in Thailand would have occurred anyway and such downturn would
have helped correct the current account deficits without a strong contrac-
tionary package initially imposed by the IMF.
Although the use of leading indicators as warning signals is widespread
in most developed countries, especially those in the Organisation for
Economic Co-operation and Development (OECD), the situation is quite
different for Thailand where many leading indicators were simply not
available and cannot be incorporated into the composite index. It is sug-
gested that the concerned authorities should spend more efforts to gather
the required data and try to improve the performance of the present com-
posite leading indicator so that it can serve as a more reliable warning
signal in the future. In addition to improving and monitoring leading eco-
nomic indicators for the real sector activities, Thailand also needs some
warning signals about the impending crisis in the financial sector and the
balance of payments as well. Kaminsky and Reinhart (1999) found that
Indicators and analysis of vulnerability to currency crisis: Thailand 71
problems in the banking sector typically precede a currency crisis and the
currency crisis deepens the banking crisis. For example, the 1996 abrupt
halt to Thailand’s export growth2 might not have caused such a severe
attack on the baht value if Thailand had not already had troubles in its
financial sector. The high ratios of non-performing loans (NPLs) among
the Thai banks and financial institutions due to the sluggishness in the
property sector must have created fear about their ability to repay debt.
This unsurprisingly leads baht traders and speculators to have growing
doubts about Thailand’s debt servicing capacity and creditworthiness.
While it would be desirable to study leading indicators for both the
banking crisis and currency crisis for Thailand, due to time and data con-
straints this study is focused on analyzing warning signals for the currency
crisis alone.
Since the banking crisis and currency crisis are found to have strong
links, many studies on currency crises also found that overstretched
financial variables provide warning signals to currency crisis. By including
several indicators from the financial sectors in the analysis of currency
crisis, it is hoped that they will provide warning signals not only to currency
crisis but also to financial problems. While most studies on currency crisis
use international cross-section data to look for signals of crisis, this study
will use the time-series data on a monthly basis. Both non-parametric and
parametric analysis will be explored. When a set of leading indicators are
confirmed from statistical analysis, they can be used to assess future vul-
nerability of the Thai economy.
2 REVIEW OF SOME SELECTED STUDIES ON

CURRENCY CRISIS
2.1 Theoretical Explanation of Currency Crisis
Prior to the Asian currency crisis in the late 1990s, there were basically
two models explaining the onset of currency attacks. They are known as
the first- and second-generation models. The former model focused on the
balance-of-payments problems created mainly through seignorage; the
latter model viewed currency crisis not as a result of bad policy but of a
shift in expectation and the model was called ‘self-fulfilling’. However, the
causes of speculative attacks on Asian currencies, with the Thai currency
being hit first, appeared to be different from those explained in the first- and
second-generation models. Therefore, economists came up with a third-
generation model to explain currency crisis. Kaminsky et al. (1998) pro-
vided a summary of the main explanation of speculative attacks and
Krugman (2001) provided a brief history of currency crisis modeling. The

following summary draws heavily from both papers.
First-generation crisis models

Krugman (1979) explained that under a fixed exchange rate system, the
expansion rate of domestic credit in excess of money demand growth led
to persistent losses of international reserves, which ultimately led to specu-
lative attacks on the currency. The root cause of the crisis appeared to be
poor government policy, that is, excessive public sector deficit that becomes
monetized under a fixed exchange rate system. The balance-of-payments
crises or near-crisis events that took place in the 1970s3 provided inspir-
ation for the first-generation modeling.
Second-generation crisis models

The series of attacks on some European currencies under the European
Monetary System (EMS) in 1992–93 could not be explained by the first-
generation models. Seignorage was not an issue and it was not the depletion
of reserves that led the authorities to abandon the parity. Rather, it was a
matter of policy choice as they may have been concerned about the adverse
consequences of policies needed to maintain the parity (for example, the
effect of higher interest rates on employment). Obstfeld (1996) offered
several variants but the main theme seemed to focus on macroeconomic
trade-offs and decisions, and the existence of multiple equilibria. If specu-
lators question the credibility of the peg, they may attack and the result can
be a self-fulfilling crisis of confidence. Within the second-generation
models, crises are no longer the result of irresponsible policy, but they occur
because market participants expect them to.
Third-generation crisis models

The Asian Crisis in 1997 appeared to be the inspiration for several variant
explanations in the third-generation crisis models. Krugman (2001) sug-
gested that there were three versions: moral-hazard-driven investment,
which leads to an excessive build-up of external debt and then to a collapse;
bank-run; and balance-sheet implications of currency depreciation.
In addition to these three main reasons as explanation for currency
crises, some papers also discuss ‘contagion effects’. Gerlach and Smets
(1995) explained that the devaluation by one country could lead its trading
partners to devalue in order to avoid a loss of competitiveness. Other chan-
nels, such as financial linkages, can also serve as transmission of contagion
effects (see Calvo and Reinhart, 1996 and Eichengreen et al., 1996 for
further discussion). Since the growth in international bank lending to
emerging markets and its sharp retrenchment was a remarkable feature of
the East Asian Crisis, Jeanneau and Micu (2002) found evidence of several
fundamental factors as determinants of international capital flows. Dodd
(2001) pointed out that the ‘carry trade’, whereby foreign exchange for-
wards and swaps were used to hedge as well as speculate on the fixed
exchange rate regimes while profiting from the interest rate differential
between pegged currencies, was the major factor preceding and precipitat-
ing the Thai currency crisis in 1997. If authorities are not mindful of their
activities, derivatives can make the economy more susceptible to financial
crisis because they create conditions for entities to raise risk in relation to
capital and to dodge prudential regulatory safeguards even though they
play the useful role in hedging and risk management.
Krugman (2001) also conjectured about a future ‘fourth-generation’ crisis
model, which may not be a currency crisis model, but may be a more general
financial crisis model in which other asset prices play the major role.
2.2 Some Recent Empirical Studies on Currency Crisis Indicators
The theoretical explanation of currency crisis contained in the first- and

second-generation models implies that there are some fundamental variables
or indicators that should help us assess the vulnerability of an economy to
currency attacks. If speculative attacks are due to the ‘self-fulfilling’ crisis of
the second-generation model, the prospect of predicting its occurrence from
various economic indicators is rather dim because attacks may occur against
a fundamentally sound economy. The Mexican Crisis in 1994 and the Asian
Crisis in 1997 revived academic interest in looking for indicators that can
help predict currency crisis.
From the author’s survey of literature published since 1997, the follow-
ing studies are found and hereby summarized.4 Goldfajn and Valdés (1997)
used the logit model to estimate the probability of a crisis and found that
lagged overvalued exchange rate was a statistically significant variable.
Esquivel and Larrain (1998) used the probit model and found the follow-
ing variables to be significant: change in reserve money as a percentage of
GDP, real exchange rate misalignment, ratio of current account to GDP,
ratio of money supply (M2) to international reserves, change in terms of
trade, growth in GNP per capita and contagion effect. Kruger et al. (1998)
also estimated the probit model and found three variables to be consistently
linked to currency crises: a measure of lending booms, real exchange rate
misalignment and the ratio of M2 over reserves. They also performed some
sensitivity analysis and found that other macroeconomic variables did not
have as robust a performance as these three.
The methodological approach used by Kaminsky et al. (1998) is different
from the above three studies. Their ‘signals’ approach is a non-parametric
method where leading indicators of currency crises are identified by their

non-normal behavior, or the so-called ‘signaling’. Their study found the
following variables to be particularly useful: international reserves, real
exchange rate, domestic credit, credit to the public sector and domestic
inflation.
All the above-mentioned studies used cross-section data from different
numbers of countries within the period 1970 to 1997. Their sample size
differs as do their methodology and definition of currency crisis. Table 2.1
provides some major comparison of these four studies.
The 1997 currency crisis also stimulated academic interest on this issue in
Thailand. Table 2.2 summarizes the two pieces of study on early warning
indicators of currency crisis for Thailand. Engwatana (1999) used both
monthly and quarterly data during 1990–98 while Poonpatpibul and
Ittisupornrat (2001) used monthly data during 1990–2000. It is somewhat
surprising that with the same methodology of probit analysis, they found
different sets of significant indicators, except for one variable. This could be
partly due to their different definitions of currency crisis. Engwatana (1999)
used ‘accumulated one month exchange rate change of 10% or more’ or ‘devi-
ation of forward premium from its three-month moving average by more
than 10% in one month’. Poonpatpibul and Itthisupornrat (2001) ‘used an
accumulated weakening of exchange rate of more than 15% in three months’.
It is also notable that the latter study did not find short-term foreign debt
to be a significant indicator in both the signals approach and probit model
even though in reality excessive short-term foreign debt was a prominent
factor in Thailand’s 1997 currency crisis. Besides, there appears to be some
inconsistency in Poonpatpibul and Ittisupornrat’s (2001) reporting of their
sample data. They reported using monthly samples between January 1990
to December 1998 for probit analysis but their regression reported only 78
total observations instead of 108 observations. In addition, following their
definition of a currency crisis, they reported having 16 observations of
crisis (where the dependent variable was set to 1) but if one examines the
exchange rate data, and using their definition, one will find only eight such
observations (this can be verified by examining the data in Table 2.3).
Since the objective of this research study is to evaluate Thailand’s future
vulnerability to currency crisis, the author does not feel comfortable relying
on the results found in other studies that still require data clarification. This
study is an additional attempt in the pool of research work on Thailand’s
leading indicators of currency crises, in the hope that they can help to fore-
warn of such events in the future.
Perhaps, it should be noted here that after this research work was completed
and while in the process of revision, the study on predicting currency crises by
Mariano et al. (2002) was called to the author’s attention. Mariano et al.
Table 2.1 Summary comparison of four studies on leading indicators of currency crises
Goldfajn and Valdés Esquivel and Larrain Kaminsky et al. Kruger et al.
(1997) (1998) (1998) (1998)
1. Sample 26 countries 30 countries 20 countries 19 countries
May 1984–May 1997 1975–96 1970–95 1977–93
2. Definition This paper follows 1. The accumulated three- The index of currency market Exchange rate pressure index is
of currency three alternative month real exchange rate turbulence is more than 3 1.5 times standard deviations
crisis procedures as follows: change is 15% or more or standard deviations above the above mean, where the index is
1. Devaluation is a 2. One-month change in mean, where the index is a defined as a weighted average
crisis when it is the real exchange rate is weighted average of monthly of percentage changes in the
larger than higher than 2.54 times percentage changes in nominal exchange rate and
• 1.96 times the country-specific standard exchange rate and monthly the negative of percentage
standard deviation deviation of real monthly percentage changes in gross changes in international
75
of the country’s growth rate, provided international reserves reserves

nominal exchange that it also exceeds 4%
rate, and
• 2% plus 1.5 times
the devaluation rate
of the previous
month. Crises are
required to be two
months apart
2. Given downward
price rigidity, large
jumps in the real
exchange rate could be
deemed to be a crisis
Goldfajn and Valdés Esquivel and Larrain Kaminsky et al. Kruger et al.
(1997) (1998) (1998) (1998)
3. The index of currency
market turbulence is
more than three standard
deviations above mean,
where the index is a
weighted average of
monthly percentage
changes in gross
international reserves
3. Methodology Logit model Probit model with Signals approach Probit model
76
random effect
4. Variables 1. Overvalued real 1. Change in reserve money 1. Real exchange rate. 1. M2/international reserves
found to be exchange rate a percentage of GDP 2. Banking crises 2. Ratio of bank claims on
significant 2. Current account 3. Exports private sector to GDP (a
indicators imbalance 4. Stock prices measure of lending boom)
3. Real exchange rate 5. M2/reserves 3. Real exchange rate
misalignment 6. Output misalignment
4. Foreign exchange reserves 7. Excess M1 balance
5. Terms of trade shock 8. International reserves
6. Poor growth performance 9. M2 multiplier
7. Regional contagion 10. Domestic credit/GDP
11. Real interest rate
12. Terms of trade
13. Real interest differential
Table 2.2 Summary of studies on currency crisis in Thailand
Engwatana (1999) Poonpatpibul and Ittisupornrat (2001)

1. Sample 1990–98 1990–2000
2. Definition of 1. The accumulated one-month nominal exchange rate Accumulated three-month depreciation in
currency crisis change is 10% or more; or nominal exchange rate is 15% or more
2. The forward premium deviates from its three-month moving
average by more than 10% in one month
3. Methodology Probit model (monthy data) Probit model (quarterly data) Signals approach Probit model
77
4. Variables found 1. Excessive domestic credit 1. High ratio of short-term 1. Export growth 1. Export growth
to be significant creation foreign debt to 2. Change in real 2. Ratio of M2 to
indicators 2. High ratio of M2 to international reserves exchange rate international
international 2. Large domestic and foreign 3. Terms of trade reserves
reserves interest rate differentials 4. Spread between 3. Percentage
3. Low ratio of international 3. Large current account deficit lending rate and change in
reserves to monthly imports 4. Reversal of portfolio deposit rate credits
4. Large domestic and foreign investment capital inflow (private sector)
interest rate differentials 4. Inflation rate
5. Real (effective) exchange
rate overvaluation
Table 2.3 Monthly data on exchange rate and international reserves
Exchange Rate International Net IMF Net Reserves

Reserves Forward Borr-
owing
Baht/US$ % Million % Position (mill $) (mill $) %

change US$ change (mill $) change
Dec-92 25.47 21 181.5 0.0 0.0 21 181.5

Jan-93 25.53 0.24 21 937.0 3.57 0.0 0.0 21 937.0 3.57
Feb-93 25.49 ⫺0.16 21 634.9 ⫺1.38 0.0 0.0 21 634.9 ⫺1.38
Mar-93 25.42 ⫺0.27 22 239.4 2.79 0.0 0.0 22 239.4 2.79
Apr-93 25.23 ⫺0.75 22 611.6 1.67 0.0 0.0 22 611.6 1.67
May-93 25.22 ⫺0.04 23 114.7 2.22 0.0 0.0 23 114.7 2.22
Jun-93 25.21 ⫺0.04 23 979.8 3.74 0.0 0.0 23 979.8 3.74
Jul-93 25.31 0.40 23 919.7 ⫺0.25 0.0 0.0 23 919.7 ⫺0.25
Aug-93 25.18 ⫺0.51 24 222.8 1.27 0.0 0.0 24 222.8 1.27
Sep-93 25.19 0.04 25 225.3 4.14 0.0 0.0 25 225.3 4.14
Oct-93 25.26 0.28 25 544.4 1.26 0.0 0.0 25 544.4 1.26
Nov-93 25.36 0.40 25 206.1 ⫺1.32 0.0 0.0 25 206.1 ⫺1.32
Dec-93 25.45 0.35 25 438.8 0.92 0.0 0.0 25 438.8 0.92
Jan-94 25.53 0.31 25 359.3 ⫺0.31 0.0 0.0 25 359.3 ⫺0.31
Feb-94 25.38 ⫺0.59 26 251.3 3.52 0.0 0.0 26 251.3 3.52
Mar-94 25.29 ⫺0.35 26 672.6 1.60 0.0 0.0 26 672.6 1.60
Apr-94 25.25 ⫺0.16 26 592.8 ⫺0.30 0.0 0.0 26 592.8 ⫺0.30
May-94 25.21 ⫺0.16 27 512.8 3.46 0.0 0.0 27 512.8 3.46
Jun-94 25.14 ⫺0.28 28 340.5 3.01 0.0 0.0 28 340.5 3.01
Jul-94 24.97 ⫺0.68 28 588.3 0.87 0.0 0.0 28 588.3 0.87
Aug-94 25.02 0.20 29 064.0 1.66 0.0 0.0 29 064.0 1.66
Sep-94 24.98 ⫺0.16 29 950.2 3.05 0.0 0.0 29 950.2 3.05
Oct-94 24.96 ⫺0.08 29 851.7 ⫺0.33 0.0 0.0 29 851.7 ⫺0.33
Nov-94 24.98 0.08 29 743.2 ⫺0.36 0.0 0.0 29 743.2 ⫺0.36
Dec-94 25.10 0.48 30 279.0 1.80 0.0 0.0 30 279.0 1.80
Jan-95 25.07 ⫺0.12 29 906.1 ⫺1.23 0.0 0.0 29 906.1 ⫺1.23
Feb-95 25.02 ⫺0.20 30 135.6 0.77 0.0 0.0 30 135.6 0.77
Mar-95 24.76 ⫺1.04 30 119.5 ⫺0.05 0.0 0.0 30 119.5 ⫺0.05
Apr-95 24.56 ⫺0.81 31 727.1 5.34 0.0 0.0 31 727.1 5.34
May-95 24.66 0.41 33 272.4 4.87 0.0 0.0 33 272.4 4.87
Jun-95 24.67 0.04 34 958.3 5.07 0.0 0.0 34 958.3 5.07
Jul-95 24.74 0.28 34 415.7 ⫺1.55 0.0 0.0 34 415.7 ⫺1.55
Aug-95 24.95 0.85 34 629.1 0.62 0.0 0.0 34 629.1 0.62
Sep-95 25.12 0.68 35 866.1 3.57 0.0 0.0 35 866.1 3.57
Oct-95 25.11 ⫺0.04 35 731.4 ⫺0.38 0.0 0.0 35 731.4 ⫺0.38
Nov-95 25.16 0.20 36 204.4 1.32 0.0 0.0 36 204.4 1.32
Dec-95 25.16 0.00 37 026.7 2.27 0.0 0.0 37 026.7 2.27
Jan-96 25.29 0.52 37 721.2 1.88 0.0 0.0 37 721.2 1.88
Feb-96 25.24 ⫺0.20 38 694.2 2.58 0.0 0.0 38 694.2 2.58
Mar-96 25.23 ⫺0.04 38 982.5 0.75 0.0 0.0 38 982.5 0.75
Apr-96 25.27 0.16 38 862.3 ⫺0.31 0.0 0.0 38 862.3 ⫺0.31

owing

May-96 25.29 0.08 39 053.8 0.49 0.0 0.0 39 053.8 0.49

Jun-96 25.35 0.24 39 830.0 1.99 0.0 0.0 39 830.0 1.99
Jul-96 25.34 ⫺0.04 39 360.6 ⫺1.18 0.0 0.0 39 360.6 ⫺1.18
Aug-96 25.27 ⫺0.28 39 370.3 0.02 0.0 0.0 39 370.3 0.02
Sep-96 25.36 0.36 39 537.0 0.42 0.0 0.0 39 537.0 0.42
Oct-96 25.46 0.39 39 902.5 0.92 ⫺500.0 0.0 39 402.5 ⫺0.34
Nov-96 25.45 ⫺0.04 39 613.3 ⫺0.72 ⫺850.0 0.0 38 763.3 ⫺1.62
Dec-96 25.56 0.43 38 724.5 ⫺2.24 ⫺4 890.0 0.0 33 834.5 ⫺12.72
Jan-97 25.69 0.51 39 233.8 1.32 ⫺8 860.0 0.0 30 373.8 ⫺10.23
Feb-97 25.90 0.82 38 149.1 ⫺2.76 ⫺12 190.0 0.0 25 959.1 ⫺14.53
Mar-97 25.92 0.08 38 065.6 ⫺0.22 ⫺13 960.0 0.0 24 105.6 ⫺7.14
Apr-97 26.03 0.42 37 320.1 ⫺1.96 ⫺13 760.0 0.0 23 560.1 ⫺2.26
May-97 25.84 ⫺0.73 33 307.6 ⫺10.75 ⫺28 010.0 0.0 5 297.6 ⫺77.51
Jun-97 25.75 ⫺0.35 32 353.0 ⫺2.87 ⫺29 510.0 0.0 2 843.0 ⫺46.33
Jul-97 30.16 17.13 30 424.2 ⫺5.96 ⫺29 280.0 0.0 1 144.2 ⫺59.75
Aug-97 32.41 7.46 25 938.6 ⫺14.74 ⫺23 460.0 1 729.0 749.6 ⫺34.49
Sep-97 36.27 11.91 29 612.2 14.16 ⫺23 380.0 1 644.2 4 588.0 512.10
Oct-97 37.55 3.53 31 287.2 5.66 ⫺24 430.0 1 804.8 5 052.4 10.12
Nov-97 39.30 4.66 26 253.6 ⫺16.09 ⫺18 280.0 1 663.6 6 310.0 24.89
Dec-97 45.29 15.24 26 967.7 2.72 ⫺18 010.0 2 534.0 6 423.7 1.80
Jan-98 53.71 18.59 26 724.3 ⫺0.90 ⫺17 420.0 2 491.8 6 812.5 6.05
Feb-98 46.30 ⫺13.80 26 156.1 ⫺2.13 ⫺16 340.0 2 245.1 7 571.0 11.13
Mar-98 41.33 ⫺10.73 27 680.0 5.83 ⫺15 740.0 2 504.6 9 435.4 24.62
Apr-98 39.48 ⫺4.48 29 530.5 6.69 ⫺15 630.0 2 638.3 11 262.2 19.36
May-98 39.14 ⫺0.87 27 450.5 ⫺7.04 ⫺13 370.0 2 740.6 11 339.9 0.69
Jun-98 42.36 8.25 26 571.7 ⫺3.20 ⫺12 010.0 2 793.2 11 768.5 3.78
Jul-98 41.19 ⫺2.77 26 776.3 0.77 ⫺11 330.0 2 763.6 12 682.7 7.77
Aug-98 41.58 0.94 26 678.8 ⫺0.36 ⫺10 170.0 2 840.9 13 667.9 7.77
Sep-98 40.41 ⫺2.80 27 290.8 2.29 ⫺9 700.0 2 928.1 14 662.7 7.28
Oct-98 38.14 ⫺5.62 28 482.1 4.37 ⫺8 400.0 2 981.5 17 100.6 16.63
Nov-98 36.46 ⫺4.40 28 891.4 1.44 ⫺7 600.0 3 000.8 18 290.6 6.96
Dec-98 36.25 ⫺0.58 29 535.9 2.23 ⫺6 600.0 3 277.0 19 658.9 7.48
Jan-99 36.59 0.94 29 013.1 ⫺1.77 ⫺5 300.0 3 215.3 20 497.8 4.27
Feb-99 37.06 1.28 28 721.4 ⫺1.01 ⫺5 000.0 3 166.5 20 554.9 0.28
Mar-99 37.51 1.21 29 936.1 4.23 ⫺4 600.0 3 132.2 22 203.9 8.02
Apr-99 37.60 0.24 30 203.8 0.89 ⫺3 900.0 3 207.6 23 096.2 4.02
May-99 37.02 ⫺1.54 30 637.2 1.43 ⫺3 500.0 3 230.7 23 906.5 3.51
Jun-99 36.91 ⫺0.30 31 433.9 2.60 ⫺3 300.0 3 336.7 24 797.2 3.73
Jul-99 37.11 0.54 31 928.8 1.57 ⫺3 200.0 3 408.3 25 320.5 2.11
Aug-99 37.98 2.34 32 216.1 0.90 ⫺3 200.0 3 453.3 25 562.8 0.96
Sep-99 39.88 5.00 32 360.2 0.45 ⫺3 000.0 3 565.4 25 794.8 0.91

owing

Oct-99 39.47 ⫺1.03 32 438.1 0.24 ⫺3 000.0 3 384.5 26 053.6 1.00

Nov-99 38.77 ⫺1.77 32 842.1 1.25 ⫺3 100.0 3 446.6 26 295.5 0.93
Dec-99 38.18 ⫺1.53 34 780.6 5.90 ⫺4 800.0 3 367.9 26 612.7 1.21
Jan-00 37.35 ⫺2.16 32 630.2 ⫺6.18 ⫺3 200.0 3 394.9 26 035.3 ⫺2.17
Feb-00 37.71 0.96 31 953.8 ⫺2.07 ⫺2 700.0 3 369.7 25 884.1 ⫺0.58
Mar-00 37.90 0.50 32 283.9 1.03 ⫺2 700.0 3 358.9 26 225.0 1.32
Apr-00 37.97 0.18 32 166.0 ⫺0.37 ⫺2 600.0 3 302.2 26 263.8 0.15
May-00 38.95 2.58 31 904.2 ⫺0.81 ⫺2 200.0 3 314.0 26 390.2 0.48
Jun-00 39.06 0.28 32 142.0 0.75 ⫺2 100.0 3 348.5 26 693.5 1.15
Jul-00 40.22 2.97 31 929.6 ⫺0.66 ⫺2 100.0 3 383.8 26 445.8 ⫺0.93
Aug-00 40.87 1.62 32 232.4 0.95 ⫺2 100.0 3 262.7 26 869.7 1.60
Sep-00 41.88 2.47 32 249.8 0.05 ⫺2 100.0 3 270.6 26 879.2 0.04
Oct-00 43.21 3.18 32 244.7 ⫺0.02 ⫺2 100.0 3 253.3 26 891.4 0.05
Nov-00 43.73 1.20 32 316.1 0.22 ⫺2 100.0 3 022.9 27 193.2 1.12
Dec-00 43.09 ⫺1.46 32 661.3 1.07 ⫺2 100.0 3 074.4 27 486.9 1.08
Jan-01 43.12 0.07 32 795.0 0.41 ⫺2 100.0 3 009.4 27 685.6 0.72
Feb-01 42.64 ⫺1.11 33 153.9 1.09 ⫺2 100.0 2 858.5 28 195.4 1.84
Mar-01 43.90 2.95 32 294.7 ⫺2.59 ⫺2 200.0 2 732.2 27 362.5 ⫺2.95
Apr-01 45.46 3.55 32 095.9 ⫺0.62 ⫺2 100.0 2 694.5 27 301.4 ⫺0.22
Source: Bank of Thailand.
criticized the probit/logit and the signals approaches for possible misclassi-
fication error in the construction of crisis dummy variables. They pointed out
several issues regarding these techniques: the need for a priori dating of crisis
occurrence, the use of arbitrary thresholds and inadequate modeling of the
dynamics in the systems, for example. The Markov-switching model of
exchange rate fluctuations with time-varying transition probabilities was used
as a predictive model of currency crises in their study.
3 DEFINITION OF CURRENCY CRISIS AND

METHODOLOGIES FOR IDENTIFYING ITS
LEADING INDICATORS
From the summary of previous studies presented in Tables 2.1 and 2.2,
we can see that researchers use either the multivariate logit/probit model or
the signals approach in analyzing crisis (except for the study by Mariano
et al.). The most commonly used approach seemed to be the estimation of
logit/probit models, and Kaminsky et al. (1998) are the first to use the signals
approach. For this research study, both methodologies are employed. The
sample period is restricted by the availability of monthly data needed for this
study, and covers the period January 1992 to December 2000.5
3.1 Definition of Currency Crisis in this Study
The definition of currency crisis used in Engwatana (1999) and

Poonpatpibul and Ittisupornrat (2001) focused on what happened to the
value of currency (including its forward premium in Engwatana). It is the
author’s opinion that his definition is not broad enough as speculative
attacks may not change the value of the currency by much if the monetary
authority could defend it with their international reserves. Furthermore,
depletion of international reserves could lead to many other economic
problems including the crisis of confidence. Therefore, we should regard the
huge reduction of international reserves as an indication of currency crisis
as well. Kaminsky et al. (1998) defined a currency crisis as ‘a situation in
which an attack on the currency leads to a sharp depreciation of the cur-
rency, a large decline in international reserves, or a combination of the two’.
From this definition, an index of exchange market pressure is constructed
as a weighted average of monthly percentage changes in the exchange rate
and the negative of monthly percentage changes in the gross international
reserves.
In the case of Thailand, pressures on the exchange rate started around
the end of 1996 and the Bank of Thailand (BOT) intervened to maintain
the value of the baht. However, the amount of the BOT’s selling of dollars
to support the value of the baht in the market did not show up as a large
depletion in the international reserves because it used the swap operation
to buy back dollars (with an obligation to sell back dollars in the future).
Some BOT officials argued that the swap operation helped ease the liq-
uidity dry-up, which would have occurred without it. We can see from
Table 2.3 that a series of speculative attacks in the first half of 1997
resulted in the BOT’s swap obligation of a huge amount, from around
4.890 billion dollars in December 1996 with gross international reserves
standing at 38.725 billion dollars to a swap obligation of 29.510 billion
dollars in June 1997. The gross international reserves at the end of June
1997 stood at 32.353 billion dollars while the net international reserves
stood at only 2.843 billion dollars. The depletion in gross reserves did not
look as severe as that in the net reserves because the swap operation con-
cealed the problem by postponing it into the future. Therefore, the crisis
index in this study will be based on the movements in net reserves and
exchange rates.
An index of currency market turbulence (I) for Thailand was constructed
based on the formula:
I ⫽ ⌬e ⌬R ␴e
e ⫺ R *␴ r
(2.1)
where ⌬e ⫽change in nominal exchange rate

⌬R ⫽change in net international reserves
␴e ⫽standard deviation of ⌬e/e
␴r ⫽standard deviation of ⌬R/R
By defining a currency crisis as a situation when the index I exceeds its

mean plus 2 standard deviations, the observations of crisis came to only 3.
And if we receded to defining the threshold at mean plus 1.5 standard devi-
ations, the observations on crisis still came to only 4. (See Appendix 2A,
Table 2A.1: Index of currency market turbulence.) Such numbers may
be adequate for the signals analysis but they are inadequate for the probit
estimate which requires a minimum number of dependent variables not
being equal to 0 (or Y ⫽1.0 when crisis occurs). Since either the severe fall
in exchange rate or the excessive depletion of reserves can be a result of
speculative attacks, it is decided that a currency crisis is a situation where
one of the following situations takes place:
1. there is an accumulated three-month depreciation in exchange rate of

15 percent or more; or
2. there is an accumulated three-month depletion in net international
reserves of 15 percent or more.
With the above definition, the Thai economy fell within the episode of cur-
rency crisis during January 1997 up to February 1998. This gives us a total
of 14 monthly observations of crisis from our sample of 108 observations
during January 1992 and December 2000. However, for some indicators
their year-on-year changes are used as signals, which means the first 12
months of the sample are lost.
3.2 Set of Possible Leading Indicators
From a comprehensive survey of empirical studies on leading indicators of

currency crises, Kaminsky et al. (1998) found a large variety of indicators
used among those studies – as many as 105 indicators were used. However,
not all of them passed the statistical significance test. The choice of vari-
ables used in this study was dictated not only by economic reasoning but
also by data availability on a monthly basis in Thailand. The set of poten-
tial leading indicators with available monthly observations during our
sample period may be grouped as follows:
● current account variables:

export growth;
import growth;
trade balance (as a ratio to GDP);
current account balance (as a ratio to GDP);
terms of trade (growth);
real exchange rate (deviation from past average or trend);
● capital account variables:
spread between domestic and foreign interest rates;
ratio of external debt to total debt of financial institutions;
short-term external debt to international reserves;
capital account balance (growth);
● financial variables:
real deposit rate;
spread between lending and deposit rates;
ratio of M2 to international reserves (growth);
ratio of domestic credit to GDP (growth);
excess real M1 balance;
● real sector variables:
fiscal balance (as a ratio to GDP);
inflation rate;
GDP growth;
Change in stock prices (represented by SET index).
It should be pointed out that Thailand does not have monthly GDP data
and the quarterly GDP data go back only to 1993. Since several indica-
tors above should be measured as a ratio to GDP, we need to generate
monthly GDP data from the quarterly data. This is done by making use
of the quarterly relationship between GDP and other variables, which
are themselves available in monthly series. The estimated quarterly rela-
tionship between GDP and exports, indirect taxes, government expendi-
tures and electricity consumption is used to estimate monthly GDP from
the monthly data of these variables. The estimated monthly GDP is
adjusted so as to make the sum of the estimated monthly series equal to
the actual quarterly data. The estimated monthly GDP series are also
used to estimate the demand for money in order to calculate excess real
M1 balance. In addition, the short-term external debt is also available on

a quarterly basis until 1999 when monthly data became available. The
monthly short-term debt is estimated by interpolation between two adja-
cent quarters.
3.3 Methodology
There are two popular approaches in the analysis of leading indicators for
currency crisis. The parametric approach utilizes the qualitative depend-
ent variable regression models (probit, logit) to identify leading indica-
tors. The non-parametric approach uses the signals analysis proposed by
Kaminsky et al. (1998). This study applied both the signals analysis and
the probit estimates to identify the leading indicators of currency crisis for
Thailand.
The signals analysis starts with the ‘signaling horizon’ of 24 months as
used in Kaminsky et al. (1998) and also performs a test on a 12-month sig-
naling horizon to see if the results are sensitive to the choice of the horizon.
The threshold value for each indicator is scanned between the 10–25 per-
centiles6 of the indicator’s distribution and the ‘optimal’ threshold is the
one that minimizes the adjusted ‘noise-to-signal’ ratio where ‘noise’ and
‘signal’ can be defined in the following matrix.
Event (within 12 or 24 months)

Crisis No crisis
Signal is issued A (good signal) B (bad signal or noise)
No signal is issued C D
BⲐ(B ⫹ D)
Adjusted noise᎐to᎐signal ratio ⫽
AⲐ(A ⫹ C)
The parametric approach employed in this study is the probit estimates.

One advantage of this approach is that it summarizes information from all
significant variables in one useful number, the probability of a crisis.
However, it also has some drawbacks as the estimates may have to exclude
some important indicators due to multicollinearity problems. Its summary
nature, on the other hand, makes it unclear which variables are sending out
the alarm unless we keep a close watch on the movement of all significant
variables.
4 EMPIRICAL RESULTS
4.1 Crisis Indicators from Signals Approach
Based on the signaling horizon of 24 months and the scan between the
10–25th percentiles, the optimal threshold for each indicator is found and
presented in Table 2.4 together with its adjusted noise-to-signal ratio. A
change of signaling horizon to 12 months is also performed to gauge the
sensitivity of our indicators and the information is presented in Table 2.5. It
can be seen that for most indicators, changing the signaling horizon does not
significantly affect their threshold levels and the adjusted noise-to-signal
Table 2.4 Performance of indicators based on 24-month signaling horizon
Indicator Threshold Adjusted Noise-

(%) to-signal Ratio
Current account
• Import growth in US$ (%) 30.2 0.45
• Export growth in US$ (%) ⫺6.9 0.95
• Ratio of current account to GDP ⫺8.1 0.13
• Ratio of trade balance to GDP ⫺9.3 0.18
• Terms of trade growth ⫺8.6 0.26
• Real exchange rate misalignment (deviation ⫺5.9 0.06
from previous 60-month average)
Capital account
• Difference between domestic (MLR) and 7.8 0.48
foreign interest rates (LIBOR in $)
• Ratio of private short-term external 127.0 0.00
debt to international reserves
• Ratio of total short-term external debt to 126.0 0.04
Financial variables
• Spread between lending and deposit rates 9.6 0.95
• Growth of M2/international reserves 11.4 0.53
• Growth of domestic credit/GDP 14.8 0.43
• Percentage excess real M1 balance 6.6 0.47
• Growth of money multiplier (M2) 8.5 1.18
Real sector
• Real GDP growth ⫺1.0 1.05
• Ratio of fiscal balance to GDP ⫺3.4 1.01
• Growth of stock prices (SET index) ⫺42.5 0.29
• Inflation rate 6.5 0.42
Table 2.5 Performance of indicators based on 12-month signaling horizon
Indicator Threshold Adjusted Noise-

(%) to-signal Ratio
Current account
• Import growth in US$ (%) * *
• Export growth in US$ (%) ⫺3.8 0.55
• Ratio of current account to GDP ⫺8.3 0.42
• Ratio of trade balance to GDP ⫺8.5 0.52
• Terms of trade growth ⫺8.6 0.25
• Real exchange rate misalignment (deviation ⫺5.3 0.85
from previous 60-month average)
Capital account
• Difference between domestic (MLR) and 7.5 0.52
foreign interest rates (LIBOR in $)
• Ratio of private short-term external 138.0 0.05
• Ratio of total short-term external debt to 128.4 0.13
Financial variables
• Spread between lending and deposit rates 9.6 0.51
• Growth of M2/international reserves 11.4 0.28
• Growth of domestic credit/GDP 15.0 0.42
• Percentage excess real M1 balance 3.6 0.69
• Growth of money multiplier (M2) 9.6 0.68
Real sector
• Real GDP growth ⫺1.0 0.57
• Ratio of fiscal balance to GDP ⫺3.4 0.55
• Growth of stock prices (SET index) ⫺42.5 0.15
• Inflation rate 6.5 0.42
Note: * No signals were found between the 10–25th percentiles with the 12-month horizon
for import growth.
ratios, with the exception of import growth M2 money multiplier, real GDP
growth and the ratio of fiscal balance to GDP.
Since the performance of these indicators are indicated by their adjusted
noise-to-signal ratio, Tables 2.6 and 2.7 rank these indicators from low to
high ratios. It can be noticed that the ratio of trade balance to GDP is not
present in these two tables because its effect is already included within the
ratio of current account to GDP. The same goes for the ratio of private short-
term external debt to international reserves, the effect of which is already
included in the ratio of total short-term external debt to international
Table 2.6 Indicators ranked according to their performance (24-month

horizon)
Rank Threshold Adjusted Noise-

(%) to-signal Ratio
1. Ratio of total short-term external 126.0 0.04
2. Real exchange rate misalignment ⫺5.9 0.06
3. Ratio of current account to GDP ⫺8.1 0.13
4. Terms of trade growth ⫺8.6 0.26
5. Growth of stock prices ⫺42.5 0.29
6. Inflation rate 6.5 0.42
7. Growth of domestic credit/GDP 14.8 0.43
8. Import growth (in US$) 30.2 0.45
9. Percentage excess real M1 balance 6.6 0.47
10. Difference between domestic and foreign 7.8 0.48
interest rates (MLR/LIBOR in $)
11. Growth of M2/international reserves 11.4 0.53
12. Export growth (in US$) ⫺6.9 0.95
13. Spread between lending and deposit rates 9.6 0.95
14. Ratio of fiscal balance to GDP ⫺3.4 1.01
15. Real GDP growth rate ⫺1.0 1.05
16. Growth of money multiplier (M2) 8.5 1.18
Note: The indicators ‘Trade balance/GDP’ and ‘Private short-term external debt/
international reserves’ are excluded from this table as their effect is implicit in ‘Ratio of
current account/GDP’ and ‘Ratio of total short-term external debt/international reserves’
respectively.
reserves. In fact, during the period covered in this study the private short-
term external debt is very close to the total short-term external debt because
the public short-term external debt was insignificantly small in comparison
to that of the private sector. It would be redundant to have duplicate meas-
ures of the same effect.
Comparing the rank and the adjusted noise-to-signal ratio of the indi-
cators in Tables 2.6 and 2.7, we can see that there are a few indicators the
performance of which is rather sensitive to the choice of signaling horizon.
Based on an evaluation of the magnitude change in threshold levels and in
the noise-to-signal ratio, the following two indicators appear to be rather
sensitive to the choice of horizon: import growth and the growth of money
multiplier. For import growth, no signals were found for the 12-month sig-
naling horizon. For growth of the money multiplier, the threshold and the
adjusted noise-to-signal ratio changed from 8.5 and 1.18 for the 24-month
Table 2.7 Indicators ranked according to their performance (12-month

horizon)
Rank Threshold Adjusted Noise-

(%) to-signal Ratio
1. Ratio of total short-term external 128.4 0.13
2. Growth of stock prices ⫺42.5 0.15
3. Terms of trade growth ⫺8.6 0.25
4. Growth of M2/international reserves 11.4 0.28
5. Ratio of current account to GDP ⫺8.3 0.42
6. Growth of domestic credit to GDP 15.0 0.42
7. Inflation rate 6.5 0.42
8. Spread between lending and deposit rates 9.6 0.51
9. Difference between domestic and foreign 7.8 0.52
interest rates (MLR/LIBOR in $)
10. Export growth (in US$) ⫺3.8 0.55
11. Ratio of fiscal balance to GDP ⫺3.4 0.55
12. Real GDP growth ⫺1.0 0.57
13. Growth of money multiplier (M2) 9.6 0.68
14. Percentage excess real M1 balance 3.6 0.69
15. Real exchange rate misalignment ⫺5.3 0.85
16. Import growth (in US$) – –
Note: The indicators ‘Trade balance/GDP’ and ‘Private short-term external debt/
international reserves’ are excluded from this table as their effect is implicit in ‘Ratio of
current account/GDP’ and ‘Ratio of total short-term external debt/international reserves’
respectively.
horizon to 9.6 and 0.68 for the 12-month horizon. The changes in thresh-
olds and noise-to-signal ratios of other variables are not so large as to cause
concern.
From the author’s judgment, the following indicators appear to be least
sensitive to the change of signaling horizon:
● the ratio of total short-term external debt to international reserves;

● terms of trade growth;
● growth of stock prices;
● inflation rate.
The remaining indicators appear to be sensitive but their sensitivity appears

moderate and does not warrant as much concern as the import growth and
the growth of money multiplier.
It is the opinion of the author that a signaling horizon of 24 months may

be rather extended as an irregular movement over the threshold at a certain
date to be read as a signal for crisis in the following 24 months. Signals
obtained from a 12-month horizon analysis may be better if we take the
viewpoint that the shorter time span between the occurrence of signals and
a possible impending crisis may compel us to be more alert about tracking
the monthly movement of these indicators. Therefore, the following dis-
cussion will be based on the results from the 12-month horizon.
Based on the 12-month signaling horizon we can see from Table 2.7 that
there are 15 variables that can serve as leading indicators of currency crisis
for Thailand. From historical experience they have issued relatively fewer
bad signals (noise) than good ones as evidenced by their adjusted noise-to-
signal ratio of less than 1. Their movement can be seen in Figures 2.2 and
2.3. However, upon detailed examination of the data’s monthly movement,
it was found that for some variables their signals were issued after July 1997.
Such observation makes it unclear whether their abnormal behavior was in
fact caused by the crisis and not vice versa. These are mainly some of the
financial variables: spread between lending and deposit rates, difference
between domestic and foreign interest rates, growth of money multiplier
and excess real M1 balance. Eliminating these from the list of our leading
indicators for currency crises, the remaining 11 indicators are deemed to
have satisfactory performance. They are grouped here as indicators with
positive and negative shocks.
Indicators with positive shocks are those whose increases over a threshold
may lead to or add to the possibility of a currency crisis. From the signals
analysis, there are four such indicators:
● ratio of total short-term external debt to international reserves;

● growth of M2/international reserves;
● growth of domestic credit/GDP;
● inflation rate.
The ratio of total short-term external debt to international reserves reflects

a constraint on the country’s liquidity in international transactions: the
higher the ratio, the more likely the crisis. The threshold level for this ratio
with minimum adjusted noise-to-signal ratio is 28.4 percent, found at the
21st percentile. So if the amount of total short-term external debt is 28.4
percent higher than reserves, it should trigger concern from the policy-
maker. This indicator is very important from at least two perspectives:
mismatch in maturity and mismatch in denomination of currencies. It is
well known that financial institutions including commercial banks face
the maturity mismatch of borrowing short and lending long. This creates a
Growth of stock prices (SET index)
120
100
80
60
40
20
0
–20
–40
–60
–80
Jan-93 Jan-94 Jan-95 Jan-96 Jan-97 Jan-98 Jan-99 Jan-00
90
Growth of terms of trade

20
15
10
5
0
–5
–10
–15
Growth of exports (in US$)
40
30
20
10
0
–10
–20
91
Ratio of current account to GDP

20
15
10
5
0
–5
–10
–15
Figure 2.2 Movement of indicators with negative shocks (threshold from 12-month horizon)
Real GDP growth rate
20
15
10
5
0
–5
–10
–15
–20
92
Ratio of fiscal balance to GDP

15
10
–5
–10
–15
Real exchange rate misalignment (deviation from previous 60-month average)
100
80
60
40
93
20
–20
Figure 2.2 (continued)

Ratio
Ratioofoftotal
totalshort-term
short-termexternal
externaldebt
debttotointernational
internationalreserves
reserves
180
180
160
160
140
140
120
120
100
100
8080
6060
4040
2020
00
Jan-93
Jan-93 Jan-94
Jan-94 Jan-95
Jan-95 Jan-96
Jan-96 Jan-97
Jan-97 Jan-98
Jan-98 Jan-99
Jan-99 Jan-00
Jan-00
94
Growth of ratio of M2 to international reserves

50
40
30
20
10
0
–10
–20
–30
Growth
Growthof
ofratio
ratioof
ofdomestic
domesticcredit
creditto
toGDP
GDP
40
30
20
10
0
–10
–20
–30
Jan-93 Jan-94
Jan-94 Jan-95
Jan-95 Jan-96
Jan-96 Jan-97
Jan-97 Jan-98
Jan-98 Jan-99
Jan-99 Jan-00
Jan-00
95
Inflation rate
12
10
8
6
4
2
0
–2
Figure 2.3 Movement of indicators with positive shocks (threshold from 12-month horizon)
Difference
Differencebetween
betweendomestic
domesticand
andforeign
foreigninterest
interestrate
rate(MLR/LIBOR
(MLR/LIBORin
inUS$)
US$)
12
12
10
10
88
66
44
22
00
Jan-93
Jan-93 Jan-94
Jan-94 Jan-95
Jan-95 Jan-96
Jan-96 Jan-97
Jan-97 Jan-98
Jan-98 Jan-99
Jan-99 Jan-00
Jan-00
96
Spread between lending and deposit rates

12
10
8
6
4
2
0
Growth of money multiplier (M2)
40
30
20
10
0
–10
–20
–30
Percentage excess real M1 balance

97
35
30
25
20
15
10
5
0
–5
–10
–15
–20
Figure 2.3 (continued)

liquidity problem when there is a bank-run even if banks are solvent. In

addition, when the short-term debts are borrowed from foreign lenders the
country inevitably faces the currency mismatch problem since developing
countries are unable to borrow abroad in their own currencies. Therefore
the high ratio of short-term external debt to international reserves erodes
investors’ confidence and incurs high probability of both banking crisis and
currency crisis.
The growth of M2/international reserves indicates the rate of monetary
expansion in the economy, which could create a harmful inflationary envir-
onment. This ratio could also grow due to the decline in foreign exchange
reserves. The threshold level for the growth of this ratio with minimum
adjusted noise-to-signal ratio is 11.4 percent, found at the 12th percentile.
The growth of domestic credit/GDP reflects the rate of credit expansion
relative to real economic activities. If this growth is excessive, it can lead to
bubbling asset prices and inflationary demand due to the wealth effect. The
threshold with minimum adjusted noise-to-signal ratio of this growth rate
is 15 percent, found at the 20th percentile.
The inflation rate may be a result of excessive expansionary policies, or
a rapid increase in demand, or from higher imported prices of inputs.
Whatever its cause, its excessive rate erodes the competitiveness of the
country and increases the vulnerability to crisis. The threshold of inflation
rate with minimum adjusted noise-to-signal ratio is 6.5 percent, found at
the 24th percentile.
Indicators with negative shocks are those whose decreases below a thresh-
old may lead to or add to the possibility of a currency crisis. There are seven
such indicators:
● growth of stock prices;

● terms of trade growth;
● export growth;
● ratio of current account to GDP;
● real GDP growth rate;
● ratio of fiscal balance to GDP;
● real exchange rate misalignment.
The negative growth in stock prices reflects lower expectation of future

earnings and lower confidence in the domestic economy. The stock prices
as measured by the SET index is also found to be an important component
of Thailand’s composite leading indicator by Tinakorn (1998). The thresh-
old with minimum adjusted noise-to-signal ratio for the SET index is –42.5
percent, found at the 18th percentile. This number appears to be of a rather
large magnitude if the SET index hovers around 200–300 points as at
present. But during the sample period, the range of the SET index was
between 1528.83 (October 1994) and 214.3 (August 1998) and the rate of
year-on-year decline went up to almost 60 percent for some months.
The terms of trade is the ratio of export price over import price. A
decline in this ratio means imports are relatively more expensive than
exports, which will have a negative impact on the trade and current
account, and ceteris paribus, the balance of payments. In a cross-country
study by Kaminsky and Reinhart (1999), it is found that crises are preceded,
on average, by a deterioration of the terms of trade with an annual decline
that is about 10 percent deeper than those observed in tranquil times before
a balance-of-payments crisis. For the case of Thailand, the threshold of the
annual decline in the terms of trade with the minimum adjusted noise-to-
signal ratio is –8.6 percent, found at the 10th percentile.
Exports account for more than 50 percent of GDP in Thailand.
Therefore its decline has grave implications for the real sector as well as the
position of the current account and the balance of payments. The thresh-
old with the minimum adjusted noise-to-signal ratio is the export growth at
–3.8 percent,7 found at the 20th percentile.
The current account includes the international exchange of both goods
and services and the current account deficit has a negative impact on the
foreign exchange earnings. This variable is measured as a ratio to GDP and
its threshold with minimum adjusted noise-to-signal ratio is found to be
–8.3 percent, at the 15th percentile meaning that current account deficit
that runs in excess of 8.3 percent of GDP is a warning for increased vul-
nerability of the economy.
Both the real GDP growth rate and the ratio of fiscal balance to GDP
are found to be rather sensitive to the choice of signaling horizon. Their
performance improves when the 12-month horizon is used. The threshold
for GDP growth rate is –1.0 percent, at the 25th percentile and that for the
ratio of fiscal balance to GDP is –3.4 percent, at the 19th percentile.
Although the deterioration in these two indicators increases the economy’s
vulnerability, it is also the case that these two indicators tend to deteriorate
after the onset of crisis if the currency crisis evolves into economic crisis.
The evolution of real exchange rate has a significant implication for the
country’s competitiveness. Kaminsky and Reinhart (1999) found that
during the year before the balance-of-payments and banking crises, the real
exchange rate shows evidence of being overvalued. This is also the case for
Thailand. In this study, the real exchange rate is measured in terms of baht
per US dollar adjusted by the ratio of US consumer price index over Thai
consumer price index. Therefore a decline in this variable means an appre-
ciation of the baht, which will have a negative impact on export earnings
and increase the vulnerability of the economy. The misalignment of the real
exchange rate is measured as a deviation from its previous 60-month

average. The threshold of this deviation with minimum adjusted noise-to-
signal ratio is –5.3 percent, at the 25th percentile.
It should be remarked here that if it were not due to the limitation of data
the exchange rate indicator to be used ought to be the real effective
exchange rate. By using the real exchange rate of the baht vis-à-vis the US
dollar only, one may lose information coming from the movement of other
currencies. For example, the Chinese adjustment of their exchange rate
regime in 1994 (by merging the official rate and the swap market rate to
produce a single exchange rate) in effect created a devaluation in the ren-
minbi from 5.8 yuan per US dollar to 8.45 yuan per US dollar. This must
have put a lot of pressure on the competitiveness of Asian countries includ-
ing Thailand. We can see in Figure 2.2 that Thailand’s export growth began
to decline since the beginning of 1995 and hit negative growth rates in many
months of 1996 but the real exchange rate misalignment did not look as
severe since it was based on the US dollar.
4.2 Composite Indicator and Probability of Crisis from Signals Approach
One way of combining the signals sent out by all the above 11 indicators is
simply by counting the number of individual indicators that have crossed
their threshold in a particular month as in equation (2.2):
n
It(1) ⫽ 兺Stj
j⫽1
(2.2)
where Sjt ⫽ 1 if variable j crosses the threshold in period t

Sjt ⫽ 0, otherwise.
However, It(1) does not take into account the fact that each variable has
different forecasting accuracy as depicted by its adjusted noise-to-signal
ratio. For example, the low noise-to-signal ratio of 0.1 of a variable Z con-
tains information that it has more forecasting accuracy than another vari-
able Y with a noise-to-signal ratio of 0.5. Therefore, the composite
indicator should give more weight to the signal sent out by Z than that by
Y. This leads us to a weighted composite indicator based on the adjusted
noise-to-signal ratio of each variable.
n
It(2) ⫽ 兺Stj · w1j
t⫽1
(2.3)
where w j ⫽ the adjusted noise᎐to᎐signal ratio of variable j.

Kaminsky (1999) also experimented with two other composite indica-

tors, and found the weighted composite indicator presented in equation
(2.3) to perform the best when predicting both currency and banking crisis.8
In this study, the weighted composite index (It(2) ) is computed and used to
calculate the conditional probability of currency crises in the formula:
(No. of months the composite index ⫽ X and

currency crisis actually occurred within 12 months)
Prob (crisisⲐI ⫽ X) ⫽
No. of months the composite index ⫽ X
where X is the value of the composite index. It was found that, in general,
the increase in the composite indicator results in higher conditional prob-
ability of crisis. However, there were a few cases where the value of the com-
posite index (X) was high but crisis did not actually occur within 12 months
and vice versa. This can be seen in Figure 2.4 (a), which plots the condi-
tional probability of a crisis, given the value of the composite index. To
obtain a smoother increasing function of probability as the composite
index increases requires some grouping of the composite index. It appears
that a grouping with a range of six (Figure 2.4f) renders a smooth increas-
ing function of probability as the composite index increases. The result of
such grouping is presented in Table 2.8.
Based on the above conditional probability we may look at the value
of the composite index computed from those 11 indicators and predict
that the probability of a crisis in the next 12 months is about 0.61 if the
value exceeds 12. And if the value of the composite index exceeds 18, it is
most likely that a crisis is coming as the conditional probability is equal
to one.
4.3 Probit Estimates
The set of indicators used as explanatory variables in the probit estimation

is the same as that used for signals analysis. However, all of the indicators
cannot be simultaneously put into an estimation as the presence of multi-
collinearity among some variables results in large variances and insignificant
t-statistics. Two of the most acceptable sets of estimates are shown in Tables
2.10, and 2.11. These two models are chosen based on the signs of coefficients
that conform to economic reasoning, and also on their statistical significance
and predictive performance. There appears to be a common set of core vari-
ables that can explain the probability of crisis. Their coefficients have the
correct (expected) signs and the coefficients are statistically significant at the
95 percent confidence level (one-tail test). These variables are summarized
below in Table 2.9:
(a) Range of composite index = 1 (b) Range of composite index = 2
1.00 1.00
0.90 0.90
0.80 0.80
0.70 0.70
0.60 0.60
0.50 0.50
0.40 0.40
0.30
0.20 0.30
0.10 0.20
0.00 0.10
0.00–1.00
3.01–4.00
6.01–7.00
9.01–10.00
12.01–13.00
15.01–16.00
18.01–19.00
21.01–22.00
24.01–25.00
0.00
0.00–2.00
4.01–6.00
8.01–10.00
12.01–14.00
>16.00
102
(c) Range of composite index = 3 (d) Range of composite index = 4

1.00 1.00
0.90 0.90
0.80 0.80
0.70 0.70
0.60 0.60
0.50 0.50
0.40 0.40
0.30 0.30
0.20 0.20
0.10 0.10
0.00 0.00
0.00–3.00
3.01–6.00
6.01–9.00
9.01–12.00
12.01–15.00
15.01–18.00
>18
0.00–4.00
4.01–8.00
8.01–12.00
12.01–16.00
>16
(e) Range of composite index = 5 (f) Range of composite index = 6
1.00 1.00
0.90 0.90
0.80 0.80
0.70 0.70
0.60 0.60
0.50 0.50
0.40 0.40
103
0.30 0.30
0.20 0.20
0.10 0.10
0.00 0.00
0.00–5.00
5.01–10.00
10.01–15.00
15.01–20.00
>20
0.00–6.00
6.01–12.00
12.01–18.00
>18
Figure 2.4 Plot of conditional probability of crisis given the value of the composite index
Table 2.8 Composite index and conditional probability of currency crisis
Value of No. of No. of Months the Index ⫽ Conditional

Composite Months the X and Currency Crisis Probability of
Index (X) Index ⫽X Occurred within 12 Months Currency Crisis
0.00–6.00 49 4 0.08
6.01–12.00 13 6 0.46
12.01–18.00 18 11 0.61
⬎18.00 4 4 1.00
*Quadratic probability score (QPS) ⫽0.266
T
1
Note: * QPS ⫽ T 兺2(Pt ⫺ Rt)2
t⫽1
Where Pt ⫽ probability of crisis between t and t ⫹h conditional on

information provided by the composite indicator
Rt ⫽ realization that equals 1 if a crisis occurs between t and
t ⫹h and equals 0 otherwise
h ⫽ time horizon, for example, 12 months.
QPS ranges from 0 to 2, with a score of 0 corresponding to perfect accuracy. See Kaminsky
(1999).
Table 2.9 Core variables
Significant Indicators Model 1 Model 2

1. Growth rate of the ratio M2/reserves 冪冪
2. Deviation of real exchange rate 冪冪
3. Ratio of current account to GDP 冪冪
4. Growth of stock price index 冪冪
5. Growth of real GDP 冪 –
6. Ratio of fiscal balance to GDP 冪冪
7. Export growth – 冪
Even though these two models seem to be about equal on statistical

grounds, the frequencies of actual and predicted outcomes appear to be in
favor of Model 2. Based on a cut-off probability of 0.5, Model 2 made a
correct prediction of 1’s for 13 observations out of 14 outcomes during the
crisis period and made an incorrect prediction of 1’s for seven observations
during the tranquil period. On the other hand, Model 1 made fewer
numbers of correct predictions for crisis and more numbers of incorrect
predictions during the tranquil time.
In the currency crisis literature, contagion effect has received consider-

able attention from academics. Kaminsky and Reinhart (2000) examined
both trade links and financial sector links and found that susceptibility to
contagion is highly non-linear. While it would be desirable to examine
whether Thailand was affected by contagion, the available data do not
permit us to test this in probit estimates. First of all, Thailand was among
the primary countries to experience the 1997 currency crisis. Secondly,
although contagion from the Tequila Crisis of 1994–95 was possible for
Thailand through links with the United States, data conditions in probit
models do not allow us to obtain a regression estimate with a contagion
dummy.9
If we regard the probit prediction of 1 as a signal for crisis, then we can
regard the incorrect signal as noise. With this perspective, the adjusted
noise-to-signal ratio can be calculated for each model and the ratios are
also presented in Tables 2.10 and 2.11. The adjusted noise-to-signal ratios
of Models 1 and 2 are, respectively, 0.145 and 0.108. Therefore, the in-
sample predictive performance appears to be in favor of Model 2. It should
also be pointed out that during the crisis period, the probability of a crisis
from probit estimation is very high and it is low during the non-crisis
period. The plots of the probability of crisis from Models 1 and 2 are pre-
sented in Figure 2.5.
4.4 Assessment of Future Vulnerability
We can use both the signals analysis and the probit estimates to assess the
future vulnerability of Thailand to currency crisis. However, the data
requirements of these two methods are somewhat different. From the
signals approach, we may use the present available data to compute the
index of currency crisis and make use of the conditional probability in
Table 2.8 to assess the likelihood of a crisis within the next 12 months.
Forecasts by the probit estimation, on the other hand, can be made for only
three months ahead since the lags of explanatory variables vary from three
to 12 months. So for the probit estimate, we only check its out-of-sample
predictive performance, given the data for 2001.
Table 2.12 presents the out-of-sample forecast for the probability of
a crisis from probit models during 2001. It can be seen that both models
predicted very low probability of a currency crisis during 2001. There
appeared to be a rise in this probability during the months of July,
September and October but the probability dropped afterwards. Overall,
the probability predicted by probit models was low and there was no cur-
rency crisis in 2001. If we want to assess future vulnerability, then we must
first forecast the values of the indicators in probit models, this is at present
Table 2.10 Probit estimate for currency crisis: Model 1 (1994.1–2000.12)
Variable Coefficient Std. Error b/Std.Err P[円Z円⬎z] x-bar

(partial
derivatives)
M2RESG6 0.0072 0.0034 2.133 0.0329 0.7728
DRER603 ⫺0.0137 0.0035 ⫺3.905 0.0001 8.8105
CAGDP6 ⫺0.0180 0.0082 ⫺2.201 0.0277 0.3084
SETG6 ⫺0.0026 0.0014 ⫺1.796 0.0726 ⫺2.3993
FBGDP12 ⫺0.0217 0.0101 ⫺2.145 0.0320 0.2302
RGDP6 ⫺0.0253 0.0078 ⫺3.226 0.0013 3.5784
Log-likelihood ⫺27.1393
Chi-squared 21.4158
Significance level 0.0007
Frequencies of actual and predicted outcomes (cut-off probability ⫽0.5)
Predicted Actual Total
1 (Crisis) 0 (No crisis)
1 (signal) 11 8 19
0 (no signal) 3 62 65
Total 14 70 84
Proportion of correct prediction ⫽0.869
Adjusted noise-to-signal ratio⫽0.145
Definition of variables
M2RESG6 ⫽ growth in the ratio of M2 over reserves (lagged six
months)
DRER603 ⫽ deviation of real exchange rate from previous
60-month average (lagged three months)
CAGDP6 ⫽ ratio of current account to GDP (lagged six months)
SETG6 ⫽ growth rate of stock price index (lagged six months)
FBGDP12 ⫽ ratio of fiscal balance to GDP (lagged 12 months)
RGDP6 ⫽ growth rate of real GDP (lagged six months)
beyond the scope of this report. But we can do it with the signals
approach, given that the signals approach helps predict a crisis within the
next 12 months (from ex post analysis). We can use the available data in
December 2001 to say something about the prospect of a crisis up to
December 2002.
Table 2.13 presents the composite index (I(2)) of 11 variables found to
be leading indicators of a currency crisis for Thailand in this study. The
Table 2.11 Probit estimate for currency crisis: Model 2 (1994.1–2000.12)
Variable Coefficient Std. Error b/Std.Err P[円Z円⬎z] x-bar

(partial
derivatives)
M2RESG6 0.0059 0.0031 1.911 0.0560 0.7728
DRER603 ⫺0.0097 0.0035 ⫺2.788 0.0053 8.8105
CAGDP6 ⫺0.0121 0.0069 ⫺1.752 0.0798 0.3084
SETG3 ⫺0.0030 0.0015 ⫺1.950 0.0511 ⫺4.3632
X$G6 ⫺0.0136 0.0039 ⫺3.438 0.0006 10.0784
FBGDP12 ⫺0.0234 0.0104 ⫺2.250 0.0244 0.2302
Log-likelihood ⫺19.28912
Chi-squared 37.11604
Significance level 0.0000
Frequencies of actual and predicted outcomes (cut-off probability ⫽0.5)
Predicted Actual Total
1 (Crisis) 0 (No crisis)
1 13 7 20
0 1 63 64
Total 14 70 84
Proportion of correct prediction ⫽0.905
Adjusted noise-to-signal ratio⫽0.108
Definition of variables
M2RESG6 ⫽ growth in the ratio of M2 over reserves
(lagged six months)
DRER603 ⫽ deviation of real exchange rate from previous
60-month average (lagged three months)
CAGDP6 ⫽ ratio of current account to GDP (lagged six months)
SETG3 ⫽ growth rate of stock price index (lagged three months)
X$G6 ⫽ export growth (US$) (lagged six months)
FBGDP12 ⫽ ratio of fiscal balance to GDP (lagged 12 months)
composite index for 2001 lies between 1.82 and 7.64 with probability of a
crisis between 0.08 and 0.46. Using 0.5 as a cut-off point, the composite
index implies that a currency crisis is not likely in the next 12 months.
Nevertheless, we should be watchful of these indicators because the terms
of trade have had a negative growth in the past 12 months and export
growth in dollar value has also continuously registered negative growth
below its threshold value of –3.8 percent since July 2001. The fiscal balance
Probability of crisis (Model 1)

1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
Probability of crisis (Model 2)

1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
Figure 2.5 Plot of probability of currency crisis from probit estimates
has also been mostly in deficit due to the stimulative fiscal policy. As a con-
sequence, the ratio of fiscal balance to GDP was running below its thresh-
old value in the last four months of 2001. There also remain some
structural problems in the financial and real sectors, and the high and
rapidly rising public debt was a significant cause for concern.
5 SUMMARY AND CONCLUSION
Based on Thailand’s available monthly data during 1992–2000, it is found

in this study that there are several leading indicators of currency crisis
that we should keep track of. The signals analysis based on the 12-month
Table 2.12 Probability of a currency crisis from probit estimates (out-of-

sample forecast)
Year 2001 Model 1 Model 2

January 0.0001 0.0001
February 0.0001 0.0001
March 0.0202 0.0030
April 0.0018 0.0325
May 0.0003 0.0000
June 0.0018 0.0109
July 0.0002 0.0914
August 0.0003 0.0134
September 0.0244 0.0482
October 0.0137 0.2110
November 0.0084 0.0351
December 0.0104 0.1149
signaling horizon yielded 11 indicators that seem to have good perform-

ance. A subset of these indicators also performed well with probit esti-
mates, which yielded high probability of crisis during the crisis period and
very low probability otherwise. These variables can be grouped by sector as
follows in Table 2.14.
In fact, these variables indicate the general health of the macroeconomy
and they should be kept track of by the concerned authorities. One can
notice that the above list is rather short in use of financial variables. Since
there seems to be a vicious cycle in which currency crisis and banking crisis
feed back on each other, an in-depth analysis of Thailand’s banking trou-
bles and their causes in the past is recommended.
There are some weaknesses in this study that should be pointed out to
the readers. The first one is that the sample period included in this study
actually covered only one episode of currency crisis, running continuously
from January 1997 to early 1998. It would have been more desirable to cover
another episode of currency crisis in 1984 when the baht had to be deval-
ued by about 14.7 percent. However, several important indicators are not
available on a monthly or even a quarterly basis before the 1990s. Therefore,
the 1984 currency crisis episode can not be used with confidence in this
study due to lack of data. Second, even when the sample period is restricted
to the 1990s, there are still some important variables that are available only
at the quarterly level. These are GDP and short-term external debt, and
their quarterly data were used to estimate the monthly data as already
explained in the text. Given the above-mentioned weaknesses, the results
Table 2.13 Composite index and conditional probability of a currency

crisis within 12 months
Total Short-term SET Index Terms of Trade M2/Reserves

Debt/Reserves
(Ratio) Signal (Growth) Signal (Growth) Signal (Growth) Signal

Threshold 128.4 ⫺42.5 ⫺8.6 11.4
Adj n-t-s 0.13 0.15 0.25 0.28
Jan-01 44.47 0 ⫺30.32 0 ⫺12.76 1 ⫺9.31 0
Feb-01 43.67 0 ⫺13.12 0 ⫺14.60 1 ⫺10.38 0
Mar-01 44.50 0 ⫺27.07 0 ⫺13.80 1 ⫺8.53 0
Apr-01 45.30 0 ⫺22.99 0 ⫺12.53 1 ⫺10.70 0
May-01 45.97 0 ⫺4.07 0 ⫺11.81 1 ⫺8.68 0
Jun-01 47.10 0 ⫺0.96 0 ⫺10.64 1 ⫺6.35 0
Jul-01 45.82 0 4.57 0 ⫺7.97 0 ⫺6.21 0
Aug-01 43.99 0 9.01 0 ⫺6.45 0 ⫺5.05 0
Sep-01 43.08 0 ⫺0.09 0 ⫺4.75 0 ⫺1.74 0
Oct-01 41.55 0 1.20 0 ⫺6.01 0 ⫺2.20 0
Nov-01 40.37 0 8.89 0 ⫺6.95 0 0.00 0
Dec-01 39.73 0 12.88 0 ⫺6.09 0 1.06 0
Domestic Credit/ Deviation of Current Inflation

GDP RER From Account/GDP
60 Months
(Growth) Signal (%) Signal (Ratio) Signal (Own level) Signal
Threshold 15.0 ⫺5.3 ⫺8.3 6.5

Adj n-t-s 0.42 0.85 0.42 0.42
Jan-01 ⫺22.32 0 23.67 0 2.79 0 1.31 0
Feb-01 ⫺17.29 0 21.20 0 8.68 0 1.46 0
Mar-01 ⫺17.54 0 24.15 0 2.77 0 1.39 0
Apr-01 ⫺19.70 0 27.00 0 1.61 0 2.55 0
May-01 ⫺21.44 0 26.03 0 3.95 0 2.78 0
Jun-01 ⫺20.49 0 24.93 0 5.59 0 2.31 0
Jul-01 ⫺25.67 0 24.50 0 3.50 0 2.15 0
Aug-01 ⫺23.48 0 21.72 0 10.53 0 1.38 0
Sep-01 ⫺0.30 0 19.32 0 3.65 0 1.37 0
Oct-01 ⫺14.86 0 19.53 0 5.13 0 1.38 0
Nov-01 ⫺10.70 0 17.83 0 5.59 0 0.99 0
Dec-01 ⫺10.06 0 15.47 0 11.05 0 0.77 0
Exports (US$) Real GDP Fiscal Balance/ Composite Conditional

GDP Index of Probability
Currency of Currency
Crisis Crisis
(Growth) Signal (Growth) Signal (Ratio) Signal
Threshold ⫺3.8 ⫺1.0 ⫺3.4
Adj n-t-s 0.55 0.57 0.55
Jan-01 ⫺3.91 1 5.35 0 ⫺3.18 0 5.82 0.08
Feb-01 ⫺3.70 0 ⫺0.86 0 ⫺2.65 0 4.00 0.08
Mar-01 3.59 0 0.66 0 ⫺2.56 0 4.00 0.08
Apr-01 ⫺8.02 1 1.61 0 ⫺5.00 1 7.64 0.46
May-01 6.81 0 2.92 0 5.19 0 4.00 0.08
Jun-01 ⫺0.96 0 0.94 0 2.67 0 4.00 0.08
Jul-01 ⫺14.18 1 7.60 0 ⫺1.75 0 1.82 0.08
Aug-01 ⫺7.59 1 5.70 0 1.13 0 1.82 0.08
Sep-01 ⫺11.38 1 ⫺8.18 1 ⫺5.94 1 5.39 0.08
Oct-01 ⫺14.16 1 6.09 0 ⫺10.09 1 3.64 0.08
Nov-01 ⫺12.30 1 0.86 0 ⫺5.00 1 3.64 0.08
Dec-01 ⫺13.70 1 ⫺0.70 0 ⫺3.72 1 3.64 0.08
Table 2.14 Leading indicators
Signals Approach Probit Estimates

External sector variables
Terms of trade growth 冪
Export growth 冪冪
Ratio of current account to GDP 冪冪
Real exchange rate misalignment 冪冪
Ratio of short-term external 冪
Financial variables
Growth of M2/international reserves 冪冪
Growth of domestic credit/GDP 冪
Real sector variables
Ratio of fiscal balance to GDP 冪冪
Inflation rate 冪
Real GDP growth rate 冪冪
Change in stock prices 冪冪
from both the signals analysis and probit estimates conducted in this study
appear quite satisfactory on both theoretical and statistical grounds.
The 1997 crisis in Thailand and its phenomenal spread has spawned a
large body of literature on its analysis and its lessons. This chapter is just a
humble attempt to use the available data in Thailand to study their leading
or signaling nature. The study finds many statistically significant indicators.
It also finds that many important indicators are lacking due to inadequate
collection and dissemination. For example, the data on non-performing
loans are not available before June 1998 and the series are simply not long
enough to establish a reliable statistical pattern. At present, the data col-
lection and dissemination by public authorities have improved a great deal
since the 1997 crisis. Although Thailand joined the IMF Special Data
Dissemination Standard (SDDS) since 1996, it was not until May 2000 that
it could meet the SDDS specifications for coverage, periodicity, timeliness
and advance release calendars. Also the present effort by the National
Economic and Social Development Board to track the short-term move-
ment of the economy by the Current Quarter Model (CQM) should
provide a better opportunity to observe some of the early warning signals.
In hindsight, the fact that there were signals for and high probability of
a currency crisis for the 1997 event indicated that there were some funda-
mental problems with the Thai economy that led speculators to attack the
baht. The fact that the concerned authorities were not mindful of the early
warning signs and decided to defend the baht almost at all costs makes
ordinary citizens shudder. Now with faster reporting of data, greater trans-
parency and improved information and indicators to monitor the economic
conditions, we can hope that monetary authorities will quickly realize an
imminent crisis and put out remedial measures before things get out of
hand.
NOTES
* This chapter is a result of the East Asian Development Network (EADN) research project
in 2002 to evaluate vulnerabilities to economic crises of East Asian countries and to learn
from the 1997–98 economic crisis that severely affected most economies in the region.
Since the project was finished in 2002, the data in this chapter are only up to December
2001. The author thanks Professor Lawrence R. Klein and Professor Tayyeb Shabbir,
both of the University of Pennsylvania, and Dr Chalongphob Sussangkarn of Thailand
Development Research Institute for their valuable comments and Miss Sanpichit
Songpaisan for her research assistance.
1. Ministry of Finance (1998). In the document, it was pointed out that Thailand’s net inter-
national reserves decreased from 33.8 billion dollars in December 1996 to around 1 billion
dollars in July 1997. Such a huge loss was concealed by the swap operation and did not
show up in the gross reserves as can be seen in the plots of Thailand’s gross and net inter-
national reserves in Figure 2.1.
2. The growth of exports (in millions of US$) during 1994–95 was 21.7 percent and 24.9
percent, respectively. It went down to –1.4 percent in 1996 before picking up to 4.3 percent
in 1997.
3. For instance in Argentina, Chile, Malaysia and Mexico, etc. See Kaminsky and Reinhart
(1999).
4. The year 1997 was chosen just for being a convenient cut-off point. Otherwise, there is a
significant amount of related literature to be reviewed. Interested readers can see a com-
prehensive review of earlier literature in Kaminsky et al. (1998).
5. However, for GDP, there are only quarterly data starting from the first quarter of 1993
and the monthly GDP data have to be estimated from a quarterly relationship of GDP
and other variables.
6. The 10–25th percentile range may appear arbitrary but it may be looked upon as the prob-
ability of rejecting the null hypothesis when it is true, or Type I error. We may regard the
normal or tranquil period as our null hypothesis and take the export growth as our indi-
cator, for example. If we find that 10 percent of the observations post an export growth
below 1 percent, we then regard any export growth below 1 percent as a signal. Such
reading of the indicator may be wrong if the null hypothesis (tranquility) is true.
Therefore, setting the maximum 25th percentile as the upper limit implies that we are
willing to accept one-fourth probability of calling a crisis when it is not true. It is found
in this study that for some indicators the percentile at which the adjusted noise-to-signal
ratio is minimum is far below the 25th percentile. See more details on methodology in
Kaminsky and Reinhart (1999).
7. It is notable that this number is quite different from the threshold of 5.6 found in
Poonpatpibul and Ittisupornrat (2001), who used the same signals approach. Such a
difference cannot be attributed to their using a 24-month horizon because in this study
the threshold from the 24-month horizon for export growth is –6.9.
8. The other two composite indicators considered by Kaminsky (1999) are the following
cases: a) Extreme signals are given more weight than mild signals. b) A time horizon of
eight months is taken into account when adding up signals because not all indicators issue
signals jointly in the same month. See more details in Kaminsky (1999).
9. The criterion for dummy variables in probit estimates is that there must be observations
for which the left-hand-side variable takes both values 0 and 1 in both groups of the
observations for the right-hand-side dummy variable. See Greene (1995, p. 416).
REFERENCES
Calvo, G. and C. Reinhart (1996), ‘Capital Flows to Latin America: Is there

Evidence of Contagion Effects?’, in G.A. Calvo, M. Goldstein and Eduard
Hochreiter (eds), Private Capital Flows to Emerging Markets, Washington, DC:
Institute for International Economics, pp. 151–71.
Dodd, R. (2001), ‘The Role of Derivatives in the East Asian Financial Crisis’,
Special Policy Report 1, Derivatives Study Center.
Eichengreen, B., A. Rose and C. Wyplosz (1996), ‘Contagious Currency Crisis’,
Centre for Economic Policy Research Discussion Paper No. 1453, August.
Engwatana, Pipat (1999), ‘Early Warning Indicators of Currency Crisis: A Case
Study of Thailand’, MA: Thesis, Faculty of Economics, Thammasat University.
Esquivel, Gerardo and Felipe Larrain (1998), ‘Explaining Currency Crises’,
Development Discussion Paper, No. 666, Massachusetts: Harvard Institute for
International Development, November.
Gerlach, S. and F. Smets (1995), ‘Contagious Speculative Attacks’, European
Journal of Political Economy, 11, 45–63.
Goldfajn, Ilan and Rodrigo O. Valdés (1997), ‘Are Currency Crises Predictable?’,
IMF Working Paper WP/97/159, December.
Greene, W.H. (1995), LIMDEP Version 7.0 User’s Manual, Econometric Software,
Inc. New York.
Jeanneau, S. and M. Micu (2002), ‘International Bank Lending to Emerging
Market Countries: Explaining the 1990s Roller Coaster’, BIS Quarterly Review,
March, 52–64.
Kaminsky, G.L. (1999), ‘Currency and Banking Crises: The Early Warnings of
Distress’, IMF Working Paper WP/99/178, December.
Kaminsky, G.L. and C.M. Reinhart (1999), ‘The Twin Crises: The Causes of
Banking and Balance-of-payments Problems’, American Economic Review,
89 (3), June.
Kaminsky, G.L. and C.M. Reinhart (2000), ‘On Crises, Contagion, and Confusion’,
Journal of International Economics, 51, 145–68.
Kaminsky, G.L., S. Lizondo and C.M. Reinhart (1998), ‘Leading Indicators of
Currency Crisis’, IMF Staff Papers, 45 (1), March, 1–48.
Kruger, M., P.N. Osakwe and J. Page (1998), ‘Fundamentals, Contagion and
Currency Crises: An Empirical Analysis’, Bank of Canada Working Paper
WP/97/159, December.
Krugman, P. (1979), ‘A Model of Balance of Payments Crisis’, Journal of Money,
Credit and Banking, 11, August, 311–25.
Krugman, P. (2001), ‘Crises: The Next Generation?’, Paper prepared for the Razin
Conference, Tel Aviv University, March.
Mariano, R.S., A.G. Abiad, B. Gultekin, T. Shabbir and A.H.H. Tan (2002),
‘Markov Chains in Predictive Models of Currency Crises – With Applications to
Southeast Asia’, PIER Working Paper No. 02-013, University of Pennsylvania.
Ministry of Finance (1998), ‘Directions of Monetary and Fiscal Measures to Solve
Economic Problems’, a mimeograph, July (in Thai).
Obstfeld, M. (1996), ‘Model of Currency Crises with Self-fulfilling Features’,
European Economic Review, 40 (3–5), April, 1037–47.
Poonpatpibul, Chaipat and Anotai Ittisupornrat (2001), ‘Early Warning System’,
BOT Working Paper WP/05/2001, June (in Thai).
Tinakorn, Pranee (1998), ‘Analysis of Leading Economic Indicators for Thailand’,
a TDRI research report in the project Short-Term Economic Model and
Forecast, financed by the National Economic and Social Development Board,
September (in Thai).
APPENDIX 2A
Table 2A.1 Index of currency market turbulence
Exchange Rate International Index of Currency Market Turbulence

Reserves
% % I I⫹ I ⬎I I⫹ I ⬎I
change change 2sd ⫹2sd 1.5sd ⫹1.5sd
Jan-93 25.53 0.24 21 937.00 3.57 ⫺3.60 12.60 0 9.47 0

Feb-93 25.49 ⫺0.16 21 634.90 ⫺1.38 1.33 12.60 0 9.47 0
Mar-93 25.42 ⫺0.27 22 239.40 2.79 ⫺3.28 12.60 0 9.47 0
Apr-93 25.23 ⫺0.75 22 611.60 1.67 ⫺2.55 12.60 0 9.47 0
May-93 25.22 ⫺0.04 23 114.70 2.22 ⫺2.43 12.60 0 9.47 0
Jun-93 25.21 ⫺0.04 23 979.80 3.74 ⫺4.07 12.60 0 9.47 0
Jul-93 25.31 0.40 23 919.70 ⫺0.25 0.67 12.60 0 9.47 0
Aug-93 25.18 ⫺0.51 24 222.80 1.27 ⫺1.88 12.60 0 9.47 0
Sep-93 25.19 0.04 25 225.30 4.14 ⫺4.41 12.60 0 9.47 0
Oct-93 25.26 0.28 25 544.40 1.26 ⫺1.08 12.60 0 9.47 0
Nov-93 25.36 0.40 25 206.10 ⫺1.32 1.82 12.60 0 9.47 0
Dec-93 25.45 0.35 25 438.80 0.92 ⫺0.64 12.60 0 9.47 0
Jan-94 25.53 0.31 25 359.30 ⫺0.31 0.65 12.60 0 9.47 0
Feb-94 25.38 ⫺0.59 26 251.30 3.52 ⫺4.37 12.60 0 9.47 0
Mar-94 25.29 ⫺0.35 26 672.60 1.60 ⫺2.08 12.60 0 9.47 0
Apr-94 25.25 ⫺0.16 26 592.80 ⫺0.30 0.16 12.60 0 9.47 0
May-94 25.21 ⫺0.16 27 512.80 3.46 ⫺3.88 12.60 0 9.47 0
Jun-94 25.14 ⫺0.28 28 340.50 3.01 ⫺3.51 12.60 0 9.47 0
Jul-94 24.97 ⫺0.68 28 588.30 0.87 ⫺1.62 12.60 0 9.47 0
Aug-94 25.02 0.20 29 064.00 1.66 ⫺1.59 12.60 0 9.47 0
Sep-94 24.98 ⫺0.16 29 950.20 3.05 ⫺3.44 12.60 0 9.47 0
Oct-94 24.96 ⫺0.08 29 851.70 ⫺0.33 0.27 12.60 0 9.47 0
Nov-94 24.98 0.08 29 743.20 ⫺0.36 0.47 12.60 0 9.47 0
Dec-94 25.10 0.48 30 279.00 1.80 ⫺1.46 12.60 0 9.47 0
Jan-95 25.07 ⫺0.12 29 906.10 ⫺1.23 1.21 12.60 0 9.47 0
Feb-95 25.02 ⫺0.20 30 135.60 0.77 ⫺1.03 12.60 0 9.47 0
Mar-95 24.76 ⫺1.04 30 119.50 ⫺0.05 ⫺0.98 12.60 0 9.47 0
Apr-95 24.56 ⫺0.81 31 727.10 5.34 ⫺6.55 12.60 0 9.47 0
May-95 24.66 0.41 33 272.40 4.87 ⫺4.83 12.60 0 9.47 0
Jun-95 24.67 0.04 34 958.30 5.07 ⫺5.41 12.60 0 9.47 0
Jul-95 24.74 0.28 34 415.70 ⫺1.55 1.95 12.60 0 9.47 0
Aug-95 24.95 0.85 34 629.10 0.62 0.18 12.60 0 9.47 0
Sep-95 25.12 0.68 35 866.10 3.57 ⫺3.16 12.60 0 9.47 0
Oct-95 25.11 ⫺0.04 35 731.40 ⫺0.38 0.36 12.60 0 9.47 0
Nov-95 25.16 0.20 36 204.40 1.32 ⫺1.23 12.60 0 9.47 0
Dec-95 25.16 0.00 37 026.70 2.27 ⫺2.44 12.60 0 9.47 0
Jan-96 25.29 0.52 37 721.20 1.88 ⫺1.50 12.60 0 9.47 0
Feb-96 25.24 ⫺0.20 38 694.20 2.58 ⫺2.97 12.60 0 9.47 0
Mar-96 25.23 ⫺0.04 38 982.50 0.75 ⫺0.84 12.60 0 9.47 0
Apr-96 25.27 0.16 38 862.30 ⫺0.31 0.49 12.60 0 9.47 0
Table 2A.1 (continued)

Reserves
% % I I⫹ I ⬎I I⫹ I ⬎I
May-96 25.29 0.08 39 053.80 0.49 ⫺0.45 12.60 0 9.47 0

Jun-96 25.35 0.24 39 830.00 1.99 ⫺1.90 12.60 0 9.47 0
Jul-96 25.34 ⫺0.04 39 360.60 ⫺1.18 1.23 12.60 0 9.47 0
Aug-96 25.27 ⫺0.28 39 370.30 0.02 ⫺0.30 12.60 0 9.47 0
Sep-96 25.36 0.36 39 537.00 0.42 ⫺0.10 12.60 0 9.47 0
Oct-96 25.46 0.39 39 902.50 0.92 ⫺0.60 12.60 0 9.47 0
Nov-96 25.45 ⫺0.04 39 613.30 ⫺0.72 0.74 12.60 0 9.47 0
Dec-96 25.56 0.43 38 724.50 ⫺2.24 2.85 12.60 0 9.47 0
Jan-97 25.69 0.51 39 233.80 1.32 ⫺0.91 12.60 0 9.47 0
Feb-97 25.90 0.82 38 149.10 ⫺2.76 3.79 12.60 0 9.47 0
Mar-97 25.92 0.08 38 065.60 ⫺0.22 0.31 12.60 0 9.47 0
Apr-97 26.03 0.42 37 320.10 ⫺1.96 2.53 12.60 0 9.47 0
May-97 25.84 ⫺0.73 33 307.60 ⫺10.75 10.84 12.60 0 9.47 1
Jun-97 25.75 ⫺0.35 32 353.00 ⫺2.87 2.74 12.60 0 9.47 0
Jul-97 30.16 17.13 30 424.20 ⫺5.96 23.54 12.60 1 9.47 1
Aug-97 32.41 7.46 25 938.60 ⫺14.74 23.32 12.60 1 9.47 1
Sep-97 36.27 11.91 29 612.20 14.16 ⫺3.33 12.60 0 9.47 0
Oct-97 37.55 3.53 31 287.20 5.66 ⫺2.56 12.60 0 9.47 0
Nov-97 39.30 4.66 26 253.60 ⫺16.09 21.97 12.60 1 9.47 1
Dec-97 45.29 15.24 26 967.70 2.72 12.31 12.60 0 9.47 1
Jan-98 53.71 18.59 26 724.30 ⫺0.90 19.56 12.60 1 9.47 1
Feb-98 46.30 ⫺13.80 26 156.10 ⫺2.13 ⫺11.51 12.60 0 9.47 0
Mar-98 41.33 ⫺10.73 27 680.00 5.83 ⫺17.00 12.60 0 9.47 0
Apr-98 39.48 ⫺4.48 29 530.50 6.69 ⫺11.67 12.60 0 9.47 0
May-98 39.14 ⫺0.87 27 450.50 ⫺7.04 6.71 12.60 0 9.47 0
Jun-98 42.36 8.25 26 571.70 ⫺3.20 11.69 12.60 0 9.47 1
Jul-98 41.19 ⫺2.77 26 776.30 0.77 ⫺3.60 12.60 0 9.47 0
Aug-98 41.58 0.94 26 678.80 ⫺0.36 1.33 12.60 0 9.47 0
Sep-98 40.41 ⫺2.80 27 290.80 2.29 ⫺5.27 12.60 0 9.47 0
Oct-98 38.14 ⫺5.62 28 482.10 4.37 ⫺10.31 12.60 0 9.47 0
Nov-98 36.46 ⫺4.40 28 891.40 1.44 ⫺5.95 12.60 0 9.47 0
Dec-98 36.25 ⫺0.58 29 535.90 2.23 ⫺2.98 12.60 0 9.47 0
Jan-99 36.59 0.94 29 013.10 ⫺1.77 2.84 12.60 0 9.47 0
Feb-99 37.06 1.28 28 721.40 ⫺1.01 2.37 12.60 0 9.47 0
Mar-99 37.51 1.21 29 936.10 4.23 ⫺3.34 12.60 0 9.47 0
Apr-99 37.60 0.24 30 203.80 0.89 ⫺0.72 12.60 0 9.47 0
May-99 37.02 ⫺1.54 30 637.20 1.43 ⫺3.09 12.60 0 9.47 0
Jun-99 36.91 ⫺0.30 31 433.90 2.60 ⫺3.10 12.60 0 9.47 0
Jul-99 37.11 0.54 31 928.80 1.57 ⫺1.15 12.60 0 9.47 0
Aug-99 37.98 2.34 32 216.10 0.90 1.38 12.60 0 9.47 0
Sep-99 39.88 5.00 32 360.20 0.45 4.52 12.60 0 9.47 0
Oct-99 39.47 ⫺1.03 32 438.10 0.24 ⫺1.29 12.60 0 9.47 0
Table 2A.1 (continued)

Reserves
% % I I⫹ I ⬎I I⫹ I ⬎I
Nov-99 38.77 ⫺1.77 32 842.10 1.25 ⫺3.11 12.60 0 9.47 0

Dec-99 38.18 ⫺1.53 34 780.60 5.90 ⫺7.88 12.60 0 9.47 0
Jan-00 37.35 ⫺2.16 32 630.20 ⫺6.18 4.49 12.60 0 9.47 0
Feb-00 37.71 0.96 31 953.80 ⫺2.07 3.19 12.60 0 9.47 0
Mar-00 37.90 0.50 32 283.90 1.03 ⫺0.61 12.60 0 9.47 0
Apr-00 37.97 0.18 32 166.00 ⫺0.37 0.58 12.60 0 9.47 0
May-00 38.95 2.58 31 904.20 ⫺0.81 3.46 12.60 0 9.47 0
Jun-00 39.06 0.28 32 142.00 0.75 ⫺0.52 12.60 0 9.47 0
Jul-00 40.22 2.97 31 929.60 ⫺0.66 3.68 12.60 0 9.47 0
Aug-00 40.87 1.62 32 232.40 0.95 0.60 12.60 0 9.47 0
Sep-00 41.88 2.47 32 249.80 0.05 2.41 12.60 0 9.47 0
Oct-00 43.21 3.18 32 244.70 ⫺0.02 3.19 12.60 0 9.47 0
Nov-00 43.73 1.20 32 316.10 0.22 0.97 12.60 0 9.47 0
Dec-00 43.09 ⫺1.46 32 661.30 1.07 ⫺2.61 12.60 0 9.47 0
Mean 31.43 0.63 31 050.33 0.53 0.06
SD 7.46 4.13 4 835.86 3.84 6.27
PART II
Cures and Reforms

3. The next financial crisis1
Barry Eichengreen
There has been considerable water under the bridge since early 1998 when
then US Treasury Secretary Rubin faced the problem of financial instabil-
ity made so evident by the Asian Crisis. In reality, these efforts long pre-
date Rubin’s speech. Some would trace them to the Mexican Crisis of
1994–95, which then IMF Managing Director Michel Camdessus referred
to as ‘the first financial crisis of the 21st century’ on the grounds that
financial structure and markets had played such a prominent role in its
development.2 Other observers would place the inauguration of this effort
even earlier, perhaps at the time of the 1982 Mexican Crisis, out of which
developed the Brady Plan, or the 1974 Herstatt Crisis that provided the
impetus for the Basel Capital Accord, or even the global financial crisis of
the 1930s that prompted the creation of the Bretton Woods Institutions.
So what makes this an appropriate time for stock taking? As I write, in
the spring of 2005, all the ingredients are in place for a classic emerging
market crisis.3 We have come to the end of a long period when the Federal
Reserve has kept interest rates low.4 Partly in response to the FED’s earlier
policies, spreads on emerging market debt have fallen to very low levels.
Now US policy rates are rising, and rising US rates are a traditional trigger
for financial problems in emerging markets. Market participants see this
coming: we have already seen a significant widening of spreads in the first
half of 2005. And there are growing worries about a disorderly correction
of the US current account deficit. On top of this, oil prices are rising, cre-
ating additional difficulties for emerging markets that are not energy
exporters. The OECD estimates that China and India are two to three times
more intensive in the use of energy than the developed countries and that
an increase of $10 in the cost of a barrel of oil shaves 0.8 percent off China’s
rate of GDP growth.5
Global economic growth of course is entirely dependent on two engines:
the United States and China. This makes the specter of a hard landing in
China as policy-makers attempt to cool down an overheated economy, and
of a dollar crash in the United States if the problem of twin deficits finally
comes home to roost, particularly alarming for emerging markets. The US
is the single most important market for exports of consumer goods by
121
122 Cures and reforms
emerging markets, while China is far and away the most rapidly growing
market for both consumer and producer goods. If either engine begins to
sputter, the result could be very serious problems for emerging markets that
are heavily dependent on exports for both growth and debt sustainability.
And if growth slows simultaneously in both countries, the consequences
could be serious indeed.
Thus, it is striking that this potentially fatal cocktail of rising US inter-
est rates, higher oil prices and growing uncertainty about Chinese policy
has not yet produced serious difficulties. One response to this observation
is ‘just be patient’. Give it more time, and the financial ambulance chasers
among us will eventually acquire a new set of clients. But another response
is that considerable progress has in fact been made in strengthening the
international financial architecture. Reforms of international financial
institutions and markets have strengthened the global financial system,
while significant reforms in emerging market economies themselves have
made these countries more resilient to mounting strains.
There is at least a kernel of truth in this point of view. Specifically,
I would point to ten important reforms that have helped to make the world
a safer financial place. However, these reforms have come at a cost. Each of
them has had unintended consequences that are not entirely salutary from
the point of view of either emerging markets or the international system.
First, emerging markets have grown more cautious since 1994, and espe-
cially since 1997, about cross-border bank borrowing as a way of accessing
external finance. Less reliance on short-term finance, and less reliance on
cross-border bank finance in particular, has reduced an important source
of financial vulnerability. Foreign banks, for their part, have grown more
cautious about cross-border lending, having been reminded of the risks of
international financial intermediation by earlier crises.6 New net flows
through money markets and new bond flotations have also fallen, albeit by
less than cross-border bank lending and less in Asia than Latin America.7
Be that as it may, the point is that emerging markets have taken valuable
steps to lengthen the maturity structure of their debts.8
Second, emerging markets have become more prudent about running
current account deficits. The Institute of International Finance, at latest
report, forecasts that emerging Asia will run a current account surplus of
$113 billion in 2005, down only slightly from its surplus of $120 billion in
2004.9 Latin America as a whole is running a balanced current account. Of
the principal Latin countries, only Ecuador has a noticeable deficit. But
elsewhere the notorious problem of current account reversals – what
happens when funding for the flow deficit suddenly dries up – has been very
considerably ameliorated. The one place where this is not the case is emerg-
ing Europe – that is, in countries like Hungary, Romania and Turkey.10
The next financial crisis 123
The problem with these solutions to the problem of financial instability,

if we may call them that, is that they imply foregoing many of the benefits
of capital mobility. In Asia we have the anomalous situation of low-income
developing countries and middle-income emerging markets financing the
deficits of a high-income country, the United States, and subsidizing the
consumption of its residents. This is not how the international financial
system is supposed to work. The growth of foreign direct investment is one
thing, but portfolio capital continues to flow upstream from the developing
to the developed world. Thus, this ‘solution’ to the problem of financial
volatility comes at significant cost.
A third stabilizing development is the massive accumulation of foreign
reserves by emerging markets.11 Reserves provide insurance against
financial shocks. Even if external debt cannot be rolled over, reserves are
still available to service it. But they provide insurance at a cost, in the short
run because the yield on US treasury bonds is less than the opportunity cost
of funds, and in the long run because the twin deficits problem in the
United States implies capital losses on dollar-denominated assets. Reserve
accumulation is therefore a sign that net capital transfer from rich to poor
countries is even less than meets the eye. And reserve accumulation in
dollars exposes emerging markets to very considerable capital losses if the
US currency falls.
Fourth, there is the fact of greater exchange rate flexibility. I regard this
as one of the most significant reforms of the international financial archi-
tecture. Greater flexibility has reduced the temptation for governments to
stick with unsustainable policies. It gives them additional capacity to use
macroeconomic policy in stabilizing ways. It creates two-way bets for spec-
ulators, removing a focal point for destabilizing financial dynamics.
Except for Ecuador and Venezuela, all of the principal Latin American
countries have now adopted flexible exchange rates. But while the Mexican
peso now floats freely and the Brazilian real has been allowed to fall by
about 9 percent since financial market participants began contemplating
rising US interest rates, other emerging markets remain reluctant to let their
currencies move. Asian countries, China in particular, fear the implications
for export growth. Latin American countries fear the financial conse-
quences, given the extent of exchange-rate-linked debt.12
Fifth, stronger fiscal policies in emerging markets have reduced the risk
of old-fashioned ‘first-generation’ crises.13 In Latin America, improvements
in fiscal discipline were long overdue, and in the short term their implica-
tions for financial stability were strongly positive. The problem is the inabil-
ity of governments to pursue their social agendas, which raises questions
about the durability of political support and therefore about the sustain-
ability of the policies. Running primary surpluses to pay off the debt also
leaves a smaller share of tax revenues for investment in infrastructure and

education, with negative implications for growth. In Turkey, impressive
fiscal consolidation has come at the expense of education spending, which
raises questions about future growth and therefore about debt sustainabil-
ity. The commitment to primary surpluses also means that fiscal policy is
dangerously procyclical, since the worse the country’s economic perfor-
mance, the larger its surplus must become.
Sixth, leverage in the international financial system has been reduced,
moderating one important threat to systemic stability. Following the Long-
term Capital Management Crisis in 1998, banks curtailed the amount of
credit that they made available to hedge funds.14 The ‘macro funds’ that
were the most voracious consumers of leverage have since gone out of busi-
ness or changed their investment styles. This raises hopes that if an unex-
pectedly rapid rise in US policy rates unleashes a flight to quality and
causes sudden deleveraging, the impact on the international system will be
lessened. The only problem is that recent data suggest that the use of lever-
age has been growing again.15
Seventh, multilateral surveillance of financial systems has been strength-
ened. I can remember visiting the IMF’s Capital Markets Division in 1995
when it occupied a small bank of offices in one of the Fund’s satellite build-
ings. Now the Capital Markets Division occupies the better part of a floor
in the Fund’s headquarters. The IMF issues a Global Financial Stability
Report twice annually. There are regular reviews of prudential supervision
and regulation in the context of the Financial Sector Assessment Program.
But there remains the problem that IMF expertise is still heavily dominated
by the macroeconomic side. The IMF continues to hire new PhD’s in eco-
nomics, as opposed to people with experience in bank supervision and
securities market regulation. And the Capital Markets Division spends too
much time writing the equivalent of low-level investment bank newsletters
for internal consumption while doing too little analytical work.
Eighth, there is the effort to promulgate standards and codes designed to
enhance transparency and strengthen market discipline, backed up by the
Bank and Fund’s Reviews of the Observance of Standards and Codes
(ROSCs). While it is still early, there are some indications that the markets
take these efforts seriously and reward countries that subscribe to the
SDDS (Special Data Dissemination Standard) or adhere to the interna-
tional financial community’s other standards and codes with lower bor-
rowing costs and improved credit ratings.16 The problem is that the benefits
may turn out to be less than meets the eye if public information crowds out
private information. At least one study suggests that when countries sub-
scribe to the SDDS or adopt internationally recognized accounting stan-
dards, there is a decline in the number of private analysts following the
country, since investing in private information generation and acquisition

become less profitable. The increase in the accuracy of information pro-
duced by the improvement in public information is partly offset by the
reduction in the accuracy of privately generated information associated
with the fall in analyst following.17
Ninth, there is the addition of collective action clauses (CACs) to debt
securities governed by New York law. Since 2003 this contractual innova-
tion, which was long dismissed as infeasible, has become the norm. (In the
latter part of 2003 and early part of 2004, sovereign bonds containing
CACs constituted more than 75 percent of the total value of total new
issues.)18 There is reason to hope that the addition of majority consent pro-
visions to sovereign debt instruments will ease holdout problems and facil-
itate orderly restructuring. The prospect of less painful restructuring gives
grounds for expecting that the IMF will be able to stand aside and let events
run their course when countries have unsustainable debts. There will be
fewer cases like Argentina in August 2001, in other words, when the Fund
feels compelled to run to the rescue of a bankrupt country in order to avert
fears that doing otherwise will destabilize the global financial system. This
will reduce the moral hazard associated by the expectation of bailouts, in
turn strengthening market discipline.
But it will take until the end of the present decade, or longer, until CACs
dominate the stock of emerging market debt securities. In addition, issue-
specific majority voting rules do little to facilitate coordination among the
holders of different bond issues, a problem whose importance is apparent
in Argentina’s default. So long as coordination remains difficult, restruc-
turing will remain difficult. It is not yet certain, in other words, that collec-
tive action clauses will make a difference.
Tenth and finally, the IMF has become more transparent. Stanley Fischer
has referred with some justification to the revolution in IMF transparency
that occurred on his watch.19 The IMF now posts so much information on
its website that it is difficult for its critics to keep up. Some would say that
this is precisely the point. Seriously, making policy in the light of day forces
the Fund to justify and defend its decisions. It enhances the institution’s
ability to make credible commitments. In principle, it also makes it more
difficult for the Fund’s principal shareholders, like the United States, to use
the institution as a vehicle for advancing its foreign policy goals as opposed
to addressing problems of financial instability.
But while the IMF has become more transparent, questions can still be
asked about whether it has become more accountable.20 The representa-
tiveness of its managing board can be questioned. The procedure used for
selecting its managing director is hardly open, even-handed and merit-
based. These problems undermine the legitimacy of the institution and
hence the credibility of its commitments. Clearly, while there has been
progress in the direction of greater transparency and openness, there is still
an accountability problem.
This brings me to the agenda for the future. What key reforms remain
incomplete? What are the most important tasks going forward? How can
we continue to strengthen the international financial system while mini-
mizing corollary damage to emerging markets?
The first item on my list, while the least exciting, is at the same time the
most important. It is more business as usual. I mean redoubling efforts to
strengthen banking systems, enhance shareholder rights, improve the
effectiveness of corporate governance and heighten the transparency and
efficiency of financial markets. This effort to improve the structure, regula-
tion and efficiency of financial markets, first and foremost at the national
level, has been the touchstone of recent efforts to strengthen the interna-
tional financial architecture. There has been progress, but much remains to
be done. While this observation is commonplace, it is no less important for
that fact.
The second item is to create a credit culture in emerging markets. When
cheap credit becomes available, it flows in large amounts into the property
market and other speculative investments, with too little attention to the
decline in credit quality that occurs with the pursuit of higher risk projects.
More investment means faster growth, at least temporarily, which attracts
more foreign investors impressed by the improvement in economic perfor-
mance. The result is self-reinforcing, procyclical and dangerously unstable
dynamics.
The solution to this problem is to price credit risk more efficiently. Doing
so is the job of bond markets.21 The implication is that emerging markets
need better bond markets. Developing them will have a number of corol-
lary benefits, including reducing dependence on bank finance and limiting
currency and maturity mismatches. While I have doubts about the efficacy
of some fashionable initiatives in this area – the Asian Bond Fund, for
example – I nonetheless regard the objective, domestic bond market devel-
opment, as a high priority.22
The third priority item should be to move to still more flexible exchange
rates as a way of discouraging procyclical capital flows and permitting the
authorities to provide better tailoring of credit conditions to meet local
needs. China is the current poster boy for this argument, but it is only the
latest example of a country that would benefit from a more flexible rate.23
Emerging markets as a class would benefit from greater exchange rate
flexibility and less fear of floating. Getting there will require putting in place
the prerequisites for flexible inflation targeting – an independent central
bank, sound fiscal policies and market-determined interest rates – in
countries where they do not already exist. It will require moving away from
outdated commitments to export-led growth in countries like Taiwan and
South Korea where the export sector is no longer the exclusive locus of
learning effects and productivity spillovers, and where educating and retain-
ing knowledge workers will require a better balance of investment between
traded and non-traded goods sectors.24 In Latin America, it will require
measures to reduce liability dollarization and to help countries acquire the
capacity to borrow abroad in local currency.
This brings me to the fourth priority, namely measures to address the
inability of emerging markets to borrow abroad in their own currencies.
The evidence is overwhelming that institutional and policy reform at the
national level, by themselves, will not solve this problem anytime soon.25
Stronger policies and institutions help to solve many problems, to be sure,
but not obviously this one. Chile has admirably strong policies and institu-
tions, but it still cannot borrow abroad in its own currency. This suggests
that the problem resides at least in part in the structure of international
financial markets. International investors reap limited diversification
benefits from adding a variety of currencies to their portfolios while incur-
ring significant transactions costs from managing positions in relatively
small, illiquid markets. The smaller a country, the greater the difficulty it
has in getting its currency added to the global portfolio.26
Part of the solution is for the World Bank and regional development
banks to borrow and lend in these currencies. Their AAA credit rating
permits them to do so with relative ease. The issuance of World Bank debt
denominated in pesos will supply the high-quality benchmark asset that
investors need in order to price riskier credits. It will increase the installed
base of local currency denominated securities. It will enhance market liq-
uidity. In turn, this will make positions in assets denominated in these cur-
rencies more attractive to international investors. From this point of view,
the decision of the Asian Development Bank to issue in Chinese yuan, Thai
baht and Indian rupees is an important step. So was the decision of the
Inter-American Development Bank to include a local currency conversion
clause in a $300 million loan to Mexico in May of 2005. These are valuable
precedents; the World Bank and other regional development banks need to
respond in kind.
The fifth and final priority should be to address governance problems at
the Bretton Woods Institutions. Admittedly, it is easier to advocate reform
of IMF governance than to specify exactly what steps should be taken.
(If nothing else, this is a salutary reminder to high-income countries urging
developing economies to strengthen corporate governance that moving
from statements of high policy to implementation is easier said than
done.) Clearly, recommendations for the reform of IMF governance should
start with a statement of which objectives of governance are most poorly

achieved. My own view is that the key problems are those of stakeholder
representation and management accountability. Seats and votes in the IMF
reflect historical circumstances, specifically what was the disproportionate
weight of a few high-income countries, notably the United States, in the
much less multi-polar world economy that existed when the Fund was
created at the end of World War II. The countries with seats and votes dis-
proportionate to their place in the world economy and their stake in the
IMF have naturally resisted suggestions for reforms that reduce their
power. This leads to the under-representation of Asian countries that are
some of the fastest-growing members of the international economy and
that loom increasingly large in the operation of the international system
and the African countries that are the subject of the bulk of IMF programs.
The 85 per cent of IMF members that are not advanced industrial coun-
tries have only 50 per cent of the seats on the executive board. The 46 sub-
Saharan African countries that are chronic users of IMF resources are
represented by just two executive directors. The result is complaints that
inappropriate policies are being foisted on crisis economies by the govern-
ments of the United States and other high-income countries that dominate
decision-making in the IMF. Even if conditionality and lending decisions
were little different in a more representative Fund, giving other members
more voice in the formulation of such policies would enhance their legiti-
macy and thereby ‘ownership’ of programs by the crisis countries.
Ultimately, it would enhance the effectiveness of IMF policy.
On the question of accountability, the IMF is both inadequately
accountable to governments and other stakeholders in general and exces-
sively accountable to the advanced industrial countries and the United
States in particular. The opacity of IMF decision-making and the multidi-
mensional nature of the institution’s objective function make it hard to
hold those responsible for the policies of the institution accountable for
their actions. Who remembers who was fired in the wake of the Argentine
fiasco? The answer is not the managing director (or those responsible for
international economic policy at the US Treasury, which pushed for the
August 2001 loan) but the head of the Latin American department.
Arguably, differences in board composition and voting shares in the Fund
would have led to different consequences and greater accountability.
Hence, my recommendations for the reform of IMF governance would
be addressed at these aspects of governance. They would include address-
ing problems of representativeness by increasing the share of basic votes
and restoring the principal of universality that was acknowledged when the
institution was originally established. At present, each country has 250
basic votes plus one additional vote for each part of its quota equivalent to
100 000 Special Drawing Rights, on which basis both its contribution and
ability to draw resources are calibrated, while quotas are calculated on the
basis of a complicated set of formulas that in practice place substantial
weight on country size and current account volatility. Quotas have been
increased in the course of 11 reviews in order to reflect better the growth of
countries and their balances of payments, but without also increasing basic
votes. Consequently, the share of basic votes has fallen from 11 percent in
1945 and 14 percent in 1955, by which time additional members had been
admitted, to only 2 percent today. Increasing the share of basic votes would
go some way toward restoring the principle of universality and enhance the
representation of the poor countries that are frequently the subject of IMF
programs. In addition, using GDP at purchasing power parity rather than
GDP at market exchange rate prices in quota calculations would enhance
the voice of rapidly growing countries like those of Asia that are chroni-
cally under-represented in the Fund.
I would also favor appointing a single executive director for the
European Union. Rationalizing Europe’s board representation in this way
will free up chairs on the Executive Board for under-represented countries.
The majority of EU members have a single currency. Hence, there is no
more possibility of balance-of-payments problems among them than there
is scope for balance-of-payments problems between US states. If a financial
crisis develops in an EU member state, the bailout, for better or worse, will
be extended by its partners in the European Union and not by the IMF. In
addition, the Fund should rely more heavily on the International Monetary
and Financial Committee for defining the objectives and strategies of the
institution (including meetings of IMFC heads of state, which can substi-
tute for G7/8 summits). This will be possible insofar as the composition of
the IMFC becomes more representative as quotas and constituencies are
adjusted.
It should also be possible to strengthen the frankness of staff surveillance
by creating a presumption that staff studies written for Article IV consul-
tations will be published. Staff performance assessments could be revised
to give greater weight to ability to make independent, candid judgments,
and a presumption could be established that reports of the IMF’s own
Independent Evaluation Office would be published. By further enhancing
transparency, these steps will strengthen the accountability of the institu-
tion. They will enhance the efficiency of IMF governance by limiting
doubts that political pressures, both internal to the Fund and from national
sources, are unduly influencing staff assessments. They will reassure those
who worry that, in the absence of adequate transparency and accountabil-
ity, a handful of advanced economies have disproportionate influence in
the Fund as a result of their personal, political and intellectual connections.
Finally, selection of the managing director and deputy managing direc-

tors should be based on considerations of technical qualification rather
than nationality. The current arrangement whereby European governments
get to nominate the managing director of the Fund and the US government
gets to pick the first deputy managing director de-legitimizes the institution
in the eyes of many governments and NGOs. An open competition would
be more likely to lead to selection of the best man or woman for the job.
Given such a lengthy list of incomplete reforms, how immediate is the
residual crisis risk? Under what circumstances might a crisis arise? Might
we be on the verge of one now, given the rise in US interest rates and global
oil prices? If so, which countries will be affected?
As will be evident from the preceding, policies and institutions have been
strengthened significantly in the last 11 years, particularly since the Asian
Crisis of 1997–98. This means that a large shock would be required to desta-
bilize a major emerging market, much less the international financial system.
Such a shock is not inconceivable. Specifically, one can imagine simultane-
ous slowdowns in the US and China, removing the two locomotives that have
been also exclusively responsible for global growth.27 This might occur if
international investors lost faith in US policies and abruptly stopped
financing its current account deficit, and if the revaluation of the Chinese
yuan was bungled.28 The effects would be particularly disruptive if it turned
out that one or more major financial institutions in the advanced countries
had large positions in derivatives that left them exposed to the subsequent
change in relative prices. This is clearly on the minds of Federal Reserve
officials such as Timothy Geithner (2004) and Alan Greenspan (2005).
While this is not the most likely scenario, neither can we rule it out. If
growth slowed significantly in both the US and China, the demand for the
exports of emerging markets would fall. Global interest rates would rise
sharply as the FED found it necessary to raise interest rates faster than cur-
rently expected in response to the collapse of the dollar and its inflationary
consequences. Doubts about the stability of major financial institutions
and markets in the face of these rapid changes would cause a flight to
quality that further curtailed capital flows to emerging markets. This in turn
would force the sudden compression of current account deficits where these
had been allowed to reach high levels. And the reallocation of resources
necessary to effect that shift would require major changes in exchange rates,
specifically significant depreciation where current account deficits had been
allowed to widen, which could cause distress for banks and firms with cur-
rency and maturity mismatches on their balance sheets. All this will sound
disturbingly familiar to those who remember the Asian Crisis.
What countries are we talking about? The answer this time around is the
emerging markets of Central and Eastern Europe that have allowed large
current account deficits to develop and that are financing them mostly or
entirely with portfolio capital inflows. Specifically, I mean Hungary,
Romania and Turkey, which have current account deficits in the order of 6
to 8 percent of GDP at the time of writing. These countries are widely
regarded as ‘convergence plays’, the assumption that their interest rates will
come down to Western European levels as a result of their recent or immi-
nent accession to the European Union and the consequent ‘Europeanization’
of their policies.29 But if capital flows turn around, growth will slow, budget
deficits will widen and the stability of financial systems will be undermined.
Again, this will sound alarmingly familiar to those who recall the Asian
Crisis. Some say that this time is different – that, unlike previous episodes,
rising US policy rates and tightening global credit conditions will not pre-
cipitate a crisis in emerging markets. I am not so sure.
NOTES
1. This chapter is a considerable revision of a speech given on ‘Economics Day’ at the

University of Pennsylvania in May 2005. The chapter also draws on my plenary address
to the 3rd Annual PECC Finance Conference, Santiago, in June 2004 and the 50th
anniversary conference of Seoul National University, Seoul 2005. I thank participants
in all these meetings for their comments and reactions.
2. See Camdessus (1995).
3. The parallels with previous crisis periods are unmistakable; see Eichengreen and Fishlow
(1998).
4. In the early 1990s the reason was to cope with the Savings & Loan Crisis in the United
States. This time around, the rationale was fear of deflation.
5. And 1 percent off India’s.
6. To be sure, international banks are even more active in emerging markets than was the
case ten years ago. But they now fund most of their lending activities in those markets
out of local deposits. However much their presence enhances the efficiency of interna-
tional financial intermediation, this does not make foreign banks a source of external
finance.
7. After peaking at $150 billion in 1995, net private debt flows to developing countries fell
steadily for three years, approaching zero in 1999. Net bond-related flows fell by half,
from $49 billion in 1996 to $19 billion in 2001, while bank loans fell even more dramat-
ically, from $51 billion in 1998 to negative levels the following year. In 2000 net debt flows
to developing countries were negative; in 2001 and 2002 they were barely positive. If one
limits one’s attention to the subgroup of developing countries classified by the Institute
of International Finance as emerging markets – supposedly the leading destinations for
capital flows – the picture is essentially the same.
8. Turkey and Brazil are cases in point. The maturity of the debt lengthened from an
average of 7.7 months in January 2001 to 15.3 months in June 2004. Fixed rate debt has
been raised from 25 percent of total domestic debt in 2002 to 38 percent today. And
foreign-currency indexed or denominated debt has declined from 32 percent of domes-
tic debt in 2002 to 19 percent today. In Brazil, the average maturity of the domestic debt
has risen to some 20 months from little more than five at the time of the 1998–99 crisis.
9. See IIF (2006).
10. I will come back to this.
11. Most obviously in Asia but in Latin America as well.
12. Fear of floating, as this phenomenon is labeled by Calvo and Reinhart (2000), is still a
problem, in other words.
13. First-generation crises as modeled by Krugman (1979) and Flood and Garber (1984) are
crises where a deterioration in domestic policies leads to a speculative attack that col-
lapses the exchange rate ex post. This is in contrast with second-generation models where
the speculative attack on the exchange rate leads to the deterioration of domestic poli-
cies that validates the attack ex post (for example, Obstfeld, 1986). I claim the credit – or
the blame – for coining this terminology (in Eichengreen et al., 1995).
14. See Eichengreen and Park (2002).
15. At least so far as we can tell, given that information on the still largely unregulated hedge
fund industry is thin on the ground.
16. See for example Cady (2004); Christofides et al., (2003); Gelos and Wei (2002); and
Glennerster and Shin (2003).
17. See Tong (2004).
18. See IMF (2004).
19. See Fischer (2001).
20. Many of the same questions apply to the World Bank.
21. See Xie (2004) for a more elaborate discussion of this point. Banks, in contrast, tend to
be too easily influenced by their customers – and the government – and in any case they
operate in the less transparent, more information-impacted segments of the economy.
22. See Eichengreen (2004a) for some of these doubts.
23. Eichengreen (2004b) elaborates this argument. Currently, when the Chinese authorities
attempt to clamp down on the provision of bank credit to reduce property market spec-
ulation and the danger of overheating, capital simply flows in from abroad and enters
the property market through non-bank channels. This is the classic ‘trilemma’ of not
being able to have an independent monetary policy when the exchange rate is pegged and
the capital account is increasingly porous (whether the authorities like it or not). A more
flexible exchange rate would go a considerable way toward squaring this circle.
24. That is to say, more investment in universities and housing.
25. For evidence on this see Eichengreen and Hausmann (2005).
26. For evidence, see Eichengreen et al. (2003).
27. This is the scenario in Goldstein (2005).
28. Which might not in fact be two independent events.
29. Hungary became a member of the EU in 2004. Romania will join at the beginning of
2007 while Turkey’s (presumably lengthy) accession negotiations are underway.
REFERENCES
Cady, John (2004), ‘Does SDDS Subscription Reduce Borrowing Costs for
Emerging Market Economies’, IMF Working Paper No. 04/58, April,
www.imf.org.
Calvo, Guillermo and Carmen Reinhart (2000), ‘Fear of Floating’, Quarterly
Journal of Economics, 117, 379–408.
Camdessus, Michel (1995), ‘Drawing Lessons from the Mexican Crisis: Preventing
and Resolving Financial Crises – the Role of the IMF’, Address at the 25th
Washington Conference of the Council of the Americas, Washington, DC,
22 May, www.imf.org.
Christofides, Charis, Christian Mulder and Andrew Tiffin (2003), ‘The Link
between Adherence to International Standards of Good Practice, Foreign
Exchange Spreads, and Ratings’, IMF Working Paper No. 03/74.
Eichengreen, Barry (2004a), ‘The Unintended Consequences of the Asian Bond
Fund’, unpublished manuscript, University of California, Berkeley.
Eichengreen, Barry (2004b), ‘Is a Change in the Renminbi Exchange Rate in

China’s Interest?’, Asian Economic Papers, 4, 40–75.
Eichengreen, Barry and Albert Fishlow (1998), ‘Contending with Capital Flows:
What is Different About the 1990s?’, Occasional paper, Council on Foreign
Relations.
Eichengreen, Barry and Ricardo Hausmann (eds) (2005), Other People’s Money:
Debt Denomination and Financial Instability in Emerging Market Economies,
Chicago: University of Chicago Press.
Eichengreen, Barry and Bokyeong Park (2002), ‘Hedge Fund Leverage Before and
After the Crisis’, Journal of Economic Integration, 17, 1–20.
Eichengreen, Barry, Ricardo Hausmann and Ugo Panizza (2003), ‘Currency
Mismatches, Debt Intolerance and Original Sin: Why They Are Not the Same
and Why It Matters’, NBER Working Paper (December).
Eichengreen, Barry, Andrew Rose and Charles Wyplosz (1995), ‘Exchange Market
Mayhem: The Antecedents and Aftermath of Speculative Attacks’, Economic
Policy, 21, 249–312.
Fischer, Stanley (2001), ‘Farewell to the IMF Executive Board’, 30 August,
http://www.imf.org/external/np/speeches/2001/083001.htm.
Flood, Robert and Peter Garber (1984), ‘Collapsing Exchange Rate Regimes: Some
Linear Examples’, Journal of International Economics, 17, 1–13.
Geithner, Timothy (2004), ‘Perspectives on the U.S. Financial System’, Speech
to the Economic Club of New York, 27 May, http://www.newyorkted.org/
newsevents/speeches/2004/gei040527.htm.
Gelos, Gaston and Shang-Jin Wei (2002), ‘Transparency and International Investor
Behavior’, IMF Working Paper No. 02/174.
Glennerster, Rachel and Tongseok Shin (2003), ‘Is Transparency Good for You,
and Can the IMF Help?’, IMF Working Paper No. 03/132.
Goldstein, Morris (2005), ‘What Might the Next Emerging-market Financial Crisis
Look Like?’, unpublished manuscript, Institute for International Economics
(April).
Greenspan, Alan (2005), ‘Risk Transfer and Financial Stability’, Speech to the
Federal Reserve Bank of Chicago’s 41st Annual Conference on Bank Structure,
Chicago, Illinois, 5 May, www.federalreserve.gov/boarddocs/speeches/2005/
20050505/.
Institute of International Finance (IIF) (2006), ‘Capital Flows to Emerging Market
Economies’, Washington, DC: Institute of International Finance (15 September).
International Monetary Fund (IMF) (2004), ‘Progress Report to the International
Monetary and Financial Committee on Crisis Resolution’, 20 April,
www.imf.org.
Krugman, Paul (1979), ‘A Model of Balance of Payments Crises’, Journal of
Money, Banking and Credit, 11, 311–25.
Obstfeld, Maurice (1986), ‘Rational and Self-fulfilling Balance of Payments Crises’,
European Economic Review Papers and Proceedings, 76, 72–81.
Tong, Hui (2004), ‘Transparency Standards and Financial Market Efficiency’,
unpublished manuscript, University of California, Berkeley.
Xie, Andy (2004), ‘The Calm before the Storm’, Morgan Stanley Global Economic
Forum (26 May).
4. Capital controls, financial crises
and cures: simulations with an
econometric model for Malaysia*
Lawrence R. Klein, Roberto S. Mariano
and Süleyman Özmucur
1 INTRODUCTION
Capital flows to developing countries grew substantially during the post-

1990 period. With liberalization of the capital account a country becomes
more vulnerable to outside effects. Domestic policy-makers have a problem
in controlling their economy as seen in many countries (Mariano et al.,
2002). Controls have been advocated to deal with financial crises. They help
to slow down capital flight and depletion of international reserves. Krugman
(1998) indicated that countries facing major crisis may benefit from tempo-
rary imposition of controls, by giving time to lower domestic interest rates
and adopt a growth package. The reaction to the imposition of controls may
be negative in international financial markets. There is a delicate balance
between liberalization and controls. As Tobin (2001) concluded ‘Integration
of finance throughout the world can be of benefit to developing, emerging,
and transition economies, but not without some limits.’ The ‘Tobin Tax’
gained considerable popularity among many international economists as a
way to raise liquid funds, but Tobin often said that his intent was ‘to put sand
in the wheels’ in order to slow down crises in early stages of trouble.
Effects of capital controls, especially on economic growth, are
highly debated (among others Calvo, 1998; Edison et al., 2002; Goldstein
et al., 2000; IMF, 1998a, 1998b; Kahler, 1998; Kaminsky and Reinhart,
1998, 1999, 2000; Klein, 1998; Krugman, 1979, 1998; Rodrik, 2001; and
Salvatore, 1998). There are cross-country studies, as well as case studies
using different methodologies. This chapter offers another method to study
possible effects of capital controls, which takes feedback mechanisms in the
economy into consideration. The proposed approach is the use of a macro-
econometric model for Malaysia. A simultaneous equation model has the
advantage of incorporating various aspects of the economy, and analyzing
134
Capital controls, financial crises and cures: Malaysia 135
possible effects on many variables as well as on the growth, which is gener-

ally the focus of attention.
The major conclusion of the chapter is that capital controls in Malaysia
had many desired effects on growth, inflation and balance of payments,
without doing much harm to the attraction of foreign direct investment, in
the medium run.
The chapter is in four sections. Section 2 is devoted to a very brief review
of earlier studies. Main characteristics of the macroeconometric model are
summarized in the third section. Simulation results are given in section 4,
and major conclusions are stated in the final section.
2 A BRIEF REVIEW OF STUDIES ON CAPITAL

CONTROLS AND FINANCIAL CRISES
It is argued that with liberalization of the capital account a country

becomes more vulnerable to outside effects. The instability caused by liber-
alization can be compensated by a higher growth rate to follow, largely due
to greater availability of capital (Summers, 2000). Stiglitz (2000) draws
attention to the need for the distinction between various forms of capital
flows, and advocates the flow of foreign direct investment [FDI]). Singh
(2003) argues that ‘even FDI, if unregulated, may do more harm than good
to many countries’.
Edison et al. (2002) survey the effects of capital account and stock
market liberalization on economic growth, and conclude that there is
support for a positive effect of capital account liberalization on growth,
especially for developing countries.
Eichengreen and Leblang (2003) state that capital controls influence
macroeconomic performance through two channels, directly with a positive
impact on resource allocation and efficiency and indirectly by limiting the
disruptive effects of crises. There are significant flows of capital toward
sectors where the rate of return is high, provided financial markets are
working well. In that case, liberalization leads to a more efficient allocation
of resources and to faster economic growth. If there are problems in
domestic and/or international financial markets, the consequences can be
less favorable. They also demonstrate the ability of controls to neutralize
the disruptive effects of crises. Since vulnerability to crises varies across
countries and with the structure and performance of the international
financial system, ‘the effects of capital account liberalization on growth are
contingent and context specific’. They present empirical results using a
panel of 21 countries covering a longer period of 1980–97, and a wider
panel for the post-1971 period to support their arguments.
Kaplan and Rodrik (2001) study how Malaysia performed compared

with Thailand and Korea when they were all undergoing IMF programs.
The reason for making that comparison was their view on Malaysian poli-
cies in the summer of 1998, which seemed to be unsustainable. They argue
that pressure against the ringgit was building up, and the economic decline
was not about to be reversed on its own. Controls were introduced to
contain the joint effects of crisis, stabilization and especially contagion
to others in the region. They use difference-in-differences method and
monthly and quarterly data to show that Malaysian policies including
capital controls produced faster economic recovery compared with its
neighbors.
Edison and Reinhart (2000) use monthly and daily data to study
effectiveness of capital controls in Malaysia, Thailand, Philippines and
South Korea. Their GARCH(1,1) models using daily data indicate that
capital controls were significant in reducing volatility in daily interest rates
and daily changes in interest rates in Malaysia, and increasing volatility in
daily stock returns in Thailand. They conclude that controls led to a greater
stability in interest rates and exchange rates in Malaysia, but not in
Thailand. A possible explanation given for the difference was the existence
of offshore banking centers that provided leakages and arbitrage opportu-
nities in Thailand. Attention was drawn to differences between experiences
of Thailand and Malaysia. First, Thailand was undergoing speculative
attacks when they tried to use capital controls as a defense mechanism;
Malaysia was not. Second, Malaysian controls were quite broad and aimed
to eliminate all loopholes and speculations, while controls in Thailand were
not as comprehensive.
Mariano et al. (2002) constructed a Markov-switching autoregressive
model that allows intercepts, lag coefficients and error variances to switch
stochastically over time according to the value taken by a latent Markov
chain describing the vulnerability of the country’s currency to speculative
attacks. Estimates of Markov-switching models for Indonesia, Malaysia,
the Philippines and Thailand were obtained using monthly data from 1974
to 1998. The estimated models are mean-switching and variance-switching
models of month-to-month percentage changes in nominal exchange rates,
with transition probabilities in the Markov chain depending on deviations
of real effective exchange rates from trend (REER), year-on-year percent-
age changes in the ratio of money supply (M2) to gross international
reserves, and year-on-year percentage changes in real domestic credit. Real
exchange rate overvaluation (relative to trend) is an important explanatory
variable for predicting a switch from normal periods to vulnerable periods.
The two other early warning variables, growth in real domestic credit and
M2/reserves, showed only moderate importance, entering into the final
specifications only in selected countries. There is good evidence for the

importance of non-performing loans in crisis prediction. However,
differences in definitions and treatment of these loans among countries
render the use of it difficult. After the Asian Crisis, there have been efforts
for standardization of the definition of non-performing loans, which will
eventually enable researchers to use it as one of the critical indicators.
For all four countries, one regime is a normal period with zero mean and
low volatility in exchange rate percentage changes. The other regime is a
financially vulnerable period with positive mean and high volatility in the
dependent variable. Forecast probabilities of a vulnerable period, given
past information, are calculated from the estimated model. Such forecast
probabilities rise substantially for Malaysia and Thailand in early 1997 –
before the start of the Asian Crisis in July 1997 – but do not send strong
signals for Indonesia and the Philippines.
3 MAIN CHARACTERISTICS OF THE

MACROECONOMETRIC MODEL
The model is a simultaneous equations system designed to address the

needs in short-term and medium-term forecasting and in scenario and
policy analysis (Klein et al., 2002). The model covers all the major sectors
of Malaysia’s economy and the feedback linkages among them. It serves as
a quantitative tool of Bank Negara in forecasting economic variables at a
disaggregated level, in analyzing the macroeconomic impact of policy mea-
sures, and in simulating scenarios of interest to the Bank. There is a very
detailed treatment of balance of payments, public sector and the financial
sector.
The macroeconometric model contains 438 equations in seven blocks. Of
these equations, 259 are stochastic, and 179 are identities. There are 607
variables, of which 438 are endogenous, and 169 are exogenous. The rela-
tively large size of the model is a direct result of the need for capturing basic
features of the Malaysian economy. These features may be summarized as
the openness of the economy, diversity of exports ranging from agricultural
commodities to semiconductors, diversity in the banking sector, and the
significance of the public sector in the production as well as in the decision-
making process.
There are ten production and employment sectors. These sectors are agri-
culture, mining, manufacturing, electricity, construction, wholesale and
retail trade, transportation and communication, finance, government ser-
vices, and other services. In addition, production of major commodities in
agriculture and mining are also treated as endogenous variables. These
commodities are rubber, saw logs, palm oil, kernel oil, cocoa, crude oil, tin
and natural gas. Industrial production indexes for major manufacturing
sub-categories are also estimated in the model. There are five private invest-
ment categories, namely, agriculture, manufacturing, mining, construction
and other.
There is a very detailed section on the export component of the balance
of payments. The capital account is also treated as endogenous. Both
foreign direct investment and portfolio investment are treated endogenously
in the model. Interest rates, exchange rates and equity prices are explained
within the model.
Because of the public sector’s significance, the model has a detailed
public sector block. Public revenues are determined within the model, using
the appropriate base for the revenue being considered. Current expend-
itures are also treated as endogenous. Development expenditures are
treated as exogenous. Non-financial public enterprises, state and local gov-
ernments’ accounts are also studied in detail. The financial sector is also
treated in detail. Monetary base and money supply are determined by the
net claims of the government, net claims of the private sector, and net
claims of the external sector. Money demand equations are estimated,
assuming equilibrium in the money market. Banking sector balance sheets
are treated in detail.
The in-sample forecasting performance of the model is assessed through
static and dynamic simulations of the model over the period 1990–2001. The
one-period simulation indicates that the major macroeconomic variables
can be predicted within reasonable error margins (in terms of mean absolute
percentage error or MAPE): GDP with 6 percent, private consumption with
5.1 percent, total investment with 5.2 percent, and prices (CPI and the
implicit GDP deflator) with 1.5 percent and 3 percent, respectively. For
the unemployment rate and the interest rate, the relevant performance indi-
cator is either mean absolute error (MAE) or root mean square error
(RMSE). The mean absolute errors are 124 basis points for the unemploy-
ment rate. These errors are rather low when compared with those from other
models; the model is relatively successful in tracking historical values.
4 SIMULATIONS WITH THE MODEL
The model was solved (dynamic solution) for the 1997–2005 period for a
baseline solution. The scenario of ‘no capital controls’ was the alternative
(simulation). In the model, there is a capital control dummy that is equal to
1 in 1994, 1998 and 1999, and 0 for other years.1 This dummy variable is set
to 0 (no controls) for the period 1997–2005 to obtain the alternative
solution. The difference (or the percentage difference) between the simula-
tion and the baseline were calculated to see the possible effects of the ‘no
capital controls’ policy. The results are obtained for all endogenous vari-
ables, but for brevity, results for only selected variables are provided here.
The real GDP would have been 0.08 percent higher in 1998 without con-
trols (Figure 4.1). This effect turns to negative 0.01 percent in 1999, nega-
tive 0.12 percent in 2000 and 0.02 percent in 2001. The overall effect is
negative, indicating that GDP would have been lower without capital con-
trols. The GDP deflator would have been 0.04 percent higher in 1998, and
0.03 percent higher in 1999. The percentage differences are negative, but
quite small in magnitude, after 1999. The overall effect is positive, indicat-
ing that GDP deflator would have been higher without capital controls.
The real exports of goods and services would have been 0.15 percent
higher in 1998 without controls (Figure 4.1). This effect turns to negative
0.01 percent in 1999, negative 0.22 percent in 2000, and 0.02 percent in
2001. The overall effect is negative, indicating that real exports of goods
and services would have been lower without capital controls. The real
imports of goods and services would have been 0.10 percent higher in 1998
without controls. This effect turns to negative 0.01 percent in 1999, nega-
tive 0.12 percent in 2000, and 0.02 percent in 2001. The overall effect is
negative, indicating that real imports of goods and services would have
been lower without capital controls.
The real private consumption expenditures would have been 0.08 percent
higher in 1998 without controls. This effect turns to negative 0.01 percent
in 1999, negative 0.11 percent in 2000 and 0.01 percent in 2001. The overall
effect is negative, indicating that real private consumption expenditures
would have been lower without capital controls. The real private fixed
investment would have been 0.23 percent higher in 1998 without controls.
This effect turns to negative 0.01 percent in 1999, negative 0.28 percent in
2000 and 0.02 percent in 2001. The overall effect is negative, indicating
that real private fixed investment would have been lower without capital
controls.
The real government consumption expenditures would have been 0.028
percent lower in 1998, and 0.022 percent lower in 1999 without controls.
This effect turns to 0.02 percent in 2001, 0.005 percent in 2001 and 0.01
percent in 2002. The overall effect is negative, indicating that real govern-
ment consumption expenditures would have been lower without capital
controls. The real government fixed investment would have been 0.010
percent lower in 1998, and 0.008 percent lower in 1999 without controls.
This effect turns to 0.006 percent in 2001, 0.002 percent in 2002. The overall
effect is negative, indicating that real government fixed investment would
have been lower without capital controls.
Percentage Deviation [(Simulation–Baseline)/Baseline*100]
Real
RealGDP
GDP GDP Deflator
0.12
0.12 0.04
0.08
0.08 0.03
0.04
0.04 0.02
0.00
0.00 0.01
–0.04
–0.04 0.00
–0.08
–0.08 –0.01
–0.12
–0.12 –0.02
–0.16
–0.16 –0.03
97
97 98
98 99
99 00
00 01
01 02
02 03
03 04
04 05
05 97 98 99 00 01 02 03 04 05
140
Real
RealExports
ExportsofofGoods
Goodsand
andServices
Services Real Imports of Goods and Services
0.2
0.2 0.12
0.08
0.1
0.1
0.04
0.0
0.0 0.00
–0.1
–0.1 –0.04
–0.08
–0.2
–0.2
–0.12
–0.3
–0.3 –0.16
97
97 98
98 99
99 00
00 01
01 02
02 03
03 04
04 05
05 97 98 99 00 01 02 03 04 05
Private Consumption Private Fixed Investment
0.12 0.3
0.08 0.2
0.04 0.1
0.00 0.0
–0.04 –0.1
–0.08 –0.2
–0.12 –0.3
97 98 99 00 01 02 03 04 05 97 98 99 00 01 02 03 04 05
141
Government Consumption Government Fixed Investment

0.02 0.008
0.01 0.004
0.00 0.000
–0.01 –0.004
–0.02 –0.008
–0.03 –0.012
97 98 99 00 01 02 03 04 05 97 98 99 00 01 02 03 04 05
Figure 4.1 Effects of capital controls on real indicators and the GDP deflator
The government’s total tax revenues would have been 0.08 percent higher
in 1998, and 0.10 percent higher in 1999 without controls (Figure 4.2). This
effect turns to negative 0.08 percent in 2000, and negative 0.12 percent in
2001. The overall effect is a small negative, indicating that government tax
revenues would have been lower without capital controls. Similar results are
obtained for total revenues of the government. The federal budget balance
would have been about 50 million ringgits higher both in 1998 in 1999
without controls. This effect turns to negative 60 million in 2000, and nega-
tive 80 million in 2001. The overall effect is negative, indicating that federal
government budget balance would have been lower without capital con-
trols. The federal government debt would have been 40 million ringgits
lower in 1998, 90 million lower in 1999 and 40 million lower in 2000 without
controls. This effect turns to 30 million in 2001, and remains in the neigh-
borhood of 10 million until 2005. The overall effect is negative, indicating
that federal government debt would have been lower without capital con-
trols. It should be noted that this is largely driven by domestic debt.
The effect of capital controls on net private capital flows is of special
interest. The net effect on long-term private capital flows is negative, but
very small in magnitude (Figure 4.3). The effect on the short-term capital
flows is also small in magnitude and negative. The net private short-term
capital flows would have been 200 million higher in 1999. The figure turns
to be negative 300 in 2000, and negative 280 in 2001, and positive 400 in
2002. The net effect is very small, especially when these are compared with
the effects on the current account and the balance of payments. There is a
significant effect on the current account, and hence the basic balance and
the overall balance of payments. The balance of payments could have been
2500 million lower in 1998 and 2800 million lower in 1999. The effects are
much smaller in following years. The corresponding figures for the basic
balance were 2600 million in 1998 and 3200 million in 1999. The major
determinant of the change in the basic balance is the change in the current
account balance.
The initial effect on interest rates would have been higher in 1998, but
generally smaller in 1999 and 2000 if there were no capital controls (Figure
4.4). The overall effect on interest rates and exchange rates is quite small.
However, since the model is an annual one it is not possible to study another
interesting question, namely the volatility of interest rates, and stock prices
as done by Edison and Reinhart (2000).
All in all, temporary controls of capital flows in Malaysia achieved
intended goals of keeping inflation under control, and resuming growth
without much distortion in foreign exchange and financial markets. A coor-
dinated stabilization policy was the key to a successful outcome.
Tax Revenue (percentage Total Government Revenue (percentage
deviation) deviation)
0.12 0.10
0.08
0.04 0.05
0.00
0.00
–0.04
–0.08
–0.05
–0.12
–0.16 –0.10
95 96 97 98 99 00 01 02 03 04 05 95 96 97 98 99 00 01 02 03 04 05
143
Federal Government Budget Balance Federal Government Debt (domestic+foreign)

(deviation) (deviation)
60 40
40 20
20 0
million ringgits
million ringgits
0 –20
–20 –40
–40 –60
–60 –80
–80 –100
95 96 97 98 99 00 01 02 03 04 05 95 96 97 98 99 00 01 02 03 04 05
Figure 4.2 Effects of capital controls on government revenues, balance and debt
Deviation (Simulation-Baseline)
Net
NetPrivate
PrivateLong-term
Long-termCapital
Capital Net Private Short-term Capital
8080 500
400
4040
300
million ringgits
million ringgits
million ringgits
00 200
100
–40
–40
0
–80
–80 –100
–200
–120
–120
–300
–160
–160 –400
9797 9898 9999 0000 0101 0202 0303 0404 0505 97 98 99 00 01 02 03 04 05
144
Balance
BalanceofofPayments,
Payments,Overall
OverallBalance
Balance Balance of Payments, Basic Balance
500
500 400
00 0
–400
–500
–500
million ringgits
million ringgits
million ringgits
–800
–1000
–1000 –1200
–1500
–1500 –1600
–2000
–2000
–2000
–2400
–2500
–2500 –2800
–3000
–3000 –3200
9797 9898 9999 0000 0101 0202 0303 0404 0505 97 98 99 00 01 02 03 04 05
Figure 4.3 Effects of capital controls on balance of payments

Base Lending Rate Rate on Savings Deposits
(deviation) (deviation)
0.002 0.0012
0.0008
0.001
0.0004
Percentage points
Percentage points
0.000 0.0000
–0.0004
–0.001
–0.0008
–0.002 –0.0012
–0.0016
–0.003
–0.0020
–0.004 –0.0024
97 98 99 00 01 02 03 04 05 97 98 99 00 01 02 03 04 05
145
Discount Rate on Treasury Bills (12-month) Exchange Rate (Ringgit/US$)

(deviation) (percentage deviation)
0.03 0.3
0.02 0.2
Percentage points
0.01 0.1
0.00 0.0
–0.01 –0.1
–0.02 –0.2
–0.03 –0.3
97 98 99 00 01 02 03 04 05 97 98 99 00 01 02 03 04 05
Figure 4.4 Effects of capital controls on interest rates and exchange rates
5 CONCLUSION
Results based on alternative methods indicate that capital controls worked

well in Malaysia, but not in Thailand. The timing, the dosage, degree of
commitment to and the duration of a policy as well as initial conditions are
critical determinants of the success of a policy. Controls over capital flows
are no exception to this general rule.
A simultaneous equation macroeconometric model is used to study pos-
sible effects of capital controls. Simulation results indicate that controls
improved the performance of the Malaysian economy. A probable channel
is the reduction of the level of uncertainty surrounding the key prices such
as the exchange rate and the interest rate.
There are lessons to be learned from the Malaysian experience. Probably,
the first requirement is to have the house in order. This is a necessary but
not a sufficient condition. A small open economy is always vulnerable to
external shocks, and policy making is always more difficult in such an open
economy because of lack of control over most of the policy instruments
usually employed. There is definitely a need for early warning indicators of
financial vulnerability such as real exchange rate overvaluation, growth in
real domestic credit and ratio of money supply to reserves.
NOTES
* The views expressed in this chapter are those of the authors and do not necessarily rep-
resent those of the Central Bank of Malaysia.
1. Using a dummy variable indicates the existence of controls, but not the intensity of con-
trols. There are attempts to measure the intensity to alleviate the limitation of using a
dummy variable. See Edison and Warnock (2002).
REFERENCES
Calvo, G. (1998), ‘Capital Flows and Capital Market Crises: The Simple Economics
of Sudden Stops’, Journal of Applied Economics, 1 (1), 35–54.
Edison, H.J. and C.M. Reinhart (2000), ‘Capital Controls During Financial Crises:
The Case of Malaysia and Thailand’, Board of Governors of the Federal Reserve
System International Finance Discussion Papers No. 662, Washington, DC,
March (mimeo).
Edison, H.J. and F.E. Warnock (2002), ‘A Simple Measure of the Intensity of Capital
Controls’, Board of Governors of the Federal Reserve System International
Finance Discussion Papers No. 708, Washington, DC, October (mimeo).
Edison, H.J., M. Klein, L. Ricci and T. Slok (2002), ‘Capital Account Liberalization
and Economic Performance: A Review of Literature’, IMF Working Paper
02/120, Washington, DC.
Eichengreen, B. and D. Leblang (2003), ‘Capital Account Liberalization and

Growth: Was Mr Mahathir Right?’, NBER Working Paper No. 9427, January.
Goldstein, M., G. Kaminsky and C.M. Reinhart (2000), Assessing Financial
Vulnerability: An Early Warning System for Emerging Markets, Institute for
International Economics, Washington, DC (June).
IMF (1998a), International Financial Markets, Washington, DC.
IMF (1998b), World Economic Outlook, Washington, DC.
Kahler, M. (ed.) (1998), Capital Flows and Financial Crises, Ithaca: Cornell
University Press.
Kaminsky, G.L. and C.M. Reinhart (1998), ‘Financial Crises in Asia and Latin
America: Then and Now’, American Economic Review, 88 (2), 444–8, May.
Kaminsky, G.L. and C.M. Reinhart (1999), ‘The Twin Crises: The Causes of
Banking and Balance of Payments Problems’, American Economic Review, 89 (3),
473–500.
Kaminsky, G. and C.M. Reinhart (2000), ‘On Crises, Contagion, and Confusion’,
Journal of International Economics, 51 (1), June, 145–68.
Kaplan, E. and D. Rodrik (2001), ‘Did the Malaysian Capital Controls Work?’,
NBER Working Paper No. 8142, Cambridge, MA, February.
Klein, L.R. (1998), ‘Financial Crises in Asia and Development Issues’, Department
of Economics, University of Pennsylvania, Philadelphia, PA (mimeo).
Klein, L.R., R.S. Mariano and S. Özmucur (2002), ‘BNM Macroeconometric
Model of Malaysia’, Bank Negara Malaysia, Kuala Lumpur, July (mimeo).
Krugman, P. (1979), ‘A Model of Balance of Payments Crises’, Journal of Money,
Credit, and Banking, 11 (August), 311–25.
Krugman, Paul (1998), ‘Saving Asia: It’s Time to Get Radical’, Fortune,
8 September.
‘Predictive Models of Economic Financial Crises’, Report for the Economic
Research Forum (ERF) for the Arab countries, Iran and Turkey, December
(mimeo).
Rodrik, D. (2001), Has Globalization Gone Too Far?, Institute for International
Economics, Washington, DC.
Salvatore, D. (1998), ‘Capital Flows, Current Account Deficits, and Financial
Crises in Emerging Market Economies’, International Trade Journal, 12 (1), 5–22,
Spring.
Singh, A. (2003), ‘Capital Account Liberalization, Free Long-term Capital Flows,
Financial Crises and Economic Development’, Eastern Economic Journal,
Spring, 29 (2), 191–216.
Stiglitz, J.E. (2000), ‘Capital Market Liberalization, Economic Growth, and
Instability’, World Development, 28 (6), 1075–86.
Summers, L. (2000), ‘International Financial Crises: Causes, Prevention and
Cures’, American Economic Review Papers and Proceedings, May, 90 (2), 1–16.
Tobin, J. (2001), ‘International Trade and Financial Markets’, Japan and the World
Economy, 13, 499–502.
5. The case of the missing market: the
bond market and why it matters for
financial development
Richard J. Herring and Nathporn Chatusripitak
1 INTRODUCTION
Over the last decade, interest in the role of finance in economic growth has
revived. Building from the pioneering work of Goldsmith (1965) and the
insights of Shaw (1973) and McKinnon (1973), the more recent work exam-
ines the role of financial institutions and financial markets in corporate
governance and the consequent implications for economic growth and
development. Levine (1997) and Stulz (2000) have provided excellent
reviews of this literature and Allen and Gale (2000) have extended it by
developing a framework for comparing bank-based financial systems with
market-based financial systems.1 Although the literature addresses ‘capital
markets’, on closer inspection the main focus is really equity markets. Bond
markets are almost completely overlooked.2
Although the omission of the bond market is not defended in the litera-
ture, one could argue that it does little violence to reality. As Table 5.1
shows, in most emerging economies in Asia, bond markets are very small
relative to the banking system or equity markets.3 Moreover, the most strik-
ing theoretical results flow from a comparison of debt contracts with equity
contracts and at a high level of abstraction bank lending can proxy for all
debt. In any event, data are much more readily available for equity markets
and the banking system than for bond markets, even in the United States.
In contrast to the academic literature, however, policy-makers have
become increasingly concerned about the absence of broad, deep, resilient
bond markets in Asia. The World Bank (Dalla et al., 1995, p. 8) has pub-
lished a study of emerging Asian bond markets urging that Asian economies
‘accelerate development of domestic . . . bond markets’, and has launched
another major study aimed at helping countries develop more efficient bond
markets. Along with Malaysia, Hong Kong has led the way. Hong Kong has
succeeded in fostering development of an active fixed-income market in
148
The case of the missing market 149
Table 5.1 The size of bond versus equity and loan markets (as a
percentage of GDP, average 1990–97)
Bank Credit Stock Market Bonds

Capitalization Outstanding
Hong Kong 140.7 193.2 11.3
Indonesia 46.0 18.4 1.8
Malaysia 103.2 197.0 68.5
Philippines 33.2 51.6 33.1
Singapore 96.6 137.5 28.4
South Korea 96.7 35.2 40.3
Thailand 87.4 57.3 9.9
Canada 92.4 57.3 81.8
Japan 176.5 77.1 90.1
United States 111.9 79.2 138.2
Euro-area 86.0 33.5 83.9
Non-euro EU 105.4 93.0 67.6
Source: World Bank Financial Structure and Economic Development Database.
exchange fund bills and notes even though the government has not run
significant deficits (Sheng, 1994 and Yam, 1997). In 1998, the Asia-Pacific
Economic Cooperation (APEC 1999) formed a study group to identify best
practices and promote the development of Asian bond markets. Much of
this official concern stems from the perception that the absence of bond
markets made several Asian economies more vulnerable to financial crisis.
The Governor of the Bank of Thailand (Sonakul, 2000) reflected this view
when he observed, ‘If I [could] turn back the clock and have a wish [list] . . .
high in its ranking would be a well-functioning Thai baht bond market’. (In
the concluding section we will discuss not only Thailand’s efforts to nurture
a bond market, but also the broader initiatives of the Executives’ Meeting of
East Asia-Pacific Central Banks [EMEAP].)
In this chapter, we consider why bond markets are so underdeveloped
relative to equity markets and the banking sector. In addition, we investi-
gate what the absence of a well-functioning bond market may imply for
savings, the quality and quantity of investment and for risk management.
Our analysis leads us to conclude that the absence of a bond market may
render an economy less efficient and significantly more vulnerable to
financial crisis.
If a government wishes to enhance efficiency and financial stability by
nurturing the development of a bond market, what are the appropriate
policy remedies? We review the key requirements for developing a broad,
deep, resilient bond market and conclude with an analysis of recent financial
development in Thailand, which is broadly representative of the wide range
of countries that have highly developed equity markets and a large banking
sector, but until very recently, only the most rudimentary bond market.
2 OVERVIEW OF THE FINANCIAL SECTOR AND

FLOW OF FUNDS ANALYSIS
The impact of the financial sector on the real economy is subtle and
complex. What distinguishes financial institutions from other firms is the
relatively small share of real assets on their balance sheets. Thus, the direct
impact of financial institutions on the real economy is relatively minor.
Nonetheless, the indirect impact of financial markets and institutions on
economic performance is extraordinarily important. The financial sector
mobilizes savings and allocates credit across space and time. It provides not
only payment services, but more importantly products that enable firms
and households to cope with economic uncertainties by hedging, pooling,
sharing and pricing risks. An efficient financial sector reduces the cost and
risk of producing and trading goods and services and thus makes an
important contribution to raising standards of living.
The structure of financial flows can be captured in flow of funds analy-
sis, a useful analytical tool for tracing the flow of funds through an
economy. This device has been used for evaluating the interaction between
the financial and real aspects of the economy for nearly half a century
(Copeland, 1955 and Goldsmith, 1965, 1985). The basic building block is
a statement of the sources and uses of resources for each economic unit
over some period of time, usually a year.
Our analysis of the relationship between the financial sector and eco-
nomic performance will proceed in stages. In the first stage, we consider
how an economy would perform without a financial sector in order to
provide a clear benchmark for comparison. The second stage introduces
direct financial claims in an environment with severe information asym-
metries. The third stage considers financial intermediaries that transform
the direct obligations of investors into indirect obligations of financial
intermediaries that have attributes that savers prefer. The fourth stage intro-
duces the government sector and the international sector.
2.1 Savings and Investment without Financial Markets or Institutions
In order to understand the role of the financial sector in enhancing eco-

nomic performance, it is useful to begin with a primitive economy in which
Table 5.2 Sources and uses of funds for the household sector
Uses (U) Sources (S)

Flows Flows
• Real assets • Net worth (savings)
• Equity • Financial liabilities
• Direct financial assets • Foreign financial liabilities
• Indirect financial assets
• Claims on government
• Foreign financial assets
• Total assets • Total liabilities and net worth
there is no financial sector. Without financial instruments, each household

would necessarily be self-financing and would make autonomous savings
and investment decisions without regard for the opportunity cost of using
those resources elsewhere in society.
In this case households are the fundamental economic unit of analysis
and the sources and uses of resources accounts (Table 5.2) reflect the
changes in each household’s balance sheet over the year. Since, at this point
financial instruments do not exist, all assets are real and there are no lia-
bilities. (Other categories of financial instruments that will be introduced
later are in italics.) Changes in real assets, here the accumulation of goods,
reflect savings or changes in net worth; dissaving results in corresponding
declines in real assets.
The fundamental decisions that influence economic performance –
(1) how much to consume and save; (2) how to allocate the flow of savings;
and (3) how to allocate the existing stock of wealth – depend on each
autonomous household’s opportunities, present and expected future
income, tastes, health, family composition, the costs of goods and services
and confidence in the future. Although barter transactions among house-
holds would permit some specialization in production, the extent of spe-
cialization would be severely limited by the necessity for each household to
be self-financing.
By aggregating sources and uses of accounts for each economic unit, a
matrix of flows of funds can be constructed for the entire economy. For
illustrative purposes, we present a primitive economy with two households
in Table 5.3. Although other sectors are listed, they are irrelevant at this
stage of the analysis because we have assumed that there are no financial
instruments that can link one sector to another. These parts of the matrix
(which will be introduced later) are in italics.
In this example, we have inserted arbitrary entries for each household.
Household 1 is saving 80 units of current income, while household 2 saves
Table 5.3 The flow of funds matrix for an economy without a financial
sector
Sectors Household Household Non- Financial Rest of Total

1 2 financial Institutions World
Firms
Flows of U S U S U S U S U S U S
Real
Income
Savings 80 40 120
Real 80 40 120
Assets
Financial
Flows
Equity
Fixed
Income
Instruments
Indirect
Financial
Instruments
Financial
Instruments
Issued by
Foreign
Residents
Totals 80 80 40 40 120 120
only 40. If productive opportunities were fortuitously distributed across

households in such a way that each household earned precisely the same
rate of return on its stock of real assets, this economy could prosper
without a financial sector. Such an outcome is highly unlikely, however,
because investment opportunities and desired savings are apt to differ
markedly across households. Moreover, there is no assurance that house-
holds with high savings have commensurately greater or more profitable
real investment opportunities.
If, for example, household 2’s desired investment exceeded its current
savings, its investment would have to be postponed until it could accumu-
late sufficient savings. This would be true even if its investment opportun-
ities offer substantially higher returns than the investment opportunities
available to household 1. Assume further that household 1’s investment
opportunities are less productive than household 2’s. Since household 1 does
not have access to the superior investment opportunities of household 2, it
may undertake inferior investment projects or save less. Society’s flow of

savings is inefficiently allocated and the stock of investment is less produc-
tive than it might otherwise be. Both the quality of capital formation and
the quantity of future output suffer, and the standard of living in this society
is less than it would be if household 1 could be induced to transfer some of
its resources to household 2 in exchange for a financial claim.
A ‘financial claim’ is a contractual agreement entitling the holder to a
future pay-off from some other economic entity. Unlike a real asset, it does
not provide its owner with a stream of physical services. Rather it is valued
for the stream of pay-offs it is expected to return over time. The financial
claim is both a store of value and a way of redistributing income over time,
which may be much more attractive to savers than the stream of services
that savers could anticipate from their own investment opportunities in real
assets.
Given the assumptions in our simple case, it is conceivable that a bargain
could be arranged between household 1 and household 2. In exchange for
household 1’s real assets, household 2 could issue a financial claim to
household 1 that would promise a more attractive pattern of pay-offs than
the investment opportunities available to household 1. This reallocation of
assets between household 1 and household 2 could increase the return on
capital formation for this society. Indeed, the possibility of investing in
financial claims that are more attractive than household 1’s own real invest-
ment opportunities might even increase the savings of household 1 and
thus increase the total quantity as well as the quality of capital formation.4
2.2 Flows with Direct Financial Claims but no Secondary Market
To examine how a financial sector affects the economy, we will introduce

the direct financial claims suggested above. The exposition is further
simplified by introducing a second sector in the economy. Assume that
firms specialize in investing in real assets financed by issuance of direct
financial claims, while households specialize in saving and investing in these
direct financial claims. Financial claims are reflected in the flow of funds
accounts as sources of funds for firms and as uses of funds for households.
Households continue to hold real assets, but most real assets appear on the
balance sheets of firms. At this stage, we will assume that direct claims
cannot be traded in well-organized secondary markets. Issues of direct
claims are, in effect, private placements that will be held by households until
they mature or the firm is liquidated.
The flow of funds matrix in Table 5.4 illustrates such a system and reflects
the sort of qualitative changes that occur when an economy first begins to spe-
cialize in production. It differs from the flow of funds matrix in Table 5.3 in
Table 5.4 The flow of funds matrix for an economy with private placement
of direct claims
Sectors Household Household Non- Financial Rest of Total

1 2 financial Institutions* World*
Institutions
Real
Income
Savings 87 43 10 140
Real Assets 7 2 131 140
Financial
Flows
Equity 60 31 91 91 91
Fixed 20 10 30 30 30
Income
Instruments
Indirect
Financial
Assets
Financial
Instruments
Issued
by Foreign
Residents
Totals 87 87 43 43 131 131 261 261
Note: * Entries for these columns are developed in later tables.
three respects: (1) firms hold most of the real assets; (2) households hold direct
financial claims on firms in lieu of most of their previous holdings of real
assets; and (3) household savings have increased by (an arbitrary) ten units to
reflect the enhanced level of income that could be gained from reallocating
real assets to more productive uses. Generally, the higher an economy’s per
capita income, the higher the ratio of financial assets to real assets.
What makes this reallocation of resources possible? What induces house-
holds to exchange real assets for direct financial claims on firms? The simple
answer is that the direct financial claims that firms offer, promise more
attractive rates of return than households could expect to earn from invest-
ing in real assets themselves. In short, they shift from real investment to the
purchase of financial claims because they expect it to be profitable to do so.
But this superficial answer ignores several important obstacles that must be
overcome in order to induce savers to give up real assets in exchange for
direct financial claims.
The fundamental problem is that, once savers no longer invest in real

assets directly, they must worry about the performance of those who act
as their agents and undertake the real investments to determine the returns
on their financial investments. Households are confronted with a prin-
cipal/agent problem in which they must deal with the possibility of hidden
actions and hidden information (Arrow, 1979). They must be concerned
about ‘adverse selection’ – the possibility that they may inadvertently invest
in incompetent firms with poor prospects instead of competent firms with
good productive opportunities. And they must be concerned with ‘moral
hazard’ – the possibility that firms may not honor their commitments once
they have received resources from investors. In order to protect against
adverse selection and moral hazard, households must spend resources in
deciding how to allocate savings. The activities involved include: (a) col-
lecting and analyzing information about firms; (b) negotiating a contract
that will limit the firm’s opportunities for taking advantage of the saver; (c)
monitoring the firm’s performance; and, if necessary, (d) enforcing the con-
tract. In the absence of strong accounting standards, good disclosure prac-
tices, strong legal protections for holders of direct claims and an efficient
judiciary and enforcement function, the information and transactions costs
may be so great that direct financing is not feasible.
In economies where the financial infrastructure – accounting and disclo-
sure practices, the legal framework and clearing and settlement arrange-
ments – is not sufficiently well developed to support arms-length direct
financial transactions, other non-market mechanisms for allocating savings
are likely to arise. Households may be linked together with firms through
family groups rather than in the marketplace.
Family ties may substitute for a strong financial infrastructure in two ways.
In the absence of strong accounting and disclosure practices, information is
likely to flow more readily within families than between unrelated parties.
Moreover, reputation within the family may substitute for information.
Thus, the adverse selection problem is likely to be mitigated for investment
in direct claims within the family group. Moreover, in the absence of strong
legal protections for creditors and minority shareholders, families have
enforcement mechanisms such as the threat of disinheritance, withholding of
affection, or expulsion from the family that may mitigate moral hazard.
In the absence of efficient capital markets, family groups may serve as a
quasi-financial system, pooling the savings of several related households to
finance a family-controlled firm in which the governance structure of the
family substitutes for capital market discipline. As the family enterprise suc-
ceeds, it will accumulate retained earnings that can be used to finance new
family enterprises. To some extent, the growth of family-controlled indus-
trial conglomerates in emerging economies can be viewed as an adaptation
to the absence of efficient capital markets. In several of the emerging

markets of Asia, more than 50 percent of publicly traded corporations are
family-controlled (Claessens et al., 1998a).
This mode of allocating capital has several potential disadvantages rela-
tive to that which would take place in a well-functioning capital market.
Firms are not confronted with the true opportunity cost of funds in the
economy, so investment may be too great or too small. Similarly, firms lose
the aggregation of information that takes place in a well-organized capital
market and may pursue inefficient investment projects far too long in the
absence of market discipline. Finally, the economy’s reliance on financial
flows within family groups raises high barriers to entry by unaffiliated firms,
which may have more attractive investment opportunities.5
As the family financial conglomerate grows in complexity, it is likely to
form an enterprise that will coordinate financial flows within the group.
This financial enterprise may also offer services to non-family members and
become a bank.
2.3 The Financial Sector with Financial Intermediaries
Banks and other financial intermediaries purchase direct financial claims

and issue their own liabilities; in essence, they transform direct claims into
indirect claims. The fundamental economic rationale for such institutions
is that they can intermediate more cheaply than the difference between what
the ultimate borrowers would pay and the ultimate saver would receive in a
direct transaction. Financial intermediaries enhance the efficiency of the
financial system if the indirect claim is more attractive to the ultimate saver
and/or if the ultimate borrower is able to sell a direct claim at a more
attractive price to the financial intermediary than to ultimate savers.
A comparison of the flow of funds matrix for an economy with only direct
financial claims (Table 5.4) with the flow of funds matrix for an economy
with both direct and indirect financial claims (Table 5.5) reveals a more
complex pattern of financing,6 characteristic of the financial deepening that
usually accompanies economic development (Goldsmith, 1965). The house-
hold sector has substituted much of its holdings of direct financial claims for
‘indirect financial claims’ on financial firms. Correspondingly, financial firms
hold most of the direct financial claims on non-financial firms. Also, the
household sector has a better opportunity to borrow from financial institu-
tions because the scale of borrowing by individual households seldom war-
rants the heavy fixed costs of issuing a direct financial claim.
But how can financial institutions link some savers and investors more
efficiently than direct market transactions between the household sector
and non-financial firms? Several factors may explain the relatively greater
Table 5.5 The flow of funds matrix for an economy with private placement
and financial institutions
Sectors Households Non- Financial Govern- Rest of Total

financial Institutions ment* World*
Firms
Flow of U S U S U S U S U S U S
Real
Income
Savings 145 12 5 0 162
Real Assets 12 148 2 162 0
Financial
Flows
Equity 10 34 28 4 38 38
Fixed 25 7 105 87 112 112
Income
Instruments
Indirect 105 3 108 108 108
Financial
Instruments
Financial
Instruments
Issued by
Foreign
Residents
Totals 152 152 151 151 117 117 418 418
Note: * Entries for these columns are developed later.
efficiency of financial intermediaries. First, financial intermediaries may be

able to collect and evaluate information regarding creditworthiness at lower
cost and with greater expertise than the household sector. And, when some
information regarding creditworthiness is confidential or proprietary, the
borrower may prefer to deal with a financial intermediary rather than dis-
close information to a rating agency or to a large number of individual
lenders in the market at large.
Second, transactions costs of negotiating, monitoring and enforcing a
financial contract may be lower for a financial intermediary than for the
household sector since there are likely to be economies of scale that can be
realized from investment in the fixed costs of maintaining a specialized staff
of loan monitors and legal and workout experts. In addition, by handling
other aspects of the borrower’s financial dealings, the financial intermedi-
ary may be in a better position to monitor changes in the borrower’s cred-
itworthiness.
Third, the financial intermediary can often transform a direct financial

claim with attributes that the borrower prefers into an indirect claim
with attributes that savers prefer. Borrowers typically need large amounts
for relatively long periods of time, while savers prefer to hold smaller-
denomination claims for shorter periods of time. By pooling the resources
of many savers, the financial intermediary may be able to accommodate the
preferences of both the borrower and savers.
Fourth, the financial intermediary often has a relative advantage in
reducing and hedging risk. By purchasing a number of direct claims on
different borrowers whose prospects are less than perfectly correlated, the
financial intermediary is able to reduce fluctuations in the value of the port-
folio of direct claims, given the expected return, relative to holdings of any
one of the direct claims with the same expected return. Diversification
reduces the financial intermediary’s net exposure to a variety of risks and
thus reduces the cost of hedging.
The upshot is that the introduction of bank deposits is likely to mobilize
additional savings that can be used to finance investment, since some
households will now substitute bank deposits for holdings of precious
metal, jewelry and other durable assets that are traditionally used as a store
of wealth. The increase in the pool of savings available to finance invest-
ment and the reduction in transactions costs in linking ultimate savers and
investors will lead to an increase in the quantity of investment. Improved
evaluation and monitoring of loans made possible by the specialization of
banks may lead to better screening and implementation of investment pro-
jects and thus improve the return on investment. These changes are
reflected in Table 5.5 where both household sector savings and real assets
have risen. Total household savings have risen from 130 units to 145 units
and retained earnings have risen from 10 units to 17 units.
Although the bank loans introduced in this section and the private place-
ments introduced in the preceding section are forms of debt, it is important
to note that they have strikingly different properties than marketable debt
securities. A ‘pure loan’ is a credit contract between a borrower and a single
lender. The contract is custom-tailored to meet the borrower’s financial
requirements and the lender’s need for assurances regarding the borrower’s
creditworthiness. Because the contract involves only one lender, it may be
renegotiated at relatively low cost should the borrower’s circumstances
change. Often the lender has specialized expertise regarding the business of
the borrower that enables the lender to monitor the borrower’s performance
at relatively low cost. The ‘pure loan’ is usually part of a relationship
between the borrower and lender in which the borrower may draw down and
repay loans over time, the lender monitors the activities of the borrower, and
the borrower may purchase other services from the lender. A pure loan is
likely to be an illiquid asset because, relative to a pure security of equal

maturity, only a small percentage of the full market value of the asset can
be realized if it is sold at short notice. The fundamental problem is that it is
difficult for a potential buyer to evaluate the credit standing of the debtor.
Moreover, the transactions costs of finding a counterparty and executing a
transaction are likely to be very high because the idiosyncratic features of a
‘pure loan’ preclude the development of dealer markets.
A ‘pure security’, in contrast, is a contract between the borrower and
many investors who may be unknown to the borrower and need have no
other relationship to the borrower. The investor need not have any special-
ized knowledge of the borrower’s business. Each investor is issued an iden-
tical type of claim on the borrower, which is readily transferable. A ‘pure
securities’ contract is much simpler than a loan agreement, containing
fewer covenants and contingent clauses, because after the security is issued,
it is impractical to renegotiate terms of the contract with the borrower; the
costs of coordinating collective action among a large number of (often
anonymous) investors are prohibitive.
A ‘pure security’ of a given maturity is likely to have a much more liquid
secondary market than a ‘pure loan’ of equal maturity. The issuance of
securities in primary markets is directed to many investors, all of whom
hold identical claims and none of whom is necessarily privy to information
about the borrower not available to the others. The standardization of
claims facilitates the development of dealer markets and leads to lower
transactions costs in selling securities. Since buyers in the secondary market
need not fear that sellers know more than they do about securities being
offered in the market, buyers can safely ignore the identity of the seller. In
contrast, loan contracts may be highly idiosyncratic, and the originating
lender may have information about the borrower, or specialized expertise
about the borrower’s business, not available to potential buyers. The loan
contract may also have contemplated some degree of monitoring by the
lender that the purchaser would be obliged to perform unless the loan were
serviced by the seller. These features severely limit the marketability of con-
ventional loans. Unless a buyer receives a full guarantee from the original
lender or some trusted third party, the buyer must make the same invest-
ment in information that the original lender made, and/or monitor the loan
agreement, perhaps without the expertise of the original lender.
2.4 The Government and International Sector
In order to complete the flow of funds matrix, we need to introduce two

additional sectors. First, the government sector affects the flow of funds in
two distinct ways. It issues direct claims to banks that serve as the reserve
base for the money supply. It also issues direct claims to finance its own
spending when desired government expenditures for purchases of goods
and services and the redistribution of income exceed current tax revenues.
Table 5.6 shows the flow of funds matrix that incorporates the government
sector. The government is shown with a deficit of 33 units that causes a cor-
responding reduction in net savings for the economy. Some economists argue
that current deficits lead to a one-for-one increase in household savings in
anticipation of higher future tax burdens (Barro, 1974). Other economists
regard this view as too extreme in light of the empirical evidence (Hausman
and Poterba, 1987). Table 5.6 depicts a case in which households make a
partial response to the government deficit: household savings rise from 145
units to 150 units. The government issues 45 units of financial liabilities to
fund its current and capital expenditures as well as its subsidies to favored
private sector borrowers. In our example, real sector investment declines in
spite of subsidies from the government to the private sector. Total real sector
Table 5.6 The flow of funds matrix for a closed economy with a
government sector

financial Institutions ment World*
Firms
Real
Income
Savings 150 12 5 33 33 167
Real Assets 7 118 2 7 134 0
Financial
Flows
Equity 10 34 28 4 38 38
Fixed 30 10 77 97 5 45 132 132
Income
Instruments
Indirect 113 5 118 118 118
Financial
Assets
Financial
Instruments
Issued by
Foreign
Residents
Totals 160 160 101 101 127 127 45 45 423 423
Note: * Entries for these columns are developed later.

assets decline from 162 units in Table 5.5 to 134 units in Table 5.6, indicative
of the ‘crowding out’ of private sector investment by government funding
demands.
Second, to complete the flow of funds, we add the international sector. As
national economies have become increasingly interdependent, cross-border
financial transactions of all kinds have become commonplace. Opening a
country to trade in financial assets offers advantages similar to those that we
observed in introducing financial instruments in the primitive economy.
World savings may be allocated more efficiently so that national income in
all countries is increased. International specialization on the basis of com-
parative advantage in financial services, like international specialization in
production, is likely to enhance efficiency. Competition from foreign insti-
tutions also stimulates innovations to cut costs and expand the range of
products. Moreover, the broader range of financial instruments available
enhances the scope for diversification to reduce country-specific risks.
Table 5.7 shows the complete flow of funds matrix. In this example the
national economy is running a current account deficit of 28 units. This
deficit is financed by net financial inflows that provide both debt and equity
Table 5.7 The flow of funds matrix for an open economy

financial Institutions ment World
Firms
Real
Income
Savings 155 33 28 33 183
Real Assets 7 134 2 7 150 0
Financial
Flows
Equity 13 41 28 5 5 46 46
Fixed 27 10 98 81 5 45 40 153 153
Income
Instruments
Indirect 116 5 136 15 136 136
Financial
Assets
Financial 2 30 32 32 32
Instruments
Issued by
Foreign
Residents
Totals 165 165 139 139 141 141 45 45 60 60 550 550
investment to the domestic economy and by drawing down some of the

domestic economy’s holdings of foreign assets. Household savings reflect
the benefits of opening the economy to the world capital market by increas-
ing to 155 units. The non-financial sector also benefits from the net inflow
of capital. Net domestic real investment increases from 134 in Table 5.6 to
150 in Table 5.7.
3 THE ROLE OF FINANCIAL INFRASTRUCTURE

AND EFFICIENT FINANCIAL MARKETS
The economy that we have sketched in the preceding section has a banking
system, but only a rudimentary capital market. The absence of an adequate
financial infrastructure meant that direct claims tended to be allocated
through extended families rather than through arm’s-length transactions in
the marketplace. Most corporate borrowing was in the form of bank loans.
The underdevelopment of capital markets in this economy limits risk-
pooling and risk-sharing opportunities for both households and firms. It
also robs the economy of a crucial source of information that helps coord-
inate decentralized decisions throughout the economy. Interest rates and
equity prices should be used by households in allocating income between
consumption and savings and in allocating their stock of wealth. And firms
should rely on financial markets for information about which investment
projects to select and how such projects should be financed (Merton, 1989).
Efficient financial markets help to allocate, transfer and deploy economic
resources across time and space in an uncertain environment (Merton,
1990). Without efficient financial markets, these functions are likely to be
performed less well and living standards will be lower than they might oth-
erwise have been.
The infrastructure to support a corporate bond market includes an
appropriate legal framework including reliable enforcement of bankruptcy
and foreclosure laws, strong accounting and disclosure standards, and
efficient and reliable clearing and settlement arrangements. It is also useful
to have a community of bond analysts and ratings agencies who can help
investors evaluate bonds. And, as we will emphasize in section 4, it is essen-
tial to develop a broad, deep, resilient secondary market.
In order for potential investors to be willing to accept a claim on future
cash flows for the repayment of principal and interest, they must be
confident that their rights to collect the promised debt payments are well
defined and enforceable. La Porta et al. (1998) have identified six measures
of creditor rights that are shown in Table 5.8a for countries in Table 5.1
along with a measure of contract enforceability. The measures focus on
creditors’ rights in the event of a default and include reorganization proced-

ures, priority rules and the scope for autonomous action by managers to
evade creditors. Based on data before the Asian Financial Crisis, on
average, the four industrialized countries score better on these indices of
creditor rights than do the eight Asian emerging economies.
La Porta et al. have also identified five indicators of the effectiveness of
the judiciary system since, in principle, strong enforcement by the courts
could compensate for weak laws. These measures (shown in Table 5.8b)
include proxies for the efficiency of the judicial system, and commitment to
the rule of law as well as indicators of the government’s attitude toward
business. Kane (2000a) also includes a measure of the quality of a country’s
bureaucracy since administrative efficiency may also affect the speed with
which rights are enforced. Again, on average, the four industrialized coun-
tries score better on these measures of the effectiveness of the judicial
system than do the eight Asian emerging economies.
In addition to assurances regarding the legal right to the promised cash
flows and the enforceability of such rights in the event of default, a potential
investor will need to form an estimate of the probability of default and the
expected recovery in the event of default. This depends on the availability of
reliable and relevant data about the firm’s current condition and prospects as
well as the availability of expert advice. La Porta et al. (1998) have identified
an index of accounting standards, which is reported in Table 5.8c. In add-
ition, Kane (2000b) has identified an index of restrictions on the press as an
indication of the openness of the society and the scope for manipulating
flows of information. Again, on average, the four industrialized countries
have much better scores than the eight Asian emerging economies, although
the average masks wide variations across the eight countries.
Finally, a potential investor must have confidence in arrangements for
the clearing and settlement of bond trades. Creditor rights, judicial
efficiency and good information will be of little use if the investor cannot
be certain of receiving the bond when payment is made. Ideally, the clear-
ing and settlement system should offer delivery against payment. Many of
the emerging markets in Asia are adopting such systems.
Generally, countries that rate higher on indices of creditor rights, judi-
cial efficiency, and quality of information have larger bond markets. As we
will see in section 4, there are many useful things that a government can do
to nurture development of a strong bond market, but these indices measure
issues of fundamental importance. Indeed, Kane (2000a) has suggested
that the official international financial institutions should help countries
improve their rankings on such indices and that the managers of these
official international financial institutions should be evaluated and com-
pensated on the basis of their success in encouraging such improvements.
Table 5.8a, b and c Indicators of the quality of the financial infrastructure
5.8a Creditor rights
Contract No Automatic Secured Restrictions Management Creditor Legal Reserves
Enforceability Stay on Creditors on Autonomous Does Not Stay in Rights Required to
Secured Assets Paid First Reorganization Reorganization Continue Operation
Hong Kong N/A 1 1 1 1 4 0
Indonesia 1.76 1 1 1 1 4 0
South Korea 2.19 1 1 0 1 3 0.5
Malaysia 2.26 1 1 1 1 4 0
Philippines 1.75 0 0 0 0 0 0
Singapore 3.22 1 1 1 1 4 0
Taiwan N/A 1 1 0 0 2 1
164
Thailand 2.23 1 1 0 1 3 0.1

Average 2.24 0.88 0.88 0.5 0.75 3 0.2
Australia 3.04 0 1 0 0 1 0
Japan 3.16 0 1 0 1 2 0.25
UK 3.43 1 1 1 1 4 0
US 3.55 0 1 0 0 1 0
Average 3.30 0.25 1 0.25 0.5 2 0.06
Notes and sources:

Contract enforceability: Measures the ‘relative degree to which contractual agreements are honored and complications presented by language and
mentality differences’. Scored 0–4, with higher scores for superior quality. Source: Business Environmental Risk Intelligence and Kane (2000b).
No automatic stay on secured assets: Equals 1 if the reorganization procedure does not impose an automatic stay on the assets of the firm upon
filing the reorganization petition. Automatic stay prevents secured creditors from gaining possession of their security. It equals 0 if such restriction
does exist in the law. Source: Bankruptcy and Reorganization Laws and LaPorta et al. (1998).
Table 5.8a, b and c (continued)
Secured creditors paid first: Equals 1 if secured creditors are ranked first in the distribution of the proceeds that result from the disposition of the
assets of a bankrupt firm. Equals 0 in non-secured creditors, such as the government and workers are given absolute priority. Source: Bankruptcy
and Reorganization Laws and LaPorta et al. (1998).
Restrictions on autonomous reorganization: Equals 1 if the reorganization procedure imposes restrictions such as creditors’ consent to file for
reorganization. It equals 0 if there are no such restrictions. Source: Bankruptcy and Reorganization Laws and LaPorta et al. (1998).
Management does not stay in reorganization: Equals 1 when an official appointed by the court, or by the creditors, is responsible for the
operation of the business during reorganization. Equivalently, this variable equals 1 if the debtor does not keep the administration of its property
165
pending the resolution of the reorganization process, and 0 otherwise. Source: Bankruptcy and Reorganization Laws and LaPorta et al. (1998).
Creditor rights: An index aggregating different creditor rights. The index is formed by adding 1 when: (1) the country imposes restrictions such as
creditors’ consent or minimum dividends to file for reorganization; (2) secured creditors are able to gain possession of their security once the
reorganization petition has been approved (no automatic stay); (3) secured creditors are ranked first in the distribution of the proceeds that result
from the disposition of the assets of a bankrupt firm; and (4) the debtor does not retain the administration of its property pending the resolution of
the reorganization. The index ranges from 0 to 4. Source: Bankruptcy and Reorganization Laws and LaPorta et al. (1998).
Legal reserves required to continue operation: It is the minimum percentage of total share capital mandated by Corporate Law to avoid the
dissolution of an existing firm. It takes a value of 0 for countries without such restrictions. Source: Company Law or Commercial Code and
LaPorta et al. (1998).

5.8b Effectiveness of judicial system
Efficiency Rule Corruption Bureaucratic Risk of Risk of
of Judicial of Quality Expropriation Contract
System Law Repudiation
Hong Kong 10.0 8.22 8.52 4.14 8.29 8.82

Indonesia 2.50 3.98 2.15 1.50 7.16 6.09
South Korea 6.00 5.35 5.30 4.18 8.31 8.59
Malaysia 9.00 6.78 7.38 3.54 7.95 7.43
Philippines 4.75 2.73 2.92 1.46 5.22 4.80
Singapore 10.00 8.57 8.22 5.11 9.30 8.86
Taiwan 6.75 8.52 6.85 N/A 9.12 9.16
Thailand 3.25 6.25 5.18 4.39 7.42 7.57
Average 6.53 6.3 5.82 3.47 7.85 7.67
Australia 10.00 10.00 8.52 6.00 9.27 8.71
Japan 10.00 8.98 8.52 5.89 9.67 9.69
UK 10.00 8.57 9.10 6.00 9.71 9.63
US 10.00 10.00 8.63 6.00 9.98 9.00
Average 10 9.39 8.69 5.97 9.66 9.26
Notes and sources:

Efficiency of judicial system: Assessment of the ‘efficiency and integrity of the legal
environment as it affects business, particularly foreign firms’ produced by the country risk-
taking agency Corporation for International Business. It ‘may be taken to represent
investors’ assessments of conditions in the country in question’. Average between 1980–83.
Scale from 0 to 10, with lower scores for low efficiency levels. Source: LaPorta et al. (1998).
Rule of law: Assessment of the law and order tradition in the country produced by the
country-risk rating agency International Country Risk (ICR). Average of the months of
April and October of the monthly index between 1982 and 1995. Scale from 0 to 10, with
lower scores for less tradition for law and order. Source: International Country Risk Guide
(ICRG) and LaPorta et al. (1998).
Corruption: ICR’s assessment of corruption in government. Lower scores indicate ‘high
government officials are likely to demand special payments’ and ‘illegal payments are
generally expected throughout the lower levels of government’ in the form of ‘bribes
connected with import and export licenses, exchange controls, tax, assessment, policy
protection, or loans’. Scale runs from 0 to 6, with lower scores indicating higher levels of
corruption. Source: ICRG and LaPorta et al. (1998).
Bureaucratic quality: Average of ‘bureaucratic quality’ assessment values assigned by
ICRG between 1982–1995. Scored 0–6, with higher scores for superior quality. Source:
Kane, (2000b).
Risk of expropriation: ICR’s assessment of the risk of ‘outright confiscation’ or ‘forced
nationalization’. Average of the months of April and October of the monthly index between
1982 and 1995. Scale from 0 to 10, with lower scores for higher risks. Source: LaPorta
et al. (1998).
Risk of contract repudiation: ICR’s assessment of the ‘risk of modification in a contract
taking the form of a repudiation postponement, or scaling down’ due to ‘budget cutbacks,
indigenization pressure, a change in government, or a change in government economic and
social priorities’. Average of the month of April and October of the monthly index between
1982 and 1995. Scale from 0 to 10, with lower scores for higher risks. Source: LaPorta et al.
(1998).

5.8c Quality of economic information
Accounting Standards Index of Restrictions
on the Press
Hong Kong 73 32.75
Indonesia N/A 71.40
South Korea 68 26.40
Malaysia 79 61.00
Philippines 64 44.60
Singapore 79 63.60
Taiwan 58 28.40
Thailand 66 39.80
Average 69.6 45.99
Australia 80 8.80
Japan 71 20.20
UK 85 22.20
US 76 12.80
Average 78 16
Notes and sources:

Accounting standards: Index created by examining and rating companies’ 1993 annual
reports on their inclusion or omission of 85 items. These items fall into seven categories
(general information, income statement, balance sheet, funds flow statement, accounting
policies, stockholders’ information and supplementary information). A minimum of five
companies in each country was studied. There are 1000 industrial companies from 41
countries. The companies represent a cross-section of various industry groups. Scores are
from 0–100. Higher scores indicate better accounting standards. Source: Center for
International Financial Analysis and Research Inc.(1995), International Accounting &
Auditing Trends, 4th edn.
Index of restrictions on the press: Assessment of repressive actions and laws, regulations,
controls and political pressures that influence media content. Score reported is the average
index assigned by Freedom House staff Annual Press Freedom Reports, 1994–98. Scale runs
from 0 to 100, with lower scores indicating greater freedom. Source: Kane (2000b).
3.1 Why Equity Markets May Exist Where Bond Markets Fail to Thrive
What are the main obstacles to developing an efficient bond market? Why,
in environments with weak financial infrastructures, which discourage all
external finance to some extent, do equity markets appear to flourish rela-
tive to bond markets? Part of the answer is inherent in the difference
between debt and equity contracts. Debt claims promise repayment of prin-
cipal and interest, while equity claims promise payment of a pro rata share
of profits and usually convey a proportionate vote in important corporate
governance matters.
The maximum return on a bond7 purchased at par value is the promised

interest payments. But the downside may include loss of the principal
amount as well as promised interest payments. The bond contract defines
the obligations of the borrower and remedies in the event of default.
Usually, the key remedy in the event of default is that the bondholders may
seize collateral or control of the enterprise from its owners. This remedy
may be of little value, however, if bankruptcy and foreclosure laws are weak
and enforcement is ineffectual.
The main challenge in pricing a bond is setting an interest rate that will
compensate for the opportunity cost of funds, default, purchasing power and
liquidity risk as well as whatever idiosyncratic features the bond may have
such as a call option or sinking fund. In the absence of an active secondary
market in risk-free debt of a comparable maturity, it will be difficult to iden-
tify the appropriate opportunity cost of funds. Estimating the probability of
default and the expected recovery from the liquidation or sale of the firm in
the event of default will also prove difficult in an economy with a weak
financial infrastructure where the ability to enforce a collateral agreement is
questionable. In the absence of credible accounting practices, good disclo-
sure practices or reliable bond ratings, it may be very difficult to estimate a
probability of default or expected loss in the event of default. This challenge
is still more difficult in the absence of clear laws setting out the bondholder’s
rights in the event of default, or an efficient judiciary that will oversee
enforcement of such rights and a reliable enforcement mechanism. If house-
holds are concerned about a high probability of default or the expected loss
in the event of default, it may not be possible to establish a viable bond
market. Borrowers may not be able to offer credibly a sufficiently high inter-
est rate to compensate for the perceived risk of loss (Stiglitz and Weiss, 1981).
In contrast to a bond in which the upside is limited by the promised inter-
est rate, an equity claim has an unlimited upside return, which can compen-
sate for the perceived riskiness of the claim. Although minority shareholders
will experience the same frustrations as bondholders in evaluating a firm’s
current condition and its earnings prospects, they can take comfort in the
fact that they share an interest with the controlling shareholders (who are
often also the managers in emerging market firms) in a rising share price.8
Thus, if there is an active secondary market and reliable clearing and settle-
ment procedures for buying and selling equity claims, an active market may
develop for a firm’s equity even though investors would not be willing to buy
its debt.
What are the consequences of operating a financial system with a
banking sector and equity market, but no bond market? The implications
are profound and far-ranging. We will analyze the impact on other markets,
savers, investors, banks and financial development more broadly.
3.2 Absence of Bond Markets: Implications for Other Markets
In the absence of a bond market, the economy will lack a market-determined

term structure of interest rates that accurately reflects the opportunity cost
of funds at each maturity. Without a term structure of interest rates, it will
be difficult to develop efficient derivatives markets that enable economic
agents to manage financial risks. Forward markets trade forward contracts
that obligate the owner to buy a given asset on a specific date at a price
specified at the origination of the contract. Since market participants always
have the option of buying the asset on the spot market and holding it until
the maturity of the forward contract, the forward price is linked to the
current price by the interest cost of holding the asset until the forward con-
tract matures. In the case of forward foreign exchange contracts, the rela-
tionship is a bit more complex because it involves both foreign and domestic
interest rates. The forward foreign exchange rate is related to the spot
foreign exchange rate by the ratio of (1 +) the home country interest rate rel-
ative to (1 +) the foreign interest rate, both corresponding to the maturity of
the forward contract. If there is no market-determined domestic interest rate,
it may still be possible to buy a forward contract, but the market will be very
thin and transactions costs will be heavy because market-makers will not be
able to hedge their positions using the bond market.
Futures markets trade futures contracts that obligate the owner to pur-
chase a specified asset at a specified exercise price on the contract maturity
date. Futures markets differ from forward markets in that changes in the
value of a futures contract are settled day by day as they occur rather than
at the maturity of the contract. Thus Black (1976) has described futures
contracts as a series of forward contracts that are settled day by day. Again,
however, a key link between spot and futures prices is the interest rate cor-
responding to the maturity of the contract. Futures contracts are exchange-
traded instruments that require a significant volume of trading to warrant
the substantial fixed costs of organizing and running an exchange.
Countries without a bond market are unlikely to generate enough activity
to support development of an active futures exchange. Although Hong
Kong, Singapore, Malaysia, the Philippines, and China all have futures
exchanges, active trading has been confined mainly to Singapore’s
International Monetary Exchange (SIMEX).
Swap contracts obligate two parties to exchange, or swap, some specified
cash flows at specific intervals. The most common form is an interest rate
swap in which cash flows are determined by the two different interest rates
specified in the swap agreement. But, the swap contract can be decomposed
into a portfolio of forward contracts (Smith et al.,1986) in which, at each
settlement date throughout the term of the swap contract, part of the
change in value is transferred between the counterparties. In contrast to

the forward contract, in which the change in value is transferred between
the counterparties at the maturity of the contract or the futures contract,
in which the change in value is transferred between the counterparties day
by day over the term of the contract, part of the value of the change in a
swap contract is conveyed between the counterparties at each settlement
date specified in the swap contract. Again, the key link between spot and
forward rates is the corresponding interest rate.
In contrast to the owners of forward, futures or swap contracts who
have an obligation to perform as specified in the contract, the owner of an
option contract has the right, but not the obligation to perform as specified
in the contract. Just as futures and swaps can be viewed as a portfolio of
forward contracts, options can be viewed as portfolios of forward contracts
and risk-free bonds. Black and Scholes (1973) have shown that a dynamic
portfolio of forward contracts on the underlying asset and riskless bonds
can replicate a call option. As the price of the asset rises, the call-option-
equivalent portfolio contains an increasing proportion of forward contracts
on the asset. As the price of the asset falls, the replicating portfolio contains
a decreasing proportion of forward contracts on the asset. Like the forward,
futures and swap markets, the options market depends critically on the bond
market for pricing and hedging positions.
In the absence of a well-functioning bond market, it may be possible to
obtain forward, swaps and options contracts that are specially tailored for
a client. But they will be very expensive relative to what they would cost in
an economy with a well-functioning bond market because they cannot be
hedged as efficiently. The consequence is that market participants will be
exposed to more financial risk than they would choose to accept if they had
access to well-functioning derivatives markets. The events of 1997 showed
that many market participants had accepted excessive exposures to foreign
exchange risk.
The absence of a risk-free9 term structure of interest rates also makes it
difficult to price credit risk by comparing a risky asset with a risk-free asset
that is alike in all other characteristics. Although inefficiencies caused by
mispricing credit risk may be second order relative to inefficiencies that
result from mispricing the risk-free rate, they nonetheless cause distortions
in the economy. Without a government bond market that establishes bench-
mark risk-free rates at critical maturities, it will be very difficult to establish
a corporate bond market, much less a market for high-yield debt or secur-
itized assets. Partly, this is a consequence of the microeconomics of market-
making. It is easier to start a new market if the activity must cover only the
marginal cost of the new market rather than the full costs of setting up and
maintaining a market. If institutions have already invested in trading in
government bonds, then the marginal cost of introducing another fixed

income market will be relatively slight.
The absence of a broad, deep, resilient bond market also causes a loss of
information to society. The differential yield between a risky bond and a
risk-free bond that is alike in all other characteristics, reflects a market con-
sensus about the appropriate credit risk (and possibly liquidity risk)
premium. This market information can be used to price comparable bank
loans and it is likely to extend the range of credit risk that can be priced in
the market rather than quantity rationed.
Although, as noted above, an equity market may flourish in the absence
of the bond market, it may not be very efficient in the sense of aligning
prices with fundamental economic values. Ideally, share prices should
reflect the present discounted value of expected future earnings. But in
economies that lack the infrastructure to support a bond market, investors
are likely to have considerable doubt about what past earnings have been
and what current earnings are, much less what expected future earnings will
be. Moreover, in the absence of a bond market it is not clear how the appro-
priate discount rate should be determined. Thus shares are likely to be
priced on the basis of expectations that are often shallowly held and subject
to considerable volatility and the usefulness of share prices in allocating
resources and corporate governance is correspondingly vitiated.
3.3 Absence of Bond Markets: Implications for Savers
Without a well-functioning bond market, savers face a diminished array of

assets. They will hold more substitute assets such as bank deposits and pos-
sibly, but less likely, equity and probably more non-financial assets such as
gold or jewelry that reduce the supply of savings that can be mobilized for
productive investment. They will be forced to accept a lower return for any
given level of risk or a higher level of risk for a given level of return relative
to an economy with a well-functioning bond market. As Hakansson (1999)
has argued with respect to the corporate bond market, ‘we can expect that a
large number of these securities will be such that we will be unable to find
any portfolio of other securities in the market which can replicate their pay-
off patterns across contingencies or states’. Applying his earlier theoretical
work Hakansson (1982, 1992) compared equilibria with and without a well-
developed bond market, and concluded that under fairly general conditions
‘the financial market richer in bonds will constitute a Pareto-improvement
over the financial market in which banks do most of the lending’.
In the absence of a well-functioning bond market, specialized financial
institutions with long-term liabilities such as life insurance companies and
pension funds will find it more difficult to acquire long-term assets that
match the maturity of their liabilities. And consequently, the insurance they
provide against future contingencies will be more costly.
3.4 Absence of Bond Markets: Implications for Investors
Without a well-functioning bond market, firms will lack a clear measure of

the opportunity cost of funds. From society’s perspective this may lead to
overinvestment if the firm’s internal rate is too low or underinvestment, if
the firm’s internal rate is too high. Evidence from the mid-1990s in several
dynamic Asian economies suggests that the internal discount rate may have
often been too low because returns on investment fell markedly.
Firms will be entirely reliant on banks for debt financing. The same
weaknesses in the financial infrastructure that impede development of a
bond market – inadequate accounting and auditing, weak disclosure laws
and uncertain enforcement of contracts – also lead banks to prefer short-
term credit. As Diamond (1991) has shown, short-term credit is an import-
ant way to control borrowers when there are hidden action and hidden
information problems since it limits the time an opportunistic firm can
exploit its creditors without being in default.
Since banks typically lend for periods much shorter than the maturity of
long-term bonds, this may affect the firm’s preferred leverage. Any given
leverage structure will be riskier the shorter the maturity of the debt out-
standing. Firms may attempt to compensate for this risk by attempting to
control the bank lender. As noted earlier, corporate conglomerates will
attempt to affiliate with a bank in part to form an internal capital market
that will substitute for the absence of an external capital market. If the sub-
sequent loans should go bad, this sort of relationship can give rise to the
charge of crony capitalism.
Another consequence of the reliance on short-term bank lending may be a
bias in firms’ investment decisions. Based on Hart and Moore (1995), Caprio
and Demirgüç-Kunt, (1997) have argued that firms will tend to match the
maturity of their assets and liabilities. This tendency has been documented in
the United States and Caprio and Demirgüç-Kunt (1997) report on World
Bank studies that affirm the pattern of matching maturities of assets and lia-
bilities holds for developing countries as well. This suggests that reliance on
short-term bank lending will bias investment toward short-term assets. As a
result, there may be too little investment in longer-term assets such as infra-
structure, public utilities, housing and capital-intensive industries.
Exclusive reliance on bank lending may bias investment in another more
subtle way. Access to the bond market may play a role in encouraging
entrepreneurial ventures by limiting the ability of banks to extract rents
from successful ventures. Black and Gilson (1998) have argued that a
dynamic venture capital sector will not thrive in a bank-based financial

system because successful venture capitalists need the option of exiting
from the project through issue of bonds or equity.
Some recent evidence for the bank-centered Japanese system (Weinstein
and Yafeh, 1998) suggests banks do extract rents from their dependent cor-
porate customers. The consequence is that the effective cost of funds is
higher than it would have been if the firm had access to a well-functioning
bond market.
The largest, best-known firms may attempt to compensate for the lack of a
domestic bond market by issuing bonds in the international market. Table 5.9
shows corporate issues outstanding as a percentage of GDP in 1998 for
several Asian economies and a benchmark group of high-income economies.
In general, reliance on international issuance of bonds was higher in this
group of Asian economies than in the benchmark group (apart from the
United Kingdom, which may be a special case because it is the locus of the
Table 5.9 Corporate borrowings in domestic and international markets

(year-end, 1998)
Domestic Corporate International Corporate

Debt Securities Debt Securities
Outstanding (% GDP) Outstanding (% GDP)
Hong Kong 1.2 8.9
Indonesia N/A 13.5
South Korea 32.2 5.7
Malaysia 33.7 15.3
Philippines 0.0 9.4
Singapore 2.5 4.7
Taiwan 1.5 2.5
Thailand 3.8* 5.2
Average 10.7 8.2
Australia 15.9 4.7
Japan 17.9 4.2
UK 8.2 9.1
US 27.4 3.2
Average 17.4 5.3
Note: * Includes data for financial institutions.
Source: IMF International Financial Statistics, IMF World Economic Outlook Database,
Bank for International Settlements.
largest concentration of international bond market activity). While this

behavior is easily explained as an accommodation to the inadequacies of
domestic bond markets, it subjects the borrowers to a heavy potential cost.
In general, issues on international bond markets are denominated in
foreign currency, usually US dollars or euros. Thus, to the extent that bor-
rowing firms use this source of financing to fund activities that will have
returns in the domestic currency, they will be increasing their exposure to
foreign exchange risk. As already noted, opportunities for hedging this risk
in derivatives markets are limited and generally quite expensive.
3.5 Absence of Bond Markets: Implications for Banks
Without competition from the bond market, the banking sector will be
larger than it would otherwise be. Banks will have more deposits at lower
cost because their customers will have very few other alternative fixed-
income investments and they will have more corporate loans because their
borrowers will have few other sources of debt financing. If the banking
market were highly competitive, the distortions from bank dominance of
debt finance might be relatively slight, but in most countries without a bond
market the banking system is highly concentrated. The deposit rate is not
likely to reflect the true opportunity cost of funds for the economy because
of cartel pricing in some countries and because in most countries, banks
benefit from access to an implicit, if not an explicit safety net. The percep-
tion that claims on the bank will receive some degree of protection from the
government, means that depositors will not be an effective source of discip-
line on bank risk-taking.
It is generally argued that bank monitoring of a borrower is superior to
monitoring by bondholders because bank lenders have lower costs of col-
lective action and can renegotiate a loan contract at lower cost in the event
that the borrower cannot meet the original repayment schedule. This may
be true in general, but recent experience has shown that if a bank is weakly
capitalized so that it cannot take a write-down in a loan renegotiation
without violating capital adequacy standards, the bank may let the bor-
rower continue negative present value projects by funding these activities to
avoid declaration of default (Herring, 1989). In this circumstance, mon-
itoring by bondholders may be preferable since they will have no motive to
sustain uneconomic activity.
The absence of a bond market precludes banks from issuing bonds,
which might reduce their exposure to liquidity risk and provide another
source of market discipline.10 The virtual absence of market discipline from
debt markets places a heavier burden on bank supervisors to curb risk-
taking. Like their counterparts in the industrialized world, however, bank
supervisors in emerging markets have seldom been up to the challenge.

Thus, the main restraint must come from shareholders of the bank. But in
a world of implicit deposit guarantees, they have an incentive to take
greater risks. This tendency is exacerbated if the bank is controlled by inter-
ests who are also heavy borrowers from the bank.
Even without this distortion of the incentives for risk-taking, a bank that
operates in an economy without bond markets has a diminished capacity
to manage risks. The thinness of derivatives markets means that most
hedging activities must involve transactions on the balance sheet. It will be
particularly challenging to deal with concentrations of credit risk, since in
the absence of a well-developed bond market, it will be difficult to sell or
securitize loans or negotiate credit derivatives. And without access to a
liquid bond market, banks will be more vulnerable to a liquidity shock
because they will not have the option of selling bonds in a liquid secondary
market and thus are more likely to be obliged to accept fire-sale losses on
the liquidation of bank loans.
Viewed from a broader perspective, the economy is at risk of crisis due
to excessive reliance on bank lending. Because banks are highly leveraged
institutions, the economy is much more vulnerable to a financial crisis than
if more corporate borrowing had taken place in the bond market and the
claims were held in well-diversified portfolios. In the event of a shock that
cripples the banking system, there will be an enormous impact on eco-
nomic activity because borrowers will not be able to substitute issuance of
bonds for bank borrowing. Instability in the banking system can halt
investment projects and reduce aggregate demand. Economic activity may
be depressed until the banking system can be recapitalized. As experience
in Asia since 1997 has shown, this can be a very painful process.
The absence of bond markets also inhibits efforts to clean up bank
balance sheets in the wake of a crisis. From Scandinavia to the United
States, Japan and several emerging economies in Asia, governments have
issued debt in exchange for non-performing loans. In the absence of well-
organized bond markets, the government debt issued is less liquid and
therefore less useful in resuscitating bank lending. More importantly, in the
absence of an active fixed-income market, it is more difficult to securitize
non-performing loans so that resources can be redeployed as rapidly as pos-
sible to restructure the economy.
3.6 Absence of Bond Markets: Summary
An economy that relies exclusively on banks for debt financing faces several
major costs. First is the loss of information that is contained in market-
determined interest rates. This impedes the development of derivatives
markets and may lead to inefficiencies in the pricing of equities. Without a

clear measure of the opportunity cost of capital firms may invest too little
or too much and the allocation of capital will be less efficient than if the
economy had the advantage of a well-functioning bond market.
Second is the loss of welfare to savers who are less well off than they
would be with the option of investing in a well-functioning bond market.
Because financial investment is less attractive than it would otherwise be,
fewer savings may be mobilized in the financial system to fund investment.
Third, firms may face a higher effective cost of funds than if they had
access to the bond market and their investment policies may be biased in
favor of short-term assets and away from entrepreneurial ventures. If firms
attempt to compensate for the lack of a domestic bond market by borrow-
ing in international bond markets, they may be obliged to accept excessive
exposure to foreign exchange risk. In any event, the underdevelopment of
domestic derivatives market will make it more difficult to manage financial
risks.
Fourth, the banking sector will be larger than it would otherwise be.
Since banks are highly leveraged, this may render the economy more vul-
nerable to crisis. Certainly, in the event that a banking crisis occurs, the
damage to the real economy will be much greater than if investors had
access to a well-functioning bond market and the financial restructuring
process will be more difficult.
If the economy would be better off with a well-functioning bond market,
what can the government do to nurture it? What policies will facilitate
development of a bond market? We turn to that topic in the next section.
4 THE ROLE OF GOVERNMENT AS ISSUER
The first major bond market to develop is usually the market in government
obligations. In many countries, the government has the largest stock of
issues outstanding. In general, it is easier for bond traders to price govern-
ment issues where credit risk is not an important consideration.
Government bond prices can then serve as a basis for pricing the issues of
other borrowers who are subject to credit risk.
In most countries, governments issue debt to fund the gap between tax
receipts and current expenditures, and sometimes to finance some extraor-
dinary current expenditure. (See Table 5.10 that shows government bor-
rowing and government borrowing relative to borrowing by other issuers in
the eight Asian emerging economies and the four industrialized countries.)
The US bond market took flight after the issuance of Liberty Bonds to
finance US participation in World War I. Rajan and Zingales (1999) note
Table 5.10 Public and total borrowings in domestic and international

markets (year-end, 1998)
Domestic Debt Securities International Debt

Outstanding (% GDP) Securities
Outstanding (% GDP)
Public All issuers Public All issuers
Hong Kong 3.3 17.4 4.7 19.5
Indonesia N/A 1.5 0.7 18.2
South Korea 16.2 75.7 7.2 16.8
Malaysia 31.3 85.4 1.4 17.5
Philippines 32.3 32.3 3.7 16.6
Singapore 20.8 23.3 0.1 6.5
Taiwan 11.7 13.2 0.0 2.8
Thailand 16.5 20.3 2.0 12.7
Australia 25.2 68.3 8.0 25.0
Germany 40.3 93.3 0.6 23.4
Japan 97.2 136.9 0.7 8.3
UK 33.1 60.8 0.9 25.8
US 88.8 159.5 1.5 9.6
Source: IMF International Financial Statistics, IMF World Economic Outlook Database,
Bank for International Settlements.
that people who would otherwise not buy a financial security, bought these
bonds for patriotic reasons. The favorable experience investors had with
these bonds left them willing to invest in securities issued by corporations.
This gave liquidity to the corporate securities market and made possible the
significant expansion of these markets during the 1920s.
Does this mean that fiscally conservative governments that do not run
deficits cannot nurture a robust bond market? Hong Kong has shown that
this need not be true. After all, it is gross debt that matters for the develop-
ment of the market, not the net debtor position of the government. Hong
Kong developed a benchmark yield curve in Hong Kong dollars through
issues of exchange fund bills and notes, the proceeds of which are used pri-
marily to invest in international markets, not to fund government spending.
If the government’s objective is the nurture of a robust bond market,
then it should aim at establishing a benchmark yield curve that can serve as
the risk-free rate for the pricing of other securities. This means committing
to a program of regular issues at the appropriate maturities – usually three
months, six months, one year, three years, five years and ultimately ten
years. It must be recognized at the outset that the goal of developing a
robust bond market may conflict with the goal of minimizing the cost of
government borrowing.11
The design of government securities should be as simple as possible
without complicated covenants and the design should be consistent across
the maturities that comprise the benchmark yield curve. This will facilitate
pricing of the risk-free rate without the complication of special features
such as sinking funds, call options or other features.
It is crucial that the interest rate on government bonds be market-
determined, not administratively determined. If the government attempts
to manipulate the bond market to reduce the cost of government borrow-
ing, important information will be lost, which may lead to distortions in the
allocation of capital. This means that the government should not require
certain institutions to hold its debt or devise special tax treatment of gov-
ernment debt that differs from that for other securities. Here again there is
a natural tension between the objectives of nurturing the development of a
robust bond market and minimizing the cost of government borrowing.
Generally, the price discovery process is enhanced by combining com-
petitive auctions of new issues with issuance through a set of primary
dealers who act as underwriters. It is useful to invite foreign firms to become
primary dealers on the same basis as domestic firms. This is likely to speed
the adoption of world-class best practices in the local bond market and
enhance the access of domestic borrowers to longer-term foreign sources
of funds. Primary dealers should be required to make markets in the issues
by continuously quoting a bid-asked spread and standing ready to buy or
sell at the stated rates.
Although the government will find a natural constituency for its longer-
term issues in the portfolios of institutions with longer-term liabilities, such
placements will not facilitate the development of a liquid secondary market
because these institutions are likely to buy and hold bonds until they
mature. Thus, it is important to attract other investors who will have a
trading orientation. Mutual funds, for example, should be encouraged to
enter the market.
4.1 Nurturing a Strong Secondary Market
The liquidity of an asset is enhanced if it is traded in a liquid secondary

market. Even if the asset is not sold, the liquidity of the secondary market
increases its value as collateral for a loan because its worth can be more
easily verified. Liquid secondary markets also raise the value of primary
securities.12 Confidence in the liquidity of secondary markets provides a
valuable option to investors. Even if the investor does not plan to sell the
primary claim before maturity, the investor’s future portfolio allocation
preferences are inevitably subject to uncertainty and thus the availability of

a deep, broad, secondary market enhances the investor’s willingness to buy
the initial, primary claim.
Empirical evidence suggests that this option may be very valuable
indeed. Pratt (1989) reports comparisons of the value of letter stocks that
are identical in all respects to the freely traded stock of public companies
except that they are restricted from trading on the open market for a
specified period.13 Pratt (1989, p. 241) concludes that ‘compared to their
free-trading counterparts, the discounts on the letter stocks were the least
for NYSE-listed stocks, and increased in order for AMEX-listed stock,
OTC reporting companies, and OTC nonreporting companies’. This
ranking of discounts corresponds roughly to perceptions of the liquidity of
these secondary markets. Using the midpoints of the discount range for
letter stocks relative to their freely traded counterparts, Pratt found that the
discount was 25.8 percent.14
The ‘liquidity of a secondary market’ is usually described in terms of its
depth and breadth. ‘Depth’ connotes the quantity that can be sold without
moving prices against the seller. ‘Breadth’ connotes the diversity of partici-
pants and the heterogeneity of their responses to new information. Both
qualities are usually positively correlated with the size of the secondary
market. Deep, broad markets are generally more resilient against disturb-
ances of any given size than thin, narrow markets; they tend to display
greater price stability in response to a shock of a given magnitude.
Liquid secondary markets are also ‘transactionally efficient’ in the sense
that the cost of a round-trip (the bid-asked spread) is low (Guttentag and
Herring, 1986). Dealer markets are usually regarded as especially trans-
actionally efficient because in addition to providing information and
matching buyers and sellers, dealers also provide immediacy by buying and
selling from inventory. The bid-asked spread charged by dealers in sec-
ondary markets must cover the opportunity cost of maintaining an inven-
tory of securities, operating costs and the risk of holding an inventory of
securities. Greater price stability, which is associated with deep, broad
markets, reduces the risk of inventorying securities and thus reduces trans-
actions costs.
A government can track its progress in fostering a liquid secondary
market by tracking the spreads quoted by dealers. The smaller the spread
and the larger the size of the transaction that dealers are willing to under-
take at the quoted spread, the more liquid the secondary market.
The liquidity of an asset also depends on the reliability of arrangements
for exchanging the asset for cash. Heightened perception of ‘settlement
risk’ – the risk that one party in a transaction will fulfill its settlement oblig-
ation while the counterparty does not – can undermine the liquidity of an
asset. In these respects the liquidity of an asset depends on the liquidity of

its secondary market. In this instance emerging markets may have an
advantage over some well-established markets with legacy clearing and set-
tlement systems. They have the opportunity to leapfrog traditional arrange-
ment by adopting modern technology to facilitate clearing and settlement
of secondary market trading. Hong Kong, for example, has established a
computerized book-entry system for bonds to reduce clearing and settle-
ment risk. This book-entry system is linked to a real time gross settlement
payment system so that it can provide real time delivery against payment
for Hong Kong dollar debt securities.
While there are many measures a government can implement to enhance
the liquidity of its secondary markets, the scope for success is inherently
constrained by the size of the economy. Most European economies have
not been of sufficient size to foster broad, deep, resilient bond markets like
those found in the United States. Early experience within the euro area,
however, indicates that the combined bond market denominated in euros
may indeed grow to rival US-dollar-denominated markets. This raises the
interesting question of whether Asia might be able to achieve similar gains
through the development of a regional bond market.
5 CONCLUDING COMMENT: THE EXAMPLE OF

THAILAND
The Thai economy is illustrative of both the problems we identified in

section 3 and the solutions we outlined in section 4. Before the crisis of
1997, Thailand had a highly developed banking sector and a buoyant stock
market, but a moribund bond market. (See italic figures in Table 5.11 that
show the amount of the bonds outstanding when the crisis erupted.)
The underdevelopment of the Thai bond market can be attributed to
several causes (see Figure 5.1). First is the lack of a benchmark, market-
determined yield curve. Until the crisis, the Thai government had a trad-
ition, dating from 1988, of fiscal surpluses. Since the government viewed
issuance of bonds solely as a means of financing deficits rather than as a
way of nurturing the development of a bond market, no government bonds
were issued from June 1990 until 1998, when the government was forced to
run significant series of deficits to recapitalize the financial system and
boost aggregate demand.
Second, the Thai government had constructed a captive market for its
securities. Banks and finance companies were required to hold substantial
reserves in the form of national government securities and often chose to
hold more bonds than required. These securities were usually held to
Table 5.11 Size of Thai financial markets
Bank Loans Stock Market Bond Outstanding GDP

Capitalization (domestic)
1992 2161.9 1485.0 215.1 2830.9
1993 2665.2 3325.4 262.0 3170.3
1994 3430.5 3300.8 339.0 3634.5
1995 4230.5 3564.6 424.4 4192.7
1996 4825.1 2559.6 519.3 4622.8
1997 6037.5 1133.3 546.8 4732.6
Note: Unit ⫽ billion baht.
Source: Thai Bond Dealing Center.
maturity so that they need not be marked to market. This discouraged sec-
ondary market trading and meant that the interest rate did not necessarily
reflect the true opportunity cost of funds.
Third, tax laws impeded the development of the secondary market. Until
1995, Thailand imposed a stamp duty on transfers of bond ownership.
Although the rate was low, approximately 0.1 percent of the value of the
bond,15 it was a powerful deterrent to secondary market trading.
Fourth, a weak legal infrastructure created doubts about creditor rights
in the event of default. Although Thailand ranked relatively well in terms
of creditor rights (see Table 5.8a), it ranked poorly in terms of judicial
effectiveness. Thailand has made a concerted effort to improve its legal
infrastructure, but survey data from a sample of local law firms and bank-
ruptcy judges in each country, reported in Doing Business in 2004 (World
Bank, 2004) show that Thailand appears to lag behind six other middle
income countries in two important respects. Table 5.12 provides data for
four Asian countries and three other middle income countries, all of which
have recently undertaken structural reforms. Based on the specified bank-
ruptcy scenario, the time to go through insolvency in Thailand (2.6 years)
is exceeded only by Indonesia (6.0 years) and Argentina (2.8 years).
Moreover, the costs of going through insolvency in Thailand (38 percent of
the bankruptcy estate) are markedly higher than in the six other countries.
Fifth, weak accounting and disclosure standards impeded the evaluation
of credit risk and made it difficult for external investors to value risky debt.
Table 5.8c shows that accounting standards in Thailand ranked below
average among the eight Asian emerging economies. Again, there have been
recent efforts to correct this weakness. Based on a study funded by the
Asian Development Bank, Thailand launched its first credit rating agency,
Weak legal infrastructure
Weak accounting standards
Low demand for corporate bond

Underdevelopment
of
PRIMARY MARKET
Tradition of fiscal surpluses
Low supply of public debt Lack of market-

determined benchmark
interest rate
Liquid asset requirement
182
Captive market
Underdevelopment
of
Tax and stamp duty SECONDARY MARKET
Limited disclosure
Inadequate clearing and

settlement system
Figure 5.1 Thailand’s structural problems and consequences for the development of the bond market
Table 5.12 Bankruptcy infrastructure: closing a business in seven

emerging markets
South
Thailand Korea Malaysia Indonesia Turkey Mexico Argentina
Time to Go 2.6 1.47 2.2 6.0 1.8 2.0 2.8
Through
Insolvency
(years)
Cost to Go 38 4 18 18 8 18 18
Through
Insolvency
(% estate)
Absolute 67 100 100 67 67 33 67
Priority
Preserved
Efficient 1 1 0 0 0 1 0
Outcome
Achieved
Goals-of- 62 91 52 35 51 61 43
Insolvency
Index
Court- 33 67 33 100 67 67 67
Powers
Index
Source and notes: Doing Business in 2004, Understanding Regulation (World Bank, 2004).
Based on responses to a questionnaire filled out by local law firms and bankruptcy judges
specifying the details of the insolvency in which the business runs a hotel in downtown real
estate, its only asset. The business is assumed to default on principal and interest 2 January,
2003. The measure for preservation of absolute priority is on a scale of 100. A score of 100
means that secured creditors are paid before labor, tax claims and shareholders. A score of 67
means that secured creditors get paid second. The efficient outcome measure takes on the
value 1 if the insolvency process results in either foreclosure or liquidation with a going-
concern sale or in a successful rehabilitation with new management. A 0 indicates the efficient
outcome was not achieved. The Goals-of-Insolvency Index goes from 0 to 100 and is the
simple average of the time of insolvency and cost of insolvency (each rescaled from 0 to 100)
and the observance of absolute priority. The Court-Powers Index is a measure of the degree to
which the court drives insolvency proceedings and is the average of three indicators: whether
the court appoints and replaces the insolvency administrator with no restrictions imposed by
law, whether the reports of the administrator are accessible only to the court and not the
creditors, and whether the court decides on the adoption of the rehabilitation plan. The index
is scaled from 0 to 100, with higher values indicating more court involvement. South Korea,
Malaysia and Indonesia are included because they experienced crises more or less
simultaneously with Thailand. Argentina, Mexico and Turkey are included because they are
middle income countries that have also undertaken structural reforms.
Table 5.13 Deteriorating corporate performance of Thai firms
Profits No. of Firms Loans of Firms Profits Over Leverage

Over with Profits < with Profits < Liabilities
Interest Interest Interest (%)
Expenses Expenses (%) Expenses (%)
1997: Q4 1.49 32.0 36.4 7.3 2.95
1997: Q3 2.59 23.3 30.8 10.2 2.95
1997: Q2 3.18 19.9 18.4 NA 2.12
1997: Q1 3.66 15.3 16.2 NA 2.01
1996: Q4 3.11 13.8 11.8 14.9 1.90
1995: Q4 4.01 9.6 7.6 18.1 1.67
1994: Q4 5.78 5.1 1.4 24.0 1.50
Note: Profit is defined as earnings before interest, taxes, depreciation and amortization
(EBITDA). Leverage is debt over equity.
Source: Alba et al. (1998).
the Thai Rating and Information Services Company Limited (TRIS) in

1993. TRIS rates both debt securities and companies. All public debt
offerings with a maturity greater than one year require a rating from
TRIS.16
The underdevelopment of the bond market may have caused serious dis-
tortions in the Thai economy. Without a market-determined interest rate
that reflected true opportunity cost of funds, and with bank loan rates
marked-up over deposit rates that were administratively determined, there
was a tendency for Thai firms to overinvest. As a result, the efficiency of
investment declined. Claessens et al. (1998b) report that the median return
on assets for Thai firms declined steadily from 11.7 percent in 1990 to 7.4
percent in 1996. Alba et al.(1998) report four indicators of enterprise per-
formance, using data for all firms listed on the Stock Exchange of
Thailand, that indicate Thai corporate performance had been deteriorating
well before the 1997 financial crisis (see Table 5.13).
The inadequacies of the bond market may have contributed to the heavy
reliance of Thai firms on family group corporate structures. Claessens et al.
(1998) documented that 46.85 percent of Thai firms were affiliated with
corporate groups in 1996.
In the absence of an efficient bond market, firms relied heavily on foreign
borrowing. Table 5.14 shows the evolution of foreign debt of the Thai
private sector from 1987–99. Between 1988–95 it grew at rates ranging from
20 to 65 percent per annum. With limited access to relevant derivatives
markets and risk management tools, foreign borrowing led to excessive
Table 5.14 Foreign debt of Thai private sector
Long-term Short-term Total Growth

1987 3175 2894 6069 ⫺2.7%
1988 3282 4492 7774 28.1%
1989 4966 5777 10 743 38.2%
1990 7633 10 160 17 793 65.6%
1991 10 382 14 686 25 068 40.9%
1992 12 189 18 364 30 553 21.9%
1993 15 302 22 634 37 936 24.2%
1994 20 153 28 999 49 152 29.6%
1995 25 155 41 011 66 166 34.6%
1996 36 172 37 559 73 731 11.4%
1997 34 855 34 238 69 093 ⫺6.3%
1998 31 293 23 373 54 666 ⫺20.9%
1999 25 506 13 546 39 052 ⫺28.6%
Note: Unit ⫽ $ million.
Source: Bank of Thailand.
build up of foreign exchange risk that contributed to the 1997 financial

crisis.
One consequence of the underdeveloped state of the bond market was
that the Thai economy was heavily reliant on bank lending. The year
before the crisis, bank lending accounted for nearly all the external
funding of Thai corporations. Banks financed about one-third of the
change in gross fixed capital formation, while new issues of equity
financed about 6 percent and net issues of bonds, a mere 2 percent.17 (See
Table 5.15.)
The consequence of this dependence on bank lending was catastrophic
for the economy. When the banks suffered heavy losses, new lending ceased
and firms were forced to halt investment projects. The result was a pro-
longed and painful economic contraction.
The Thai authorities have learned a costly lesson about the dangers of
over-reliance on banks. They have begun to implement reforms designed to
stimulate development of both the primary and secondary bond markets.
The Bank of Thailand, acting as agent and advisor to the Ministry of
Finance, has made an effort to introduce a quarter-ahead calendar of
regular issuance of government bonds in the primary market, and with the
government taking on responsibility for many of the costs of financial
sector restructuring, there has been no shortage of supply. The Bank of
Table 5.15 The dominance of bank lending in the external financing of

Thai corporations, 1996
Domestic Credit Stock Market Domestic

Provided by Capitalization Corporate
Banking Sector Debt Securities
Amount Change Total Equity raised Outstanding Net issues
(% GDP) (% GCF) (% GDP) (% GCF) (% GDP) (% GCF)
100.0 31.3 65.8 6.0 3.9* 1.9
Notes and sources: Includes financial institution bonds. 1996, end of year data. The
banking sector includes monetary authorities, deposit money banks, and other banking
institutions for which data are available (including institutions that do not accept
transferable deposits but do incur such liabilities as time and savings deposits). Examples of
other banking institutions include savings and mortgage loan institutions and building and
loan associations. The data are as reported on line 32d in the IMF International Financial
Statistics (IFS). GDP is the gross domestic product as reported on line 99b in the IFS. GCF
is the gross fixed capital formation as reported on line 93e in the IFS. Corporate debt
securities are debt securities that were issued in domestic currency by residents of the
country indicated, including short-term paper (for example, commercial paper).
Thailand has succeeded in developing a yield curve for government bonds

that extends from less than one year out to 15 years.
In June 1999, the Bank of Thailand allowed financial institutions to
conduct securities borrowing and lending business, which has helped
promote risk management and market liquidity. The Thai Bond Dealing
Center (BDC) was granted a Bond Exchange license from the Thai
Securities and Exchange Commission to act as self-regulatory organiza-
tion, institute a code of conduct for market participants and settle any dis-
agreements between participants in the secondary market.
The Bank of Thailand has developed a primary dealership system to
facilitate the conduct of open market operations. Primary dealers are also
obliged to submit reference yields on government securities to the BDC at
the conclusion of each business day to refresh information about the
benchmark yield curve.
The Bank of Thailand has launched a fully automated delivery versus
payment (DVP) settlement system. It has been supplemented by the intra-
day liquidity facilities and queuing mechanism, employing digital signature
technology to ensure secure and smooth delivery and payments trans-
actions, but is not yet available on a real time basis. Clearing and settlement
for corporate bonds has been centralized at the Thailand Securities
Depository Corporation (TSD). Clearing and settlement of government
bonds is being shifted from the Bank of Thailand to the TSD as well.
These changes have markedly improved the liquidity of the government

bond market. As the turnover data in Figure 5.2 show, the volume trading
is substantially higher than at the end of the 1990s. While the volume of
trading in corporate bonds is substantially lower, it is also trending
upwards.
In addition to measures aimed directly at improving the functioning of
local currency bond markets, the Thai government has also placed empha-
sis on upgrading accounting and disclosure standards and improving cor-
porate governance. New accounting standards have been introduced since
the crisis and implemented. Now all companies listed on the Stock
Exchange of Thailand (SET) have audit committees in place. And the
Institute of Directors has provided training to large numbers of corporate
directors. But corporate governance practices in Thailand still fall short of
world standards in some regards. Although Thailand has not yet adopted
the OECD Code of Good Corporate Governance, the SET has adopted a
Code of Best Practices for Directors of Listed Companies. The Institute of
Directors released its second benchmarking survey on corporate gover-
nance practices of the 234 listed corporations with regard to the SET’s
‘Principles of Good Corporate Governance’. Relative to 2001 the survey
showed improvement with regard to five standards and deterioration with
regard to one (Controlling System and Internal Audit).18 A cross-border
survey of corporate governance practices, conducted by CLSA Asia-Pacific
Markets,19 shows that Thai practices compare favorably with those in
China, the Philippines and Indonesia, but fall short of those in Malaysia,
Korea and India.20 Thailand earned particularly low ratings with regard to
‘committed and effective enforcement of rules and regulations’, ‘the polit-
ical and regulatory environment affecting corporate governance and the
ability of companies to maximize value without arbitrary restrictions’, and
‘the adoption of International Generally Accepted Accounting Principles’,
although in each case the rankings showed improvement from 2002 to 2003.
The Thai authorities have also been active participants in regional initia-
tives to strengthen Asian bond markets. For example, Thai officials worked
with the Asian Policy Forum (2001) to improve corporate bond markets in
the region.21 The proposal recognized that bond market development
requires that governments make a number of policy trade-offs. Instead of
borrowing only when a fiscal deficit arises and market conditions appear
favorable, a government must borrow at regular intervals whether it needs
the money or not. Instead of minimizing the cost of debt through requir-
ing some institutions to hold the debt, a government must submit to a
market determination of the interest rate. Similarly, rather than attempting
to minimize the cost of the debt by playing the yield curve – which usually
means borrowing at the short end – a government must issue regularly at
200
180
160
140
Turnover (in %)
120
100
80
188
60
40
20
0
1998 1999 2000 2001 2002 2003
Corporate Bonds Government Bonds
Source: Thai Bond Dealing Center.
Figure 5.2 Trading volume in Thai bond markets

each maturity interval. And, instead of raising revenue through taxing bond
transactions, which discourages trading, the government must remove
stamp taxes. Indeed, the Asian Policy Forum proposal even contemplated
positive tax incentives.
More recently, along with ten22 other Asian central banks, the Thai
Central Bank formed the Executives’ Meeting of East Asia-Pacific Central
Banks (EMEAP) to consider ways to overcome the limitations inherent in
the small size of most of the existing national markets and to work toward
greater integration of these markets to form a regional market. This group
established an Asian Bond Fund 1 (ABF 1) on 2 June, 2003 to invest about
$1 billion in reserve assets in Asian bonds. On 16 December, 2004, a second
fund was launched to invest in the local currency bonds of eight EMEAP
members including Thailand. ABF 2 allocated about $2 billion to a Pan-
Asia Bond Index Fund and a Fund of Bond Funds. ABF 2 is designed to
encourage investment by public investors outside the EMEAP region and
by private sector investors both in Asia and from around the world in local
currency bonds of the EMEAP group.23 The Thai government has also
opened its market to issuance of local currency-denominated bonds by
international financial institutions and borrowers from several other coun-
tries to further encourage the cross-border supply and trading of bonds.
The success of these efforts can be seen in the growth of bond markets
in Asia. Relative to the pre-crisis situation, Table 5.16 shows that six of the
Table 5.16 The size of bond markets (as a % of GDP, average 1990–97
versus 2003)
Bonds outstanding Bonds outstanding

1990–97 2003
Hong Kong 11.3 28.7
Indonesia 1.8 27.5
Malaysia 68.5 95.1
Philippines 33.1 30.7
Singapore 28.4 63.2
South Korea 40.3 73.6
Thailand 9.9 39.7
Canada 81.8 81.4
Japan 90.1 181.8
United States 138.2 160.5
Euro Area 83.9 102.5
Non-euro EU 67.6 83.9
Source: World Bank Financial Structure and Economic Development Database.

Asian economies (the exception being the Philippines) have made substan-
tial progress in developing their bond markets. The Thai bond market is
more than four times larger, relative to GDP, than before the crisis.
As the Thai example has shown, bond markets do matter for financial
development. Certainly, an economy can grow rapidly without an active
bond market. But the cost is an increased vulnerability to a financial crisis
and a loss of information to guide savings and investment decisions. Heavy
reliance on banks means a correspondingly heavy exposure to banking
crises. And the consequence can be catastrophic for the real economy. But
the example of Thailand also shows that it may be possible to rebuild the
financial system with an expanded role for the bond market.
NOTES
1. Hoontrakul (1996) provides a case study for Thailand.

2. Exceptions include Boot and Thakor (1997) and Hakansson (1999). Eichengreen,
Chapter 3, this volume, also highlights the importance of bond markets.
3. Malaysia and South Korea appear to be exceptions to this generalization, but upon
closer inspection neither country’s bond market was as robust as these aggregate data
indicate. The Malaysian market was dominated by issues of Malaysian government
securities, which were placed in a, ‘captive market’ in which most financial intemediaries
faced mandatory investment requirements (Harun, 2002). Until 2000, private issuance
of bonds was deemed ‘deposit taking’ and subject to strict control by the Bank Negara
Malaysia (Dalla et al., 1995, p. 81). The Korean bond market at this time also had an
active corporate sector, but it was dominated by ‘guaranteed’ corporate bonds, for which
the principal and interest were guaranteed by financial institutions that were assumed to
be fully backed by the government. At that time, moreover, the financial sector held
about four-fifths of the bonds and turnover was very low (Kim, 2000).
4. Higher returns on financial instruments may encourage saving; but higher returns also
enable savers to achieve a target stock of wealth with a lower rate of saving. Thus in
theory the impact of expected returns on the overall savings rate is ambiguous. Empirical
studies across a number of countries have not been able to resolve the question.
Nonetheless, higher returns on financial instruments will induce households to allocate
more savings to financial instruments than to real assets such as jewelry and precious
metals that do not contribute to productive investment (and, in an open economy, to
shift from foreign to domestic assets). Efficient financial markets will allocate financial
claims to projects that offer the highest, risk-adjusted returns and so income and total
savings are likely to rise even though the savings rate may not.
5. Rajan and Zingales (1999) suggest that family groups may oppose financial development
because improvements in capital markets would undermine the value of entrenched
positions and increase competition.
6. Yet much of the complexity is obscured by the convention of aggregating flows by sector.
Financial flows among financial firms are often very large relative to flows vis-à-vis other
sectors. For example, interbank trading in the foreign exchange markets is roughly 90
percent of total volume and interbank transactions in the eurocurrency markets are vir-
tually two-thirds of the total.
7. We shall use the term ‘bond’ in the broadest sense to include all tradable fixed income
instruments such as bills and commercial paper (usually with maturities of less than one
year), notes (with maturities of one to five years) and bonds (usually with maturities
greater than five years).
8. Note that when the principal owners are the managers of the firm, the usual arguments
about the disciplinary role of debt in constraining agency problems between the owner
principal and manager agents do not apply.
9. Technically, it is not essential that the term structure be default risk-free. It is necessary,
however, that the benchmark bonds that price the term structure share the same risk of
default. In most markets, government issues, which are approximately default risk-free
in domestic currency terms, provide the benchmarks for estimating the term structure of
interest rates.
10. See, for example, the recent proposal by the Shadow Financial Regulatory Committee
to require that all internationally active banks be required to issue subordinated debt
(Calomiris et al., 2000).
11. The US, for example, is currently reducing the effective maturity of its outstanding
debt. Because the US has a highly developed bond market with abundant issues by
government-sponsored enterprises that serve as close substitutes for government debt, it
may be able to reduce gross debt without undermining the efficiency of the bond market.
This would not be a wise policy in an emerging market, however.
12. When Citibank introduced the Certificate of Deposit, it was careful to make
arrangements with dealers to establish active secondary markets in which CDs could be
traded.
13. Publicly traded corporations issue letter stock frequently in making acquisitions or
raising capital when the time and cost of registering the new stock with the SEC
(Securities and Exchange Commission) would make the transaction impractical. Even
though such stock cannot be sold to the public on the open market, it may be sold in
private transactions under certain circumstances. Such transactions must be reported to
the SEC where they become a matter of public record Pratt (1989, p. 240).
14. Fernando and Herring (2002) have shown how liquidity depends on the diversity of
market participants and provide evidence from a natural experiment in which the sec-
ondary market for perpetual floating rate notes collapsed and prices fell from 15 to 30
percent even though the credit standing of the issuers had not changed.
15. Emery (1997).
16. In 1999, the Securities and Exchange Commission passed a resolution requiring private
placement debt of more than 100 million baht to be rated.
17. Retained earnings, of course, financed the remainder of the change in gross fixed capital
formation.
18. Thailand: Economic Monitor, World Bank Thailand Office (2003), p. 33, Table 8.
19. CLSA Asia-Pacific Markets is a provider of brokerage and investment banking services,
headquartered in Hong Kong.
20. Thailand: Economic Monitor, World Bank Thailand Office (2003), p. 34.
21. See also Yoshitomi and Shirai (2001) for an extensive discussion of many of the issues
that underlie the proposals.
22. The other nine include the Reserve Bank of Australia, People’s Bank of China, Hong
Kong Monetary Authority, Bank of Indonesia, Bank of Japan, Bank of Korea, Bank
Negara Malaysia, Reserve Bank of New Zealand, Bangko Sentral ng Pilipinas, and the
Monetary Authority of Singapore.
23. For additional discussion, see Asia Bond Monitor, April 2005, Asian Development Bank.
REFERENCES
Alba, Pedro, Stijn Claessens and Simeon Djankov (1998), ‘Thailand’s Corporate
Financing and Governance Structures: Impact on Firms’ Competitiveness’,
Paper for the World Bank Conference on Thailand’s Dynamic Economic
Recovery and Competitiveness.
Allen, Franklin and Douglas Gale (2000), Comparing Financial Systems,

Cambridge, MA: The MIT Press.
APEC (1999), Compendium of Sound Practices, Guidelines to Facilitate the
Development of Domestic Bond Markets in APEC Member Economies, Report of
the Collaborative Initiative on Development of Domestic Bond Markets.
Arrow, Kenneth (1979), ‘Pareto Efficiency With Costly Transfers’, Economic
Forum, 10, 1–13.
Asian Development Bank (1999), ‘Governance in Thailand: Challenges, Issues and
Prospects’, http://www.adb.org/Work/Governance/thailand.pdf, April.
Asian Policy Forum (2001), ‘Policy Recommendations for Designing New and
Balanced Financial Market Structures in Post-crisis Asia, Proposals on How to
Foster Bond Markets through Strengthening the Banking Sector’, Asian
Development Bank Institute, Forum Secretariat, October.
Bank for International Settlements, International Financial Statistics,
http://www.bis.org/statistics.
Barro, Robert (1974), ‘Are Government Bonds Net Wealth?’, Journal of Political
Economy, November/December, 1095–117.
Black, Bernard and Ronald Gilson (1998), ‘Venture Capital and the Structure of
Capital Markets: Banks Versus Stock Markets’, Journal of Financial Economics,
47, 243–77.
Black, Fischer (1976), ‘The Pricing of Commodity Contracts’, Journal of Financial
Economics, 3.
Black, Fischer and Myron Scholes (1973), ‘The Pricing of Options and Corporate
Liabilities’, Journal of Political Economy, 81, 637–59.
Boot, Arnoud and Anjan Thakor (1997), ‘Financial System Architecture’, Review
of Financial Studies, 10 (3), 693–733.
Calomiris, Charles et al. (2000), Reforming Bank Capital Regulation, Proposal by
the US Shadow Financial Regulatory Committee, Washington DC: American
Enterprise Institute.
Caprio, Gerard, Jr and Aslï Demirgüç-Kunt (1997), ‘The Role of Long Term
Finance: Theory and Evidence’, World Bank Working Paper, February.
Claessens, Stijn, Simeon Djankov and Larry H.P. Lang (1998a), ‘Who Controls
East Asian Corporations?’, Working paper, World Bank.
Claessens, Stijn, Simeon Djankov and Larry H.P. Lang (1998b), ‘East Asian
Corporates: Growth, Financing and Risks over the Last Decade’, Working paper,
World Bank.
Claessens, Stijn, Simeon Djankov, Joseph P.H. Fan and Larry H.P. Lang (1998),
‘Corporate Diversification in East Asia: The Role of Ultimate Ownership and
Group Affiliation’, Working paper, World Bank.
Copeland, Morris (1955), A Study of Money Flows in the United States, New York:
National Bureau of Economic Research.
Dalla, Ismail, Deena Khatdhate, D.C. Rao, Kali Kondury, Lwang Jun and Terry
Chuppe (1995), The Emerging Asian Bond Market, Washington DC: The World
Bank.
Diamond, Douglas (1991), ‘Debt, Maturity and Liquidity Risk’, Quarterly Journal
of Economics, 106, 709–37.
Emery, Robert F. (1997), The Bond Markets of Developing East Asia, Boulder,
Colorado: Westview Press.
Fernando, Chitru and Richard Herring (2002), ‘Commonality in Liquidity and
Market Collapse: Theory and Application to the Market for Perps’, Working
paper, Wharton Financial Institutions Center, The Wharton School, University

of Pennsylvania.
Goldsmith, Raymond (1965), The Flow of Capital Funds in the Postwar Economy,
New York: National Bureau of Economic Research, Chapter 2.
Goldsmith, Raymond (1985), Comparative National Balance Sheets: A Study of
Twenty Countries, 1688–1978, Chicago: University of Chicago Press.
Guttentag, Jack M. and Richard J. Herring (1986), ‘Financial Innovations
to Stabilize Credit Flows to Developing Countries’, Studies in Banking and
Finance, 3.
Hakansson, Nils H. (1982), ‘Changes in Financial Markets: Welfare and Price
Effects and the Basic Theories of Value Conservation’, Journal of Finance, 37 (4),
September, 977–1005.
Hakansson, Nils H. (1992), ‘Welfare Economics of Financial Markets’, in The New
Palgrave Dictionary of Money and Finance, John Eatwell, Murray Milgate and
Peter Newman (eds), London: MacMillan Press, pp. 790–96.
Hakansson, Nils H. (1999), ‘The Role of a Corporate Bond Market in an Economy
– and in Avoiding Crises’, China Accounting and Finance Review, 1 (1), March.
Hart, Oliver and John Moore (1995), ‘Debt and Seniority: An Analysis of the Role
of Hard Claims in Constraining Management’, American Economic Journal,
June, 567–85.
Harun, Dato’ Salleh (2002), ‘The Development of Debt Markets in Malaysia’, BIS
Papers No.11, Bank for International Settlements, 147–50
Hoontrakul, Pongsak (1996), ‘Should Thailand Have Bank Based or Stock Market
Based Financial Systems?’, Discussion paper, Sasin-GIBA.
Hausman, Jerry A. and James M. Poterba (1987), ‘Household Behavior and the Tax
Reform Act of 1986’, Economic Perspectives, 1 (1), summer, 101–19.
Herring, Richard (1989), ‘The Economics of Workout Lending’, Journal of Money,
Credit and Banking, 21, February, 1–15.
International Monetary Fund (1999), International Financial Statistics Yearbook.
International Monetary Fund (2005), Global Financial Stability Review, April.
Kane, Edward J. (2000a), ‘Architecture of Supra-governmental International
Financial Regulation’, Working paper, Boston College.
Kane, Edward J. (2000b), ‘Designing Financial Safety Nets to Fit Country
Circumstances’, Working paper, Boston College.
Kim, Yongbeom (2000), ‘Korea’s Bond Market Following the Onset of Financial
Crisis’, Building Local Bond Markets: An Asian Perspective, International
Finance Corporation, pp. 63–83.
La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer and Robert Vishny
(1998), ‘Law and Finance’, Journal of Political Economy, 106, 1113–55.
Levine, Ross (1997), ‘Financial Development and Economic Growth: Views and
Agenda’, Journal of Economic Literature, 36, 688–726.
McKinnon, Ronald (1973), Money and Capital in Economic Development,
Washington: Brookings Institution.
Merton, Robert (1989), ‘On the Application of the Continuous-time Theory of
Finance to Financial Intermediation and Insurance’, The Geneva Papers on Risk
and Insurance, 14 (52), July, 225–61.
Merton, Robert (1990), ‘The Financial System and Economic Performance’,
Journal of Financial Services Research, 4, December, 263–300.
Pratt, Shannon P. (1989), Valuing a Business, The Analysis and Appraisal of Closely
Held Companies, Homewood: Dow Jones-Irwin, 2nd Edition.
Rajan, Raghuram and Luigi Zingales (1999), ‘The Politics of Financial

Development’, Working paper, University of Chicago.
Shaw, Edward (1973), Financial Deepening in Economic Development, New York:
Oxford University Press.
Sheng, Andrew (1994), ‘Future Directions for Hong Kong’s Debt Market’, Speech
at the First Annual Pan-Asia Bonds Summit 29–30 November.
Smith, Clifford, Jr, Charles Smithson and Lee Wakeman (1986), ‘The Evolving
Market for Swaps’, Midland Corporate Finance Journal, winter.
Sonakul, M.R. Chatu Mongol (2000), ‘Keynote Address on the Occasion of the
ADB Conference on Government Bond Market and Financial Sector
Development in Developing Asian Economies’, Manila, 28–30 March.
Stiglitz, Joseph E. and Andrew Weiss (1981), ‘Credit Rationing in Markets with
Imperfect Information’, American Economic Review, LXXXI (3), June, 393–410.
Stulz, René (2000), ‘Does Financial Structure Matter for Economic Growth? A
Corporate Finance Perspective’, Working paper, 24 January.
Thai Bond Dealing Center, http://www.thaibdc.or.th.
Weinstein, D.E. and Y. Yafeh (1998), ‘On the Costs of a Bank Centered Financial
System: Evidence from the Changing Main Bank Relations in Japan’, Journal of
Finance, 53 (2), 635–72.
World Bank (2004), Doing Business in 2004: Understanding Regulation,
Washington: The International Bank for Reconstruction and Development/The
World Bank.
World Bank (2005), Financial Structure and Economic Development Database,
http://www.worldbank.org/research/projects/finstructure/database.htm.
World Bank Thailand Office (2003), Thailand: Economic Monitor, October, http://
siteresources.worldbank.org/INTTHAILAND/Resources/Economic-Monitor/
2003oct.pdf.
Yam, Joseph (1997), ‘Development of the Debt Market, Keynote Address at the
Asian Debt Conference’, http://www.info.gov.hk/hkma/eng/speeches/speechs/
joseph/speech_140797b.htm, 14 July.
Yoshitomi, Masaru and Sayuri Shirai (2001), ‘Designing a Financial Market
Structure in Post-crisis Asia’, ADB Institute Working Paper 15, 27 March.
PART III
Analytical Issues Pertaining to the

Recent Crises
6. Investment, growth and productivity
during the East Asian Financial
Crisis
F. Gerard Adams and Tayyeb Shabbir
1 INTRODUCTION
Almost ten years have passed since the East Asian Financial Crisis so
rudely interrupted the rapid expansion of the East Asian economies. Most
scholars have sought to evaluate the crisis phenomenon and its impact in
terms of financial flows, international exchange exposure, contagion, etc.
All of these factors seem to have had some relevance (Corsetti et al., 1998a
and 1998b; Radelet and Sachs, 1998). Comparatively, little analysis has
been devoted to the crisis from a production input/factor productivity per-
spective. Such a view has important implications for the growth prospects
of the East Asian countries.
In this chapter, we will consider the East Asian Crisis from the perspec-
tive of macroeconomic performance, its impact on investment, growth and
productivity. First, we discuss the East Asian growth process and its rela-
tionship to underlying growth forces. Then, we look at developments from
an empirical perspective, computing total factor productivity and testing
various explanations for it. Finally, we will speculate about the long-run
implications for future development in East Asia.
2 ECONOMIC DEVELOPMENT, GROWTH AND THE

FINANCIAL CRISIS
East Asian growth has been a process of transformation, turning low

income subsistence agriculture countries into export industry powerhouses.
Traditional development theory sees East Asian growth as export-led
industrialization (Herrick and Kindleberger, 1983). High rates of domestic
investment and capital inflows, improved production competence, and,
often, international entrepreneurship enabled these countries, one after the
197
198 Analytical issues pertaining to the recent crises
other, to use their comparative advantage in low labor costs to compete

effectively in world markets. Rising living standards then contributed to a
growing domestic market that has also benefited East Asian industry.
The economic development of East Asia has sometimes been described
as a sequential growth pattern followed by the countries in the region.
Akamatsu (1961, 1962) pioneered the concept of the ‘Flying Geese Pattern’
(Kojima, 2000), an idea that has often been interpreted as a sequence of
policy steps following the leader country, presumably Japan. But
Akamatsu’s early work already pointed to a rationale based on the chang-
ing patterns of comparative advantage as countries advance along the
development ladder (Adams, 1998). The basic notion assumes that rela-
tively low wages and favorable exchange rates initially provide comparative
advantage in labor-intensive products like apparel and consumer electron-
ics exports to the world market. As development advances, rising wage
levels and appreciating exchange rates reduce comparative advantage in
labor-intensive products. At this stage of development, production of tra-
ditional products is shifted to countries that still have lower costs and the
more advanced East Asian countries turn to more sophisticated and, typi-
cally, more capital- and technology-intensive products. As the required
technology comes closer to the technological frontier, growth is somewhat
slower. Finally, we can envision a still more advanced stage of development
with emphasis on sophisticated services like finance, management and high
technology. The essential ingredient during the transition period is the shift
of production toward industrial output and the acquisition of knowledge
and experience (and capital) that makes possible the production of items
that are competitive in the world market. High rates of investment, redi-
rection of economic activity toward industrial products for domestic and
export markets, and direct foreign investment appear to be important in
facilitating these developments.
Another perspective on East Asian growth that we will emphasize, is the
role of inputs and of total factor productivity (TFP), a term that summa-
rizes structural changes and economies of scale, as well as technological
change. The growth accounting framework has been applied to examine
East Asian growth in a number of earlier papers (Collins and Bosworth,
1996; Crafts, 1999). The principal issue in these papers is to explain East
Asia’s rapid growth, specifically the explanation for East Asian growth as
‘accumulation versus assimilation’1 and to evaluate whether there remains
substantial room for a productivity catch-up. The magnitude of an unex-
plained total factor productivity growth component in East Asian growth is
a matter of controversy (World Bank, 1993) ever since empirical economists
(Krugman, 1994; Young, 1995) have reported that, for some East Asian
countries, rapid growth has been primarily a matter of more intensive input
Investment, growth and productivity during the East Asian Financial Crisis 199
use rather than increased Total Factor Productivity (TFP).2 By and large,
the later empirical work suggests that while a large share of the growth in
the region is attributable to growth of inputs, there remains substantial
room for productivity catch-up (Timmer and Szirmai, 1997; Lau, 1998).
The growth accounting papers have used the traditional production func-
tion approach and then have attempted to explain the unexplained residual
in terms of a number of general and country-specific variables. Some of
these variables relate to the quality of inputs, for example, mean years of
schooling acquired by labor, but others relate to macroeconomic issues like
the exchange rate and the government budget deficit. Interestingly, these
studies have not taken into account the impact of the business cycle, despite
the fact that business fluctuations are known to be an important variable in
explaining short-term variations in productivity in the industrial countries.3
Being aware of the sometimes heated debate that can involve the concept
of total factor productivity (TFP), we would like to note that we intend to
use the TFP concept non-judgmentally, that is, only as a mechanical way for
summarizing growth that is not directly the result of increased factor inputs
and not associate any one specific interpretation to this residual. Then we
will try to explain computed TFP in terms of variables that relate to the
speed of growth, industrialization and exports. An important aspect of our
explanation will be the impact of the fluctuation in real output that was
associated with the 1997 financial crisis. We stress that an emphasis on real
output and productivity does not necessarily contradict the linkages
between the rate and patterns of development and international financial
considerations like capital flows, exchange rates, competitiveness and expec-
tations that lie behind traditional explanations of the East Asian Crisis.
3 THE FINANCIAL CRISIS AND THE GROWTH

RECORD
As is well known, the rapid development of East Asia was interrupted in the
late 1990s by the East Asian Financial Crisis, introducing a sharp reversal
in output growth and investment throughout the region. The impact was
uneven with little apparent influence on China but with serious recessionary
effect on the domestic economies of some countries, particularly Thailand,
South Korea and Indonesia. Rapid growth of exports, in part, a result of
financial crisis-induced exchange rate readjustments and a gradual return to
higher rates of investment (though not as a share of GDP) have enabled
most East Asian countries to resume their rapid growth in recent years.
Let us start by looking here at the East Asian growth record (Figure 6.1).
The 1997 and 1998 period shows a severe interruption of real growth at the
20.0
15.0
10.0
5.0
% per year
0.0
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
200
–5.0
–10.0
–15.0
China, People’s Rep. of Korea, Rep. of Indonesia

Malaysia Philippines Thailand
Figure 6.1 GDP growth rates

Table 6.1 Growth of real GDP, pre- and post-crisis (% per year)
Pre-crisis Financial Post-crisis Post vs.

Period Crisis Period Pre-crisis
1987–91 1991–96 1997 1998 1999 1999–2003 Difference*
China, People’s 8.0 12.1 8.8 7.8 7.1 8.2 ⫺3.9
Rep.
Hong Kong, 6.6 5.3 5.1 ⫺5.0 3.4 3.9 ⫺1.4
China
Korea, Rep. of 9.1 7.0 4.7 ⫺6.9 9.5 5.6 ⫺1.4
Indonesia 7.5 7.6 4.7 ⫺13.1 0.8 4.0 ⫺3.6
Malaysia 8.6 9.6 7.3 ⫺7.4 6.1 4.5 ⫺5.1
Philippines 3.9 3.5 5.2 ⫺0.6 3.4 4.1 0.6
Singapore 9.3 9.3 8.6 ⫺0.9 6.9 2.8 ⫺6.5
Thailand 10.9 8.1 ⫺1.4 ⫺10.5 4.4 4.8 ⫺3.3
Taipei 8.4 6.8 6.7 4.6 5.4 2.6 ⫺4.2
Note: * (1999–2003) minus (1991–96).
Source: Asian Development Bank.
time of the East Asian Financial Crisis for all the East Asian countries
except China. We show the relevant averages in Table 6.1 for the pre-crisis
period, for the time of the financial crisis – 1997, 1998 and 1999 – and for
a post-crisis period. Before the crisis, the growth rates are extraordinarily
high in all the countries except the late-blooming Philippines, ranging from
6.8 percent in Taipei to 12.1 percent in China. The crisis-induced down-
swing is clearly apparent in 1997–98 particularly in Indonesia, Thailand,
Malaysia and South Korea. The period 1998–99 is a time that marks the
start of the recovery. The post-crisis period shows renewed growth but, with
the exception of the Philippines, at substantially lower growth rates than in
the pre-crisis period.
Now let us examine labor productivity which provides a simple way to
measure improvements in a nation’s productive power. Figure 6.2 and
Table 6.2 show the movement of labor productivity in East Asia. For most
countries, except China, the rate of productivity growth shows fluctuations,
in relation particularly to the 1997–98 crisis. There appears to be a down-
ward trend, with growth in the post-crisis period substantially lower than
in the years that preceded the crisis.
The 1997–98 crisis is dominated by a huge swing in the share of fixed
investment in GDP (Table 6.3, Figure 6.3). Before the crisis, gross fixed
capital formation had risen to high levels relative to GDP, between 30 and
40 percent, in all the East Asian countries, except Thailand and the
15.0
10.0
5.0
0.0
% per year
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
–5.0
202
–10.0
–15.0
–20.0

Figure 6.2 Labour productivity growth

Table 6.2 Improvement of labor productivity, pre- and post-crisis

(% per year)
Pre-crisis Post-crisis
Period Period
1987–91 1991–96 1997 1998 1999 2000–03 Difference*
China, People’s 2.8 11.1 7.6 6.6 6.1 7.2 ⫺3.9
Rep. of
Hong Kong, 5.6 3.1 2.1 ⫺3.7 3.7 3.0 0.0
China
Korea, 5.4 4.8 2.9 ⫺0.8 7.7 3.4 ⫺1.4
Rep. of
Indonesia 5.3 5.2 5.0 ⫺15.8 ⫺0.5 3.5 ⫺1.8
Malaysia 4.7 5.5 5.3 ⫺7.7 3.4 1.7 ⫺3.7
Philippines 1.7 ⫺0.2 3.6 ⫺1.9 5.2 0.8 0.9
Singapore 4.7 6.5 3.9 ⫺3.1 6.0 0.7 ⫺5.8
Thailand 7.8 7.4 ⫺4.3 ⫺7.4 4.6 2.8 ⫺4.6
Taipei 6.6 5.4 5.5 3.3 4.4 2.1 ⫺3.3
Note: * (2000–2003) minus (1991–96).
Table 6.3 Investment as share of GDP: pre- and post-crisis (% of GDP)
1986–90 1991–96 1997 1998 1999–2003 Difference*
China, People’s 36.3 39.3 38.2 37.7 39.4 0.1
Rep. of
Hong Kong, 26.7 30.3 34.5 29.2 25.1 ⫺5.2
China
Korea, 30.4 31.3 31.8 16.8 15.4 ⫺15.9
Rep. of
Indonesia 32.4 37.6 36.0 25.0 29.6 ⫺8.0
Malaysia 27.5 39.8 43.0 26.7 23.8 ⫺16.0
Philippines 19.4 22.7 24.8 20.3 19.7 ⫺3.0
Singapore 36.1 35.1 39.2 32.3 24.7 ⫺10.4
Thailand 21.7 24.8 24.2 24.9 19.6 ⫺5.2
Taipei, China 32.6 41.2 33.7 20.4 23.3 ⫺17.9
Note: * (1999–2003) minus (1991–96).

50.0
45.0
40.0
35.0
30.0
% per year
25.0
20.0
15.0
204
10.0
5.0
0.0
86
87
88
89
90
91
92
93
94
95
96
97
98
99
00
01
02
03
19
19
19
19
19
19
19
19
19
19
19
19
19
19
20
20
20
20
China, People’s Rep. of Indonesia Taipei
Malaysia Philippines
Thailand Korea, Rep. of
Figure 6.3 Gross domestic fixed investments as % of GDP

Table 6.4 Growth of capital stock:** pre- and post-crisis (% per year)
1986–91 1991–96 1997 1998 1999 1999–2003 Difference*
China, People’s 9.30 9.60 10.10 9.40 9.00 8.80 ⫺0.90
Rep. of
Hong Kong, 4.70 5.50 7.00 7.60 5.90 3.80 ⫺1.80
China
Korea, 7.30 9.20 12.10 9.90 5.40 5.90 ⫺3.30
Rep. of
Indonesia 3.00 6.90 8.70 8.70 3.40 2.50 ⫺4.40
Malaysia 5.70 11.50 12.80 12.20 3.80 3.80 ⫺7.80
Philippines 2.30 3.00 4.20 4.80 3.30 3.30 0.30
Singapore 7.40 8.20 10.70 10.70 8.30 5.80 ⫺2.30
Thailand 8.70 10.70 10.10 5.90 0.70 0.80 ⫺9.90
Taipei 2.70 5.50 5.70 6.20 6.40 4.60 ⫺0.90
Notes:
* (1991–96) minus (1999–2003).
** Computed from investment, an initial capital output ratio of 3 and an assumed
depreciation rate of 0.05 per year.
Philippines. This represents an extremely high rate since it implies an esti-

mated increase in capital stock and, in capital, on the basis of our estimates,
of almost 10 percent annually (Table 6.4). This rate of capital accumula-
tion was maintained even in 1997 but dropped sharply in 1998 except in
China. In fact, the post-crisis period shows relatively small recovery of the
investment/GDP ratio. As compared with the pre-crisis period, capital
investment in all East Asian countries has been accounting for a substan-
tially lower share of GDP post-crisis with the glaring exception of China.
Also, the rate of capital stock accumulations has been substantially lower
post-crisis in all East Asian countries except the Philippines (Table 6.4).
Remarkably, the mostly adverse swing in investment was not matched by
a similar swing in domestic saving, which remained comparatively stable as
a share of GDP, so that post-crisis domestic saving greatly exceeded domes-
tic investment in some East Asian countries, particularly in Malaysia and
Singapore (Figure 6.4). The reconciling figures were, obviously, capital
flows – foreign inflows turned sharply to outflows after the financial crisis.
Finally, note that export growth declined sharply in advance of and
during the crisis (Figure 6.5, Table 6.5). In fact, it has been argued that the
lag in the performance of export markets as early as in 1996, was one of the
40.0
35.0
30.0
25.0
20.0
% per year
15.0
10.0
5.0
206
0.0
1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
–5.0
–10.0
–15.0
China, People’s Rep. of Indonesia Korea, Rep. of

Malaysia Philippines Singapore
Thailand
Figure 6.4 Domestic saving less domestic investment (% of GDP)

50.0
40.0
30.0
% per year
20.0
10.0
207
0.0
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
–10.0
–20.0
China, People’s Rep. of Indonesia Korea, Rep. of
Taipei
Figure 6.5 East Asian export growth (in US$)

Table 6.5 Growth of exports: pre- and post-crisis (In US $, % change

per year)
1986–91 1991–95 1996 1997 1998 1999 1999–2003 Difference*
China 18.4 21.8 17.9 20.9 0.5 6.1 22.9 1.1
Hong Kong 23.0 15.2 4.0 4.1 ⫺7.5 0.1 6.9 ⫺8.4
Indonesia 14.5 11.8 9.8 7.1 ⫺8.6 ⫺0.4 10.7 ⫺1.0
Korea 16.6 16.4 4.6 4.8 ⫺7.9 8.2 8.3 ⫺8.1
Malaysia 19.8 21.1 6.1 0.9 ⫺6.9 15.1 9.5 ⫺11.5
Philippines 13.2 18.6 18.3 22.8 16.9 20.3 5.1 ⫺13.5
Singapore 21.7 19.2 5.9 0.2 ⫺12.3 4.4 6.6 ⫺12.6
Thailand 26.8 18.8 ⫺2.5 3.3 ⫺5.3 7.3 8.8 ⫺10.1
Taiwan 14.3 10.2 3.8 5.3 ⫺9.4 10.0 5.4 ⫺4.8
Note: * (1999–2003) minus (1991–95).
major contributing factors to the onset of the 1997 crisis (Adams, 1998).
The recovery of exports after devaluation provided an important offset to
the lower investment share in the post-crisis period, but export growth in
the post-crisis period has been lower than before the crisis. China is, again,
a glaring exception. The importance of exports here is that export produc-
tion is hypothesized to contribute to the growth of productivity.
4 DISENTANGLING EAST ASIAN GROWTH: AN

INPUT/TOTAL FACTOR PRODUCTIVITY
APPROACH
In view of the complex interactions between the various forces to which the
East Asian Crisis can be attributed, it is helpful to look at the phenomenon
from the perspective of the underlying forces influencing the production
potential of the region’s economies. For this purpose, we take an input/total
factor productivity view of the developments. Growth of output can be dis-
entangled into that attributable to increased inputs and the residual factor,
often termed ‘total factor productivity’. (Denison, 1974; Solow, 1956).
Following Solow, the residual represents the difference between the growth
of total output and a weighted sum of growth in labor and capital inputs:
% change of TFP ⫽ % change of GDP – (wl* % change

of labor ⫹ wk * % change of capital).
The weights wl and wk correspond to factor shares assuming perfect com-

petition and constant returns to scale.4
This decomposition analysis represents a very simple yet quite useful way
to characterize the growth process. With standard assumptions about the
weighting of the inputs and returns to scale, it provides a framework for
empirical analysis. Crafts (1999) and Collins and Bosworth (1996) consider
the possibilities for appropriate coefficients at some length. A range of
weights for the capital component, from 0.3 to 0.4 is considered, with the sug-
gestion that the elasticity of output with respect to capital might be on the
high end of the range in East Asia. We have assumed wk ⫽ 0.4 and wl ⫽ 0.6.5
The impact of the coefficient assumptions that underlie the calculation was
validated with a sensitivity analysis.6 As noted earlier, we are interpreting the
meaning of the residual in a limited sense since it includes all elements that
have not been taken into account in the computation of the growth of inputs.
In addition to technological change, these include such factors as possible
economies of scale, the composition of output, the role of exports, and,
importantly, the cyclical position of the economy.7 It is possible to hypothe-
size about the forces that might influence TFP, some closely related to indus-
trial growth and consequent shifting from inefficient to efficient sectors, and
others representing technical or managerial change that is associated with
development of science and production techniques and which is more diffi-
cult to measure. In summary, we can define unexplained change in TFP as the
remaining effect unaccounted for in such ‘systematic’ influences as business
cycle effects as well as industrial composition and export effects. Thus:
Unexplained change of TFP ⫽ change of TFP – business cycle effect –

industrial composition minus export
effect.
Unexplained TFP then includes interaction effects, increasing returns and

pure technical change.
Statistics on estimated total labor and capital inputs are presented in
Figure 6.6 and, on a pre-crisis and post-crisis basis in Table 6.6 and corre-
sponding figures for the residual growth factor are shown in Figure 6.7 and
Table 6.7. The financial crisis period shows a sharp decline in the weighted
sum of the growth rate of factor inputs in all countries except China and
Taipei. Notable is the fact that input growth remains lower in all East Asian
countries post-crisis than in the pre-crisis period, particularly in Indonesia,
Malaysia and Thailand.
The growth of the residual (TFP) shows considerable variability from
year to year and, particularly, over the financial crisis period. This reflects
changes in efficiency of capital and labor use during the cycle. Capital stock
14.0
12.0
10.0
% per year
8.0
6.0
4.0
210
2.0
0.0
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003

Taipei
Figure 6.6 Growth of East Asian inputs of labor and capital (% per year, wk ⫽ 0.6 and wk ⫽ 0.4)
Table 6.6 Growth of capital and labor inputs in East Asia: pre-crisis and
post-crisis (% per year assuming wk ⫽ 0.4 and wl ⫽ 0.6)
1987–91 1991–96 1997 1998 1999 1999–2003 Difference*
China, People’s 6.1 4.5 4.8 4.5 4.2 4.1 ⫺0.3
Rep. of
Hong Kong, 2.1 3.5 4.5 2.3 2.1 2.0 ⫺1.5
China
Korea, 4.6 5.0 5.9 0.4 3.2 3.7 ⫺1.3
Rep. of
Indonesia 2.3 4.2 3.3 5.1 2.2 1.3 ⫺2.8
Malaysia 4.1 7.1 6.3 5.1 3.2 3.2 ⫺3.9
Philippines 2.1 3.4 2.7 2.7 0.3 3.3 ⫺0.1
Singapore 5.2 4.9 7.1 5.6 3.9 3.6 ⫺1.4
Thailand 4.7 4.7 5.8 0.5 0.2 1.5 ⫺3.2
Taipei, China 1.9 3.1 3.0 3.2 3.2 2.1 ⫺0.9
Note: Computed. *(1999–2003) minus (1991–96).
continues to grow during the first years of the financial crisis (note the esti-
mates in Table 6.4) but output declines. Similarly, procyclical variation of
labor productivity is well-known – employers delay hiring during the
upswing and postpone layoffs during the downswing, producing cyclical
variations in labor productivity over the cycle. In the post-crisis period,
TFP growth is considerably lower than in pre-crisis China, Singapore and
Taipei, but not in the other East Asian countries. In China, this may reflect
statistical changes towards a ‘norm’, given the astonishingly high rate
recorded earlier. In Taipei and Singapore, the decline in TFP growth may
be a result of the fact that these countries are approaching the technologi-
cal frontier in their high-tech industries.
5 STATISTICAL ANALYSIS OF ‘DETERMINANTS’

OF TFP
We proceed next to a statistical analysis of factors associated with the

growth of TFP. As noted, TFP is measured as a residual from the standard
growth accounting relationship and may encompass a number of factors –
systematic as well as stochastic. Moreover, technical change is difficult to
measure directly, since changes in production methods are difficult to pin
down numerically. We test a number of factors suggested in our discussion
15.0
10.0
5.0
% per year
0.0
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
–5.0
212
–10.0
–15.0
–20.0

Taipei
Figure 6.7 East Asian total factor productivity 1987–2003

Table 6.7 Total factor productivity growth in East Asia: pre-crisis and
post-crisis (% per year assuming wk ⫽ 0.4 and wl ⫽ 0.6)
1987–91 1991–96 1997 1998 1999 1999–2003 Difference*
China, People’s 1.9 7.6 4.1 3.4 2.9 4.1 ⫺3.5
Rep. of
Hong Kong, 4.5 1.8 0.5 ⫺7.2 1.3 1.9 0.1
China
Korea, 4.5 2.0 ⫺1.2 ⫺7.2 6.2 1.9 ⫺0.1
Rep. of
Indonesia 5.2 3.4 1.4 ⫺18.2 ⫺1.4 2.9 ⫺0.5
Malaysia 4.4 2.5 1.0 ⫺12.5 3.0 1.3 ⫺1.2
Philippines 1.9 0.1 2.5 ⫺3.3 3.1 0.8 0.7
Singapore 4.1 4.4 1.4 ⫺6.5 3.0 ⫺0.8 ⫺5.2
Thailand 6.3 3.4 ⫺7.1 ⫺11.0 4.2 3.3 ⫺0.1
Taipei, China 6.4 3.8 3.7 1.3 2.2 0.5 ⫺3.3
Note: Computed. * (1999–2003) minus (1991–96).
above that might affect TFP, either by moving businesses away from the his-
torical production path, such as in the financial crisis or shifts into new
industries, or that might move the productivity path such as technical
change incorporated in investment.
5.1 Financial Crisis Impacts
The 1997 financial crisis is the sharpest and most pervasive business
fluctuation episode that affected the East Asian countries during the past two
decades. As we have noted, there were severe declines in real output in most
East Asian countries in 1997–98. We have examined the effect of these events
on productivity using regressions linking TFP to a series of dummy variables
covering the 1997–2001 period (Table 6.8). The significant coefficients for
1997 and 1998, amounting to 1 percent in 1997 (largely the effect on
Thailand) and 9 percent for 1998, show that declines in production have clear
impacts on TFP, probably through underutilization of capital and labor.
Note that these effects are in addition to effects related to use of less labor
and slower growth of capital that have been incorporated into inputs.
A closely related question is whether the impact of the crisis was sym-
metrical, in the sense that the loss in productivity suffered in 1997–98 was
made up at a later point. That does not appear to be the case. With refer-
ence to Equation 2, the regression showing dummies for 1997, 1998 and
Table 6.8 Financial crisis impact on TFP
Equation 1
Coefficient Standard error Significance
Dependent variable
% change TFP
Independent variables
Dummy 1997 ⫺2.516 1.155 Significant
Dummy 1999 ⫺0.474 1.155 Non-significant
Constant 3.213 0.298
Adjusted R2 0.343
Equation 2
Dependent variable
% change TFP
Dummy 1999 ⫺1.055 1.136 Non-significant
Dummy 2000–03 ⫺2.031 0.640 Significant
Adjusted R2 0.385
Equation 3
Dependent variable
% change TFP
Time ⫺0.313 0.113 Significant
Dummy 1997 ⫺1.371 1.274 Marginal
Dummy 1999 1.297 1.399 Non-significant
Dummy 2000–03 1.105 1.294 Non-significant
Adjusted R2 0.416
2000–03 shows negative coefficients, significant for 2000–03. A time trend,

Equation 3, is significantly negative, suggesting that productivity growth is
slowing over a longer period. In this case, the coefficients of the 1999 and
the 2000–03 dummies are positive but not significant. In any case, the
decline in TFP as a result of the financial crisis in 1997 and 1998, amount-
ing to more than 10 percent was not made up in the recovery period.
Table 6.9 Regression test on residual factor
Equation 4
Dependent variable
% change TFP
China dummy 0.501 1.037 Non-significant
Dummy 1997 ⫺1.109 0.871 Marginal
Dummy 2000–03 ⫺0.282 0.575 Non-significant
Income pc. ppp basis ⫺2.735E⫺06 3.500E⫺05 Non-significant
Industry share ⫺0.035 0.032 Marginal
Change of ind. share ⫺1.426 0.254 Significant
FDI/employment 2.163E⫺05 3.382E⫺05 Non-significant
Investment/GDP ⫺0.055 0.035 Marginal
% change exports 5.895 2.076 Significant
% change industrial output 49.774 5.729 Significant
Adjusted R2 0.683
5.2 Impact of Industrial Production, Exports and Investment
The question is whether other variables like increasing exports, share of

investment, or industrialization contribute to explaining the productivity
(TFP) residual. Our fundamental hypothesis is that much of East Asian
growth can be explained in terms of inputs and growth of production of
industrial products and particularly products directed at the export market
that involve greater productivity than traditional goods and services. We
have tested numerous variables that might explain the unexplained part of
growth of productivity in East Asia. In each case, we have included
dummy variables to account for the productivity changes associated with
the financial crisis. The computations are pooled time series/cross-section
regressions covering eight East Asian countries over the period 1987–2003.
The coefficients for the country dummies were not significantly different
from zero.
A number of variables that might relate to the residual were tested sepa-
rately and in combination. (Table 6.9). As anticipated, change in industrial
output and change in exports make significant contributions to residual
productivity change, 0.5 percent for each 1 percent increase in industrial
production and 0.05 percent for each 1 percent change in exports.
Investment change has already been taken into account as an input, prior
to the computation of the residual productivity change and it does not yield
a significant result in the form of investment as a share of GDP. Notable
are the results for industry as a share of GDP. The effects of the industry
share variable are negative, though with only marginal significance. This
suggests, that countries that already have a high level of industrialization
are less likely to show technical change-based productivity change. More
puzzling, however, is the result for change in the share of industry, since
here also there is a negative coefficient and it is statistically significant.
Given the level of industrial development, an increase in the share of indus-
try appears to have a negative impact on TFP. Again, this may reflect the
lessened potential for productivity change as high income countries shift
further resources into industry. On the other hand, the income per capita
variable does not show significant effects.
We also hypothesized that foreign direct investment would influence
changes in TFP. It is difficult, however, to measure FDI across the East
Asian countries since some, like Taipei, show negative values for FDI, while
others benefit from positive values. In any case, we were not able to show a
significant positive effect, though that may have been incorporated into the
significant effect of export growth, many of which are produced by firms
benefiting at some stage from foreign investment or management. The
dummy variable for 1998 is significantly negative, the dummy for 1997 is
only marginally significant. The dummy for China and the dummy for
2000–03 did not show significant effects.
The variables that showed statistically significant effects were combined
into the regression shown in Table 6.10 and the differences between actual
and predicted values are shown for the pre- and post-crisis periods in
Table 6.11. Much of the variation in TFP can be explained in terms of
growth of industrial output and exports (adjusted R-squared of 0.67) that
appear to be associated with high TFP growth in the East Asian countries.
Presumably, the resources employed in the new industries have a higher
level of productivity than that prevailing elsewhere in the economy, causing
TFP to increase with industrial and export growth.
The unexplained element of TFP growth (difference between actual TFP
growth and the estimated value) can be said to indicate pure technical
progress as well as interactions and possible increasing returns to scale. It
is relatively small, accounting for less than a third of the variance of calcu-
lated TFP and a much smaller share of the total variance of productivity,
for most countries throughout the entire estimation period. This is in
accord with the results noted in earlier growth accounting studies. While
there is no comprehensive evidence of a change in unexplained TFP
growth, some countries are clearly doing better recently in this respect than
Table 6.10 Basic equation explaining TFP
Equation 5
Dependent variable
% change TFP
Dummy 1997 ⫺1.237 0.828 Marginally significant
% change exports ($) ⫺6.458 1.945 Significant
% change industrial output 45.783 4.819 Significant
Change in industrial share ⫺1.379 0.234 Significant
Constant ⫺0.833 0.405
Adjusted R2 0.674
Table 6.11 Difference between actual and estimated TFP, pre-crisis and
post-crisis
1987–91* 1992–96* 1997 1998 1999 2000–03 Difference*
China, 0.2 1.9 0.9 4.4 0.1 ⫺0.1 ⫺1.9
Rep. of
Hong Kong, 1.5 0.9 1.8 0.2 1.0 1.5 0.5
China
Korea, 0.2 ⫺1.3 ⫺1.5 1.8 2.3 ⫺0.6 0.7
Rep. of
Indonesia 1.6 ⫺0.4 1.0 ⫺6.7 ⫺1.5 1.3 1.8
Malaysia 0.1 ⫺2.2 ⫺0.1 ⫺4.0 0.7 ⫺0.5 1.7
Philippines 0.0 ⫺1.8 0.8 0.8 1.2 ⫺0.9 1.0
Singapore ⫺0.7 ⫺0.1 ⫺0.4 ⫺0.3 0.5 ⫺1.8 ⫺1.7
Thailand 0.3 ⫺0.4 ⫺4.8 ⫺1.1 2.8 1.5 1.9
Taipei 2.2 0.8 2.3 4.8 0.0 ⫺0.2 ⫺1.0
Note: * Omitting outlier values for 1990 (China) and 1995 (Korea) (2000–03) minus
(1992–96).
others. China shows almost zero unexplained TFP since the mid-1990s,
meaning that all of its growth reflects either an increase in inputs or is
closely associated with the expansion of industrial production and exports.
Of course, one possible explanation of this phenomenon may lie in the fact
that China has proven, in a systemic or institutional sense, to be a much
sought-after destination for foreign direct investment and the resulting
relatively large FDI inflows may have reduced China’s imperative to worry
about ways to improve productivity per se. Again, while the financial crisis
had significant negative impacts on productivity growth, there is little evi-
dence to suggest that the unexplained component of productivity growth is
structurally different in the post-crisis period than in earlier years; it is
lower in Taiwan and Singapore, where it may reflect the increasing difficulty
of these relatively advanced countries in catching up with advanced tech-
nology. It is also lower in China, where it may reflect statistical problems,
but it is higher in Indonesia, Malaysia and Thailand.
6 CONCLUDING REMARKS
The object of this study was to examine the effect of the East Asian Crisis
on the growth of productivity change in East Asia. Our approach has been
one based on measurement of inputs and explanation of total factor pro-
ductivity (TFP), associating it with the cyclical impact of the financial crisis
and growth in industrial production and exports. These considerations had
statistically significant explanatory power. The coefficient estimates for
other measured factors, including FDI and investment, did not show up
statistically significant in the regressions.
As has also been shown by earlier empirical work, our calculations
suggest that rapid growth in East Asia depends greatly on the expansion of
inputs of labor and capital. Productivity growth also appears to be closely
related to expansion of industrial production and exports and to vary con-
siderably with cycles in business activity. A relatively small share of TFP
remains unexplained and that may be associated with increasing returns
and technological change.
The impact of the financial crisis is clearly apparent in 1997 and 1998.
The loss in productivity growth associated with this period is not made up
in later years. In the post-crisis years, investment is lower as a share of GDP
and growth of total inputs is slower, except in the Philippines, than before
the crisis. Moreover, in China, Singapore and Taipei, TFP is substantially
lower in the post-crisis than in the pre-crisis period. In these countries, this
reduction is also apparent in the unexplained part of TFP, though in other
parts of East Asia, the unexplained part of TFP is somewhat higher post-
crisis than before. Lower growth of unexplained TFP in the most advanced
countries, Taipei and Singapore, suggests that they are finding it increas-
ingly difficult to shift their production functions to take advantage of new
technology. The reduced unexplained TFP element in China, which reflects
a downward trend from an extraordinarily high pre-crisis rate of unex-
plained productivity growth, calls for further study.
NOTES
1. The subtitle of the Collins and Bosworth paper.

2. For instance, with particular reference to Singapore, Krugman comments that the
country’s rapid growth was a matter of ‘perspiration rather than inspiration’. He then
went on to suggest that diminishing returns will limit further growth as in the Soviet
Union!
3. Since Collins and Bosworth work with data through 1994, before the East Asian Crisis,
this may be understandable, but Crafts (1999) who is specifically concerned with before
and after the crisis periods does not take this factor into account either.
4. More recent work also includes a factor for the development of human capital (H), that
is, the production function assumes Y ⫽ f(K, H, L). The difficulty with introducing H here
is that, with the exception of high-tech industries, the human capital constraint has not
been effective in East Asia as workers have moved from simple agricultural work to simple
industrial tasks. Moreover, the aggregate level of educational attainment, years of school-
ing, as used by Collins and Bosworth, provides little information on the current growth of
productive abilities.
5. The depreciation rate for capital is another necessary assumption. We have assumed that
real depreciation of capital is at an annual rate of 0.05. The World Bank’s capital stock
statistics assume a depreciation rate of 0.04.
6. A sensitivity analysis postulating various values of the relevant coefficients showed that
choice of the coefficients did not greatly affect the pattern of values produced.
7. Modern growth theory would also note the interaction between technical change
and the rate of growth as well as the possibility of increasing returns to scale (Romer,
1986).
REFERENCES
Adams, F.G. (1998), ‘The East Asian Development Ladder: Virtuous Circles and
Linkages in East Asian Economic Development’, in F.G. Adams and
S. Ichimura, (eds), East Asian Development, Westport, CN: Praeger, pp. 3–18.
Akamatsu, K. (1961), ‘A Theory of Unbalanced Growth in the World Economy’,
Weltwirtschaftliches Archiv, 86, 195–213.
Akamatsu, K. (1962), ‘Historical Pattern of Economic Growth in Developing
Countries’, The Developing Economies, 1, 3–25.
Collins, Susan M. and Barry P. Bosworth (1996), ‘Economic Growth in East Asia:
Accumulation Versus Assimilation’ Brookings Papers, Washington: Brookings
Institution, http://www.brookings.edu/=views/articles/collins/19960828.htm.
Corsetti, G., P. Pesenti and N. Roubini (1998a), ‘What Caused the Asian Currency
and Financial Crisis? Part I: A Macroeconomic Overview’, NBER Working
Paper, W 6833, Cambridge, MA: NBER.
Corsetti, G., P. Pesenti and N. Roubini (1998b), ‘What Caused the Asian Currency
and Financial Crisis? Part II: The Policy Debate’, NBER Working Paper, W 6834,
Cambridge, MA: NBER.
Crafts, Nicholas (1999), ‘East Asian Growth Before and After the Crisis’, IMF Staff
Papers, 46 (2) 139–66.
Denison, Edward F. (1974), Accounting for United States Economic Growth,
1929–1969, Washington: Brookings.
Herrick, Bruce and Charles P. Kindleberger (1983), Economic Development,
New York: McGraw-Hill.
Kojima, K. (2000) ‘The “Flying Geese” Model of Asian Economic Development:

Origin, Theoretical Extensions, and Regional Policy Implications’, Journal of
Asian Economics, 11, 375–401.
Krugman, Paul (1994), ‘The Myth of East Asia’s Miracle’, Foreign Affairs, 73,
62–78.
Lau, Lawrence (1998), ‘The Sources of East Asian Economic Growth’, in
F.G. Adams and S. Ichimura (eds), East Asian Development, Westport: Praeger,
pp. 41–68.
Radelet, Steven and Jeffrey D. Sachs (1998), ‘The East Asian Financial Crisis:
Diagnosis, Remedies, Prospects’, Brookings Papers on Economic Activity, 1, 1–90.
Romer, Paul (1986), ‘Increasing Returns and Long Run Growth’, Journal of
Political, Economy, 94 (5), 1002–37.
Solow, Robert M. (1956), ‘A Contribution to the Theory of Economic Growth’,
Review of Economics and Statistics, 70, 65–94.
Timmer, M.R., and A. Szirmai (1997), ‘Growth and Divergence in Manufacturing
Performance in South and East Asia’, Discussion paper GD37, Groningen:
University of Groningen Growth Center.
World Bank (1993), The East Asian Miracle: Economic Growth and Public Policy,
New York: Oxford University Press.
Young, A. (1995), ‘The Tyranny of Numbers: Confronting the Statistical Realities
of the East Asian Growth Experience’, Journal of Political Economy, 110,
641–80.
7. What really happened to Thai wage
rates during the East Asian
Financial Crisis?*
Jere R. Behrman, Anil B. Deolalikar and
Pranee Tinakorn
1 INTRODUCTION
For most members of developing countries that experience negative macro

shocks, such as the East Asian Financial Crisis of the late 1990s or the ‘Lost
Decade’ of the 1980s in much of Latin America and Africa, the most
important manifestations of the macro shocks are through the labor
market. This is the case because the most important earning asset of most
members of such societies – particularly for the poorer members of these
societies – is their labor. Therefore, it is important to understand the effects
of economic crises on the price of that asset, namely the real wage rate.
What happens to this price is important because it directly affects marginal
time allocation decisions, directly affects the income of many members of
society, and is an important consideration for the design of effective anti-
poverty and safety net policies. In this study we explore what happened to
real wage rates for the particular case of Thailand during the East Asian
Financial Crisis. Thailand was the first of the Asian developing economies
to experience the financial crisis of the second half of the 1990s.1 Indeed,
the East Asian Financial Crisis is often said to have begun with events in
Thailand in 1997. In part for this reason, the Thai experience has been
viewed as particularly interesting and has been investigated in a number of
studies. While our empirical analysis is for Thailand, the questions we
address about what happened to real wage rates due to the macro crisis are
of relevance for other developing countries that experience such macro
shocks.
Most of the previous literature claims that real wage rates declined a fair
amount in Thailand due to the crisis, as well as in various degrees in the
other Asian economies most afflicted by the crisis. The most prominent
and visible of these studies for Thailand is World Bank (2000). This study
221
uses a methodology that other analysts have used as well for other studies
of the Thai experience and of the experience of other countries (for
example, Betcherman and Islam, 2000; also see Fallon and Lucas, 2000;
Horton and Mazumdar, 1999; Mahmoud and Aryah, 1999). In this
approach, the percentage change in the number of employed workers
between the pre- and post-crisis period2 is subtracted from the percentage
change in aggregate real wage earnings between the pre- and post-crisis
period to obtain the percentage change in the real wage rate due to the
crisis. Using this methodology, the World Bank estimates that the Thai real
wage rate fell by 4.6 percent and Betcherman and Islam estimate that real
wages fell by –1.1 percent in Malaysia, –9.3 percent in Korea and –41.0
percent in Indonesia (Table 7.1). But we argue that the methodology used
in these and other studies is subject to at least four possibly important lim-
itations:
1. First, it assumes that percentage changes in overall employment are the

same as percentage changes in the employment of wage workers.
2. Second, it attributes changes in hours worked to changes in real wage
rates.
3. Third, rather than weighting each wage worker equally, it weights wage
rates differentially across wage workers depending on how many hours
they work.
4. Fourth, it ignores important compositional changes in wage recipients
with regard to gender, age and schooling, as well as other possibly
related factors such as ability and motivation.
Table 7.1 Percentage year-on-year changes in employment, real wages,

real wage labor earnings and GDP, 1997–98 (% ratio to
% change in GDP in parentheses)
Country % Change In
Employment Real wages Real wage labor GDP
earnings
Indonesia 2.7 (⫺20) ⫺41.0 (299) ⫺38.3 (280) ⫺13.7
South Korea ⫺5.3 (91) ⫺9.3 (160) ⫺14.6 (252) ⫺5.8
Malaysia ⫺2.7 (36) ⫺1.1 (15) ⫺3.8 (51) ⫺7.5
Thailand ⫺3.0 (30) ⫺4.6 (46) ⫺7.6 (76) ⫺10.0
Sources: World Bank (2000, Table 1). Figures for Thailand calculated from Labor Force
Survey data, and represent data for the first and third quarters of 1998. Figures for other
countries obtained from Betcherman and Islam (2000), and refer to the full 1998 year.
What really happened to Thai wage rates? 223
While these may seem to be dry technical points, they may make a consid-
erable difference in our understanding of the answer to the important
empirical question of what did happen to real wage rates and what the
implications are for policy.
We demonstrate that, indeed, in the Thai case, these four methodologi-
cal aspects of the approach used by the World Bank and others each make
differences in estimates of what happened to real wage rates. These
differences are particularly large in the Thai case for the second and the
fourth of these methodological limitations, resulting in absolute biases in
measured real wage declines of 5.5 and 9.8 percentage points, respectively.
In the particular case of Thailand, these biases are partially offsetting.
Even so, we estimate that for the combination of these four reasons, the
World Bank (2000) overestimates the decline in real wage rates by about 40
percent. Our preferred estimate, using the same data and, except for these
four aspects of methodology, the same assumptions as the World Bank is
that the Thai aggregate real wage rate declined by 7.8 percent due to the
crisis.
In our judgment, these individual biases and their combined impact on
the estimated decline in real wage rates due to the crisis are NOT small
differences that should be relegated to a footnote in a country evaluation of
the impact of the crisis. Such differences might significantly affect how ana-
lysts think about the impact of the crisis and what policy responses seem
appropriate. Therefore, we claim that how one calculates the real wage rate
is important in the Thai case, and quite possibly in other cases.
Finally, the method that we use to obtain our preferred estimates also
permits us to investigate the frequent claim that the more vulnerable and
poorer members of the work force suffered the largest real wage declines in
the crisis: females, those with less schooling, and the young and the old. Our
examination of this question suggests that the usual characterization is par-
tially, but only partially, correct. Those who had limited schooling and the
young did experience relative declines in real wage rates. But so did those
with university schooling. And females and older wage earners experienced
relative increases in their real wages in comparison with others.
This chapter is organized as follows. Section 2 reviews the previous
studies on what happened to aggregate Thai real wage rates due to the crisis,
with emphasis on World Bank (2000) because that is the most prominent
study and the one that is clearest about the procedures followed to obtain
the estimates. Section 3 presents estimates of the extent of the four biases
noted above in World Bank (2000) estimates for the Thai case, presents our
preferred estimates of the change in aggregate real wage rates due to the
financial crisis, and presents our estimates of what happened to real wage
rates for some of the more vulnerable workers. Section 4 concludes.
2 SUMMARY OF PREVIOUS LITERATURE ON

REAL WAGE DECLINES DUE TO THE THAI
CRISIS
Studies by other authors on the impact of the crisis on Thai labor markets
have suggested – and in most cases stated explicitly – that wage rates fell
considerably in the immediate aftermath of the crisis, though in some cases
there is some ambiguity about whether wage rates or wage earnings are
being discussed. We begin by summarizing the results of World Bank
(2000) because, as noted, that is the most visible and the clearest on the
methodology used among the available studies. We then summarize more
briefly five other studies.
2.1 World Bank (2000) Estimates of Thai Real Wage Declines
As noted in the introduction, the World Bank (2000) estimates that the Thai
real wage rate fell by 4.6 percent due to the crisis by subtracting the per-
centage change in the number of employed workers between the pre- and
post-crisis period from the percentage change in aggregate real wage earn-
ings between the pre- and post-crisis period. The pre-crisis period that is
used is the average over three Labor Force Survey (LFS) rounds prior to the
crisis (that is, the first and third quarters of 1996 and the first quarter of
1997). The post-crisis period that is used is the average over three survey
rounds after the initiation of the crisis (that is, the first and third quarters
of 1998 and the first quarter of 1999). Both periods cover five quarters, or
1.25 years, ending about six months before and starting about six months
after, respectively, the July 1997 date that is often referred to as the start of
the crisis period. For comparability, both are for the same duration and
both include the same combination of seasonal peak and slack LFSs.
Monthly wage earnings were calculated from wages reported for other
reporting periods (for example, hourly, daily, weekly) by multiplying by
standard factors that assume full-time work.3 Then other monetary
benefits were added and the resulting total monthly wage earnings were
deflated by the consumer price index (CPI) to obtain the total real monthly
wage earnings.
2.2 Other Studies on Thai Real Wage Rate Declines in Response

to the Crisis
We summarize here more briefly several other studies. They all are based
on pre- and post-crisis comparisons with LFS data and they all convey the
impression that the empirical evidence suggests that Thai real wage rates
declined due to the crisis. But they present a range of estimates, refer to
somewhat different periods and in most cases are not very clear about their
procedures. So it is not possible to determine why they differ in some
respects from the World Bank (2000) estimates, nor whether they are
subject to some or all of the same biases.
1. Kittiprapas and Intaravitak (2000, Appendix 1) define real monthly

wages by translating wages reported for other periods (for example,
hourly, daily, weekly) into monthly wages using: (a) hourly wages ⫻
hours worked per week ⫻ 30/7, (b) daily wages ⫻ days of work per
week ⫻ 30/7, and (c) weekly wage ⫻ 30/7, respectively for the payment
periods used in the reports in the LFS. In all cases, the monetary value
of monthly bonuses plus overtime plus other monetary fringe benefits
then were added, and the total deflated by the CPI. Their calculations
suggest significant declines of 4.6 percent and 3.9 percent in real wage
earnings for Rounds 1 and 3, respectively, of the LFS in 1998 as com-
pared with 1997.
2. Mahmood and Aryah (1999) state: ‘The crisis lowered real wage growth
in 1997 to 1 percent, and then cut the real wage by over 7 percent in
1998. The cut in the real wage is expected to persist over 1999 [p. 4] . . .
Again Labor Force data shows that between February 1997 and
February 1998, the nominal wage dropped by 6 percent, while the real
wage rate dropped by 4 percent’ [p. 16].4 We note that it is not clear
whether the first statement refers to wage earnings or wage rates, but
the second explicitly refers to wage rates.
3. Siamwalla (2000) states: ‘[N]ominal wages among the males in the
educated group fell by about 6–8 percent between August 1997 and
August 1998, and among females by about 4–7 percent. Among the less
educated, the fall was less than 2 percent’ [p. 27]. The fall in real wages,
of course, would be almost 8 percent greater (given the 7.8 percent
increase in the CPI between 1997 and 1998).
4. Kakwani (1998) presents in Table 6 in his paper, data on ‘wage earn-
ings’ that indicate that they fell from 6273 baht per month in the first
quarter of 1997 to 5825 baht per month in the first quarter of 1998, a
decline of 7.1 percent.
5. Pongsapich and Brimble (1999, section 3.2) state: ‘Wages fell back to
near 1996 levels. The total nominal wage bill in the private sector
in 1998 was about 1 1/4 billion baht lower than in 1997, falling back
to little more than the levels of 1996. The loss of real income
per income earner reached 21 percent by the wet season of 1998,
declining further from the fall of 17 percent in the previous dry
season.’
3 RE-EXAMINATIONS OF THAI REAL WAGE RATE

CHANGES DUE TO CRISIS
We have questions, as we note above, about four aspects of the World Bank
(2000) estimates of real wage rate decline of 4.6 percent due to the crisis. In
this section we first consider each of these four aspects in turn and what
they imply for the estimated change in real wages. We then consider what
our estimates imply for what happened to the real wage rates for selected
‘more vulnerable’ groups.
3.1 Possible Biases in the World Bank (2000) Procedure and Our
Preferred Estimates of Aggregate Thai Real Wage Rate Changes
due to Crisis
The essence of the World Bank procedure about which we have questions
can be summarized as follows. The World Bank calculates the change in
real wage rates between the pre- and post-crisis periods as:
% change in Wmw ⫽ % change in Ym ⫺ % change in N, (7.1)
where Wmw is the real wage rate (with the superscript ‘w’ referring to the
World Bank 2000 estimate), Ym is the total real wage earnings based on all
wage earners (⫽ ⌺Yi, where i is summed over all wage recipients), Nm is the
total number of employees, and the subscript ‘m’ denotes means). Relation
(7.1) is consistent with the real wage rate being defined in each period as:
Wwm ⫽ Ym Ⲑ(HmE) ⫽ Ym Ⲑ(HmaN), (7.2)
where Hm is the average hours worked per pay period by wage recipients, E
is the number of wage recipients in each period and ‘a’ is the ratio between
the number of wage recipients in each period E and the total number of
employees in that period N, and both Hm and a are assumed to be constant
across periods.
We argue that the best available estimate is to calculate for each period
the average wage rate for individuals who are observably the same as:5
Wm ⫽ (⌺Wi )ⲐE ⫽ (⌺(Yi ⲐHi )) ⲐE, (7.3)
where Wi is the real hourly wage rate for the i-th wage recipient that is cal-
culated by dividing real wage earnings Yi for each wage recipient by the
hours worked per pay period Hi for each wage recipient. We then calculate
the percentage change in Wm between the pre- and the post-crisis periods.
Before we turn to why we think that our procedure is preferable and the
empirical importance of each of the four differences between what we argue
is the preferred procedure and the procedure that the World Bank utilized,
we wish to emphasize that we are not disagreeing with a number of other
important aspects of the World Bank estimates – the definition of the pre-
and post-crisis periods, the use of the LFS and of particular rounds of the
LFS, the deflation procedure, the definition of real wage compensation, the
definition of wage recipients, the definition of who is employed – so we are
following exactly the same procedures as the World Bank with regard to all
these aspects of the estimates. Therefore, the differences on which we focus
below are NOT due to other differences in the calculations.
3.1.1 Number of wage recipients are a fixed share of total employment

The World Bank use of the total numbers employed (N) instead of the total
number of wage recipients (E) in relation (7.1) is fine as long as the share
of wage recipients among all those employed (the parameter ‘a’ in relation
(7.2)) is constant across periods. The problem is that share may not be con-
stant between the pre- and the post-crisis period. The World Bank (2000,
Table 1) estimates that employment declined by 3 percent, but the compo-
sition of employment shifted from wage to non-wage workers, so the wage
employment declined by 5.5 percent. This by itself causes a bias of 2.5 per-
centage points (⫽ 5.5 – 3.0) in the World Bank estimates. That is, if this bias
alone is corrected, the World Bank estimate of a decline in the real wage
rate of 4.6 percent would change to a decline of 2.1 percent.
3.1.2 Hours worked are constant

The World Bank treatment (or non-treatment) of hours worked is fine if
average hours worked did not change between the pre- and the post-crisis
period. The problem is that mean hours worked did change, and in fact
declined by 5.5 percent. This by itself causes a bias of 5.5 percentage points
in the World Bank estimates. That is, if this bias alone is corrected,
the World Bank estimate of a decline in the real wage rate of 4.6 percent
would change to an increase of 0.9 percent. If both of these first two biases
are corrected, the World Bank estimate would change to an increase of
3.4 percent.
3.1.3 Average of wage rates across wage recipients with these individuals
weighted proportional to their hours worked
A more subtle difference between our preferred approach and that of the
World Bank is how individuals are weighted when wages or wage rates are
summed across individuals in relation (7.2) versus relation (7.3). The World
Bank effectively weights wage recipients proportionately to their hours
worked in relation (7.2) while we think that it is more appropriate to weight

individuals equally to obtain the mean wage rates across wage recipients in
relation (7.3). We think that for characterizing the changes in the average
of the wage rates faced by wage recipients, each individual wage recipient
should be weighted the same. We also note that treating each individual as
an observation (rather than each hour worked as is implicit in relation (7.2))
is consistent with standard practices for the estimation of micro wage rate
and other labor market relations (some of which we utilize below). Our esti-
mate is that this by itself causes a bias of 1.4 percentage points in the World
Bank estimates. That is, if this bias alone is corrected, the World Bank esti-
mate of a decline in the real wage rate of 4.6 percent would change to a
decrease of 6.0 percent. If all three of these first three biases are corrected,
the World Bank estimate would change to an increase of 2.0 percent.
The result from correcting the first three sources of biases (or the first
two, for that matter) that the change in real hourly wage compensation rate
due to the financial crisis was positive is striking and puzzling. It contrasts
sharply with previous suggestions that are summarized in sections 1 and 2
that real wage rate declines were an important aspect of the adjustment
process, as well as with causal observations and the implications of most
models of labor market outcomes given a negative macro shock. We have
therefore examined our estimates very carefully. Figure 7.1 plots the real
hourly wage rate over the 1987–99 period for which we have comparable
data. Generally, this figure suggests a secular upward trend in real wage
rates that continued after the crisis and that is consistent with an increase
in the reported real wage rate after the initiation of the crisis, with the
exception of a blip in the third quarter of 1997 (the first quarter after the
usual dating of the start of the financial crisis).6 For the three included
survey periods subsequent to this blip, the real hourly compensation rate
was decreasing,7 but the average over these three periods exceeded the
average over comparable periods before the blip.
3.1.4 No change in composition of wage recipients between pre- and

post-crisis periods
The World Bank (2000) procedure compares the real wages obtained from
all wage recipients in the pre-crisis period with the real wages obtained by
all wage recipients in the post-crisis period. This comparison is fine if the
composition of wage recipients remained the same in both periods. But this
would be surprising. As noted above, the number of wage recipients
declined by 5.5 percent between the two periods. Standard analysis of
hiring and labor supply decisions suggests that this decline is likely to have
been selective, not random. It is possible, and even likely, that the workers
who left wage employment status tended to receive lower wages than others
38 38
36 36
34 34
In 1994 baht per hour
32 32
30 30
28 28
26 26
24 24
22 22
20 20
18 18
88 3
88 1
89 3
89 1
90 3
90 1
91 3
91 1
92 3
92 1
93 3
93 1
94 3
94 1
95 3
95 1
96 3
96 1
97 3
97 1
98 3
98 1
99 3
99 1
3
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
19 -Q
-Q
87
19
Note: No first round was conducted by the NSO in 1990 causing a gap in the plotted
section.
Figure 7.1 Real hourly wage rate, 1987–99
before leaving wage employment because they were less productive regard-
ing observed characteristics such as schooling or unobserved characteris-
tics such as innate ability and motivation. If so, it is possible that mean
wage rates for those in wage employment in the post-crisis period increased,
even though the post-crisis wage rates for those who maintained wage
employment through both periods hardly changed or even declined.
Table 7.2 summarizes relevant data for all workers and for subcategories
defined by three characteristics of workers that are observed in the data and
that are not likely to change (or at least not change much) due to labor
market conditions: gender, age and schooling.8 The first column gives the
average wage rate for the indicated subgroup as a percentage of the national
average wage rate for the pre-crisis period. These data permit the charac-
terization of which types of workers received relatively high wage rates
before the crisis. The next two columns include the percentage changes
between the pre- and the post-crisis periods for wage employment and
hours worked in wage employment.
Examination of this table suggests that wage employment shifted rela-
tively from females to males, from younger to older workers, and from
lower-schooled to higher-schooled individuals. Each of these three
shifts was from lower to higher real wage rate categories.9 This suggests that
Table 7.2 Wages as percentages of national average wages in the pre-crisis

period and percentage changes in wage employment and hours
worked in wage employment before and after July 1997 for
selected subgroups of workers
Subgroups Wage Rate as % Percentage Change Between

of Workers of National Pre- and Post-crisis in
Average Wage
Rate in Pre-crisis Wage Hours worked in
Period employment wage employment
All workers 100.0 ⫺5.5 ⫺5.4
By gender
Female 92.0 ⫺13.5 ⫺5.4
Male 105.3 ⫺4.0 ⫺5.4
By age
15–19-year-olds 47.6 ⫺29.1 ⫺4.6
20–24-year-olds 66.8 ⫺14.2 ⫺4.9
25–49-year-olds 110.2 ⫺0.6 ⫺5.4
50+ year-olds 139.0 ⫺0.4 ⫺5.4
By schooling
Less than primary 43.7 ⫺22.5 ⫺5.6
Primary 59.3 ⫺17.4 ⫺5.2
Secondary 99.5 14.8 ⫺4.0
Vocational 149.4 1.3 ⫺3.0
University 255.9 20.0 ⫺1.8
compositional changes may have been important and, because they involved
a shift from much lower wage rate categories to higher wage rate categories,
the failure to account for them in the World Bank estimates mean that the
estimated overall average real wage rate change is biased upwards.
Therefore, we obtain estimates that permit the control for changes in the
composition of wage recipients with respect to age, gender and schooling.
We estimate a ln real hourly wage rate equation, pooling individual data
from all six quarters covering the pre- and post-crisis periods. By construc-
tion, this approach does not have any of the first three problems in the
World Bank estimates that are discussed above. All coefficients of the wage
equation are allowed to vary across the pre- and post-crisis periods. Using
this equation, the actual change in real hourly ln wages (which would be the
percentage change in real hourly wages between the pre- and post-crisis
periods) can be decomposed into changes that might have occurred in the
absence of any compositional changes in wage employment and changes
Table 7.3 Decomposition of crisis-induced change in ln real hourly

wage rates
Change Value % Contribution

Changes due to varying coefficients (or wage returns)
Intercept ⫺0.133 ⫺701.2
Female 0.013 66.4
Primary schooling 0 0
Secondary schooling 0 0
Vocational training 0 0
University education ⫺0.011 ⫺56.1
Ages 20–24 0 0
Ages 24–49 0.037 196.2
Ages 50 & over 0.016 83.9
Subtotal ⫺0.078 ⫺410.8
Changes due to shifts in wage employment
Female ⫺0.005
Primary schooling ⫺0.027 ⫺142.8
Secondary schooling 0.031 163.6
Vocational training 0.004 19.4
University education 0.082 431.5
Ages 20–24 ⫺0.002 ⫺11.7
Ages 24–49 0.012 64.4
Ages 50 & over 0.002 11.5
Subtotal 0.097 510.8
Total explained change in ln wages 0.019 100.0
Note: See Appendix 7A, Table 7A.1 for the estimates of the underlying ln wage equation.
that occurred because of compositional shifts. The estimates of the ln wage

equation are presented in Appendix 7A, Table 7A.1, while the results of the
decomposition analysis are given in Table 7.3.10
These estimates suggest that, controlling for the compositional shifts in
wage employment, real hourly wage rates would have declined by 7.8
percent between the pre- and immediate post-crisis periods.11 This estimate
of a ‘no-compositional shift’ in the decline in real aggregate wage rates of
7.8 percent is strikingly different from the estimated increase in the aggre-
gate real wage rate of the World Bank estimate adjusted for the first three
biases discussed above. This comparison suggests that the World Bank esti-
mate is biased upwards by 9.8 percentage points due to the failure to control
for compositional shifts related to gender, age and schooling. This is a con-
siderable bias. Of course, the actual unadjusted World Bank estimate was
a decline of 4.6 percent because the biases due to the assumptions of no
change in the share of employees who were wage workers and of no change
in hours worked partially offset the biases due to the compositional shift
and the weighting by hours worked instead of by wage recipients. But even
with this serendipitous partial canceling of different biases, the estimate as
presented in World Bank (2000) understates by about 40 percent the real
wage rate decline that we obtain in our preferred estimate.
In summary, thus, our best estimate of how much real hourly wage rates
would have declined had there been no change in the composition of wage
recipients is 7.8 percent. Of course if the composition changed in terms of
unobserved variables, such as ability and motivation, in the same way that
it changed in terms of observable variables, such as schooling, as is sug-
gested by some micro empirical estimates for other countries, the true
decline for the same individuals controlling also for these unobserved
factors would have been greater than 7.8 percent.
3.2 Who Fared Relatively Poorly in Terms of Real Wage Rate Changes?
Many have claimed that the poorer and more vulnerable workers – females,
low-schooled, the young and the old – suffered most in the crisis (for
example, Horton and Mazumdar, 1999; Kittiprapas, 1999; Kittiprapas and
Intaravitak, 2000; Knowles et al., 1999; Pongsapich and Brimble, 1999;
Sussangkarn et al., 1999). The compositional shifts in wage employment
described near the end of section 3.1 are generally consistent with that
claim. But that does not necessarily mean that the poorer and most vul-
nerable suffered the largest relative real wage rate declines.
Table 7.4 presents some evidence on this question with reference to
gender, age and schooling. The first column gives the percentage change
between the pre- and post-crisis periods in real wage rates. The second and
third columns give regression estimates from the pooled data described for
the decomposition in section 3.1 (and presented in Appendix 7A, Table
7A.1) for the interaction between the post-crisis dummy variable and
gender, age and schooling (all relative, again, to males younger than 20 with
less than complete primary schooling). Column 2 includes coefficient esti-
mates for all of these characteristics; column 3 includes only those charac-
teristics that have statistically significant coefficient estimates at least at the
10 percent level in column 2.
The percentage changes in real hourly wage compensation rates in the
first column of Table 7.4 suggest that some of the groups that are often
characterized as ‘more vulnerable’ did fare relatively badly in terms of wage
rate changes. In particular, youth fared worse than prime-age adults. On the
other hand, some of the groups usually characterized as more vulnerable
fared relatively well: those with primary or less schooling (as well as those
Table 7.4 Percentage changes in real wage rates and coefficient estimates
for interactions between crisis dummy variable and different
groups in pooled real wage regression
Change Percentage Change Between Change in Real Wages

Pre- and Post-crisis Relative to Reference
Periods in Real Wage Group (estimated
Rates coefficients for
interactions with
crisis dummy)
By gender
Female 3.8 3.5* 0.030*
Male 1.2 – –
By age
15–19-year-olds ⫺4.6 B B
20–24-year-olds ⫺3.7 ⫺0.010
25–49-year-olds 0.0 0.046*
50+ year-olds 12.8 0.153* 0.161*
By schooling
Less than Primary ⫺5.8 B B
Primary ⫺4.7 0.027
Secondary ⫺9.5 0.006
Vocational ⫺4.6 0.025
University ⫺11.5 ⫺0.033* ⫺0.055*
Reference group (males ⫺0.146* ⫺0.133*
⬍ 20 years old, less
than primary school)
Notes: * Calculations from the LFS data tapes. Changes are defined between the period
before the crisis (first quarter of 1996, third quarter of 1996 and first quarter of 1997) to a
comparable period after the initiation of the crises (first quarter of 1998, third quarter of
1998, first quarter of 1999). Real hourly wage rate includes cash wages plus monetary
benefits, adjusted for the CPI, per hours worked. The second and third columns are from
regressions 2 and 3 in Appendix 7A, Table 7A.1. * implies significantly non-zero at least at
the 10 percent level. B means that included in reference group in bottom row.
with vocational schooling) had smaller declines than those with secondary
or university schooling, females had larger increases than males, and older
adults had much larger increases than any other age group. The first
column gives the gross associations without controlling for other charac-
teristics. The second and third columns give the changes for various groups
with multivariate controls for other changes. These multivariate estimates
again suggest that some of the more vulnerable groups fared relatively
poorly. Based on column 3, males under 20 years of age with less than
primary schooling had significant declines of 13.3 percent in wage rates.
But females had significantly greater increases than males by 3.0 percent-
age points and older adults had significantly and substantially greater
increases than any other age group. The only other significant effect is that
those with university schooling had declines that were 5.5 percentage points
greater than the declines experienced by the reference group (those with less
than completed primary school).
3 CONCLUSIONS
What happens to real wage rates at times of macro shocks is of consider-

able importance because of (1) the implications for purchasing power of
workers and the design of appropriate anti-poverty and social safety net
policies, (2) the impact on time allocations of workers, and (3) the implica-
tions for the extent to which labor market adjustments in response to the
shock occur through price rather than quantity effects, which in turn is
likely to affect how broadly the effects are spread and how quickly adjust-
ment occurs. Conventional wisdom is that real wage rates fell, in some cases
substantially, in Thailand and in the other Asian economies most hit by the
financial crisis of the late 1990s, as well as in response to many other nega-
tive macro shocks experienced in developing countries.
The pre–post comparison of earnings per worker used by the World Bank
(2000) and others to estimate the change in real wage rates due to the crisis
in Thailand (and in other countries – see Table 7.1) has four problems that
may result in substantial biases: (1) ignoring change in the share of wage
employees in total employment, (2) ignoring changes in hours worked in
wage employment, (3) weighting changes in real hourly wage rates by hours
worked rather than treating individuals equally, and (4) ignoring compos-
itional changes among wage recipients. We quantify the magnitude of these
biases for the Thai experience and find that they are considerable – in
absolute terms 2.5, 5.5, 1.3 and 9.8 percentage points, respectively. The
usual methods, thus, are likely to give a misleading picture of what hap-
pened to average real wage rates. These methods do not provide good guides
regarding the impact of real wage changes on purchasing power or time
allocations of workers, nor of the extent of price adjustments in labor
markets, nor do they inform policy choices well.
Our best estimate is that real wage rates declined by at least 7.8 percent
due to the crisis, which is probably an underestimate of the true decline
because of the probable composition changes with regard to unobserved
characteristics such as ability and motivation that in many studies are
found to be correlated with characteristics such as schooling. Though in the
World Bank (2000) estimate the biases mentioned above are serendipitously
partially offsetting, the severity of the impact of the crisis on declines in the
real wage rate is underestimated by 3.2 percentage points or by over 40
percent even if the added problem of unobservables is ignored. In order to
understand the impact of such crises and to inform policy-makers and
other interested parties, it is important that the informational basis and the
methodologies for estimating real wage rate changes are improved.
The best solution would be for longitudinal labor force data to be col-
lected as a matter of routine so that comparisons could be made of the
wage rates for the same individuals over time, thereby controlling for all
individual unobserved as well as observed characteristics. Such data are
collected in Labor Force Surveys for some developing countries (for
example, Brazil has a rolling panel), but not many. Until such longitudinal
surveys are available in other cases, a second-best solution is to follow the
methods in this chapter, particularly with regard to controlling for compo-
sitional changes with respect to observed characteristics such as age, gender
and schooling. An added by-product of such an approach is estimates of
the extent to which different groups identified by such characteristics fared
relatively poorly – which in the present case gives some, but limited, support
to the widely-held understanding that the more vulnerable fared worse –
and, in fact, females and older adults fared relatively well, while highly edu-
cated individuals fared relatively poorly.
NOTES
* This chapter builds upon work in Behrman et al. (2000) that was prepared for the World
Bank by the Thailand Development Research Institute (TDRI). The authors alone – and
not TDRI nor the World Bank – are responsible for all interpretations in this chapter.
The authors thank Dr Worawan Chandoevwit and Mr Rangsiman Kingkaew for
research and computational assistance on this project.
1. In 1997 speculation on currency devaluation intensified, official foreign reserves were
rapidly depleted, the currency was subsequently allowed to float and devalued consider-
ably, and over half of the finance companies were closed. The annual growth rate in real
GDP per capita declined to –2.3 percent in 1997 and to –11.4 percent for 1998 from
4.8 percent for 1996 and from an average of 7.2 percent for 1990–95. Real GDP per capita
relative to the underlying secular growth trend for the 1990s fell (from the peak in 1996)
by 5.2 percent in 1997, 20.2 percent in 1998, and 20.5 percent in 1999 (based on the data
for the 1990s as presented in Behrman and Tinakorn, 2000, Table 8; these estimates focus
on how much real GDP per capita differs from the secular trend each year, and, because
this secular trend is positive, incorporate the secular growth not realized in addition to
any decline in measured real GDP per capita; see also Kakwani and Pothong, 1998).
2. To avoid long and perhaps awkward terminology, the ‘post-crisis period’ is used here and
throughout this study to refer to what is really the ‘post-initiation-of-the-crisis period’
and not the period subsequent to the recovery from the crisis. That is, comparisons are
made between average real wage rates for three labor force surveys prior to the initiation
of the crisis in mid-1997 and for three labor force surveys subsequent to the initiation of
the crisis in mid-1997 (see section 2.1 below).
3. Specifically, hourly wages were multiplied by 208 work hours per month, daily wages
were multiplied by 26 days per month, and weekly wages were multiplied by 4.2 weeks
per month.
4. The source is given as ILO (1999), ‘Country Employment Policy Review for Thailand’.
Given that inflation was positive (5.6 percent in 1997 and 8.1 percent in 1998 according
to the CPI based on data from the Bank of Thailand, see Behrman et al. (2000, Table 2),
it is not clear how the real drop in percentage terms can be less than, rather than greater
than, the nominal drop in percentage terms.
5. As is standard in the literature, this procedure is using the average real wage per hour for
each wage recipient rather than the marginal real wage rate per hour. The latter concep-
tually is preferable at least for some uses, but it is not clear how to calculate the marginal
real wage rate per hour with information that is generally available in Labor Force
Surveys.
6. The 1997 third-quarter real wage rate was strikingly high: 23.5 percent above the post-
crisis average and 38.2 percent above the pre-crisis average. It should be emphasized,
however, that if there is some anomaly for the 1997 third-quarter report, that does not
affect our, nor the World Bank’s (2000), pre–post comparisons because that period is not
used for these comparisons.
7. The real wage rates declined between rounds within the post-crisis period: by 14.1
percent from 1997-Q3 to 1998-Q1, by 2.8 percent from 1998-Q1 to 1998-Q3, and by 7.7
percent from 1998-Q3 to 1999-Q1.
8. Schooling may be affected a little by labor market conditions for those near the margin
of ending their schooling, but not much for older individuals in the prime working age
ranges. We do not include in this table characteristics such as location, migration status,
work status and occupation because these are likely to be affected by current and recent
labor market conditions.
9. The third column indicates a further shift in average hours worked per worker in wage
employment by schooling levels that reinforces this tendency, though no such shift in
hours worked per wage worker by gender and a small partially offsetting shift in hours
worked per wage worker by age groups. This shift in hours worked does not affect the
estimate of the change in real wage rates if each wage recipient is weighted equally as we
argue is preferable. But it affects the estimate if wage recipients are weighted by hours
worked as in World Bank (2000).
10. This type of decomposition is different in some basic respects from the racial and gender
decompositions of wage differentials along the lines pioneered by Oaxaca (1973) and
Blinder (1973) and subsequently extended and refined by others such as Jones (1983),
Cotton (1988) and Neumark (1988). In the Oaxaca–Blinder tradition, for example, the
comparison is between two exclusive groups at the same point of time (not two overlap-
ping groups in different time periods as in our case) and it makes a difference whether
the estimates for one group or the other is used as the base for making the decomposi-
tion. Not only does the decomposition change depending on which estimates are used,
but in the counterfactual in which ‘discrimination’ was changed, the relevant coefficients
might change. In the present case we are merely describing the changes from the pre-
crisis to the post-crisis periods, not attempting to indicate what would happen with
movement in the opposite direction or with some counterfactual. Our procedure is
equivalent, of course, to estimating separate ln real wage rate relations for the pre- and
post-crisis period and comparing the coefficient estimates.
11. On the other hand, had real wage rates for every population subgroup remained
unchanged during the crisis, the observed average wage rate would still have shown an
increase of 9.7 percent. Most of this increase (8.2 out of 9.7, or 85 percent) would have
occurred because of a shift in wage employment from lower-schooled to university-
educated individuals (who typically earn substantially more). The difference between the
increase of 9.7 percent (with only real wage rates unchanged) and the decline of 7.8
percent (with only the composition changed) is basically the same as the 2.0 percent
increase with no concern about composition and with the other three adjustments in
place.
REFERENCES
Behrman, Jere R. and Pranee Tinakorn (2000), ‘The Surprisingly Limited Impact
of the Thai Crisis on Labor, Including on Many Allegedly “More Vulnerable”
Workers’, Indian Journal of Labour Economics, 43 (3) (October), 513–44.
Behrman, Jere R., Anil B. Deolalikar and Pranee Tinakorn with Worawan
Chandoevwit (2000), The Effects of the Thai Economic Crisis and of Thai Labor
Market Policies on Labor Market Outcomes, Bangkok: Thailand Development
Research Institute.
Betcherman, G. and R. Islam (2000), ‘East Asian Labor Markets and the Crisis:
Impacts, Responses, and Lessons’, Synthesis paper prepared for the World
Bank/ILO/Japanese Ministry of Labor/Japan Institute of Labor seminar on The
Economic Crisis, Employment, and the Labor Market in East and South-East
Asia, Tokyo, February.
Blinder, Alan (1973), ‘Wage Discrimination: Reduced Form and Structural
Estimates’, Journal of Human Resources, 8 (Fall), 436–55.
Cotton, Jeremiah (1988), ‘On the Decomposition of Wage Differentials’, The
Review of Economics and Statistics, 70 (2), 236–43.
Fallon, Peter R. and Robert E.B. Lucas (2000), ‘The Impact of Financial Crises on
Labor Markets, Household Incomes and Poverty: A Review of the Evidence’,
Discussion Paper Number 103, Boston: Boston University, Institute for Economic
Development.
Horton, S. and D. Mazumdar (1999), ‘Vulnerable Groups and the Labor Market:
The Aftermath of the Asian Financial Crisis’, Paper prepared for the World
Bank/ILO/Japanese Ministry of Labor/Japan Institute of Labor Seminar on
Economic Crisis, Employment, and the Labor Market in East and South-east
Asia, Tokyo, November.
Jones, F.L. (1983), ‘On Decomposing the Wage Gap: A Critical Comment on
Blinder’s Method’, Journal of Human Resources, 18 (1), 126–30.
Kakwani, N. (1998), ‘Impact of Economic Crisis on Employment, Unemployment
and Real Income’, Bangkok: National Economic and Social Development
Board, mimeo.
Kakwani, N. and Jaroenjit Pothong (1998), ‘Impact of Economic Crisis on the
Standard of Living in Thailand’, Indicators of Well-Being and Policy Analysis, 2
(4) (October), 1–20.
Kittiprapas, Sauwalak (1999), ‘Social Impacts of Thai Economic Crisis’, in
Chalongphob Sussangkarn, Frank Flatters and Sauwalak Kittiprapas (eds),
Social Impacts of the Asian Economic Crisis in Thailand, Indonesia, Malaysia and
the Philippines, Bangkok: Thailand Development Research Institute.
Kittiprapas, Sauwalak and Chedtha Intaravitak (2000), ‘Social Responses to
Economic Crisis: Some Evidence during 1997–99 of Thai Crisis’, Bangkok:
Thailand Development Research Institute, mimeo.
Knowles, James C., Ernesto M. Pernia and Mary Racelis (1999), ‘Social
Consequences of the Financial Crisis in Asia: The Deeper Crisis’, Manila: Asian
Development Bank, mimeo.
Mahmoud, Moazam and Gosah Aryah (1999), ‘An Appraisal of the Labor Market
in the Context of the Macro Economy: Growth, Crisis and Competitiveness’,
Country paper prepared for the World Bank/ILO/Japanese Ministry of
Labor/Japan Institute of Labor seminar on The Economic Crisis, Employment,
and the Labor Market in East and South-East Asia, Tokyo, 13–15 October.
Neumark, David (1988), ‘Employers’ Discriminatory Behavior and the Estimation

of Wage Discrimination’, Journal of Human Resources, 23 (3), 279–95.
Oaxaca, Ronald (1973), ‘Male–Female Wage Differentials in Urban Labor
Markets’, International Economic Review, 14 (3) (October), 693–710.
Pongsapich, Amara and Peter Brimble (1999), ‘Assessing the Social Impacts of the
Financial Crisis in Thailand’, Bangkok: Chulalongkorn University and The
Brooker Group, prepared for the Asian Development Bank, Regional Technical
Assistance, mimeo.
Siamwalla, Ammar (2000), ‘Anatomy of the Thai Economic Crisis’, in Peter C.
Warr, (ed.) Thailand Beyond the Crisis, London: Routledge.
Sussangkarn, Chalongphob, Frank Flatters and Sauwalak Kittiprapas (eds) (1999),
Social Impacts of the Asian Economic Crisis in Thailand, Indonesia, Malaysia and
the Philippines, Bangkok: Thailand Development Research Institute.
World Bank (2000), Thai Workers and the Crisis, Bangkok: Social Monitor 4.
APPENDIX 7A
Table 7A.1 Estimated real hourly ln wage functions, using pooled data for six quarters (1Q96, 3Q96, 1Q97, 1Q98,
3Q98 and 1Q99)
Independent Variable Regression 1 Regression 2 Regression 3 Sample Means

Parameter T-ratio Parameter T-ratio Parameter T-ratio Pre-crisis Post-crisis
Whether female ⫺0.193 ⫺45.2 ⫺0.208 ⫺34.6 ⫺0.208 ⫺9.7 0.421 0.398
Whether primary schooled 0.348 36.6 0.334 24.5 0.346 31.5 0.585 0.507
Whether secondary schooled 0.822 78.8 0.820 54.9 0.821 87.6 0.172 0.210
Whether having vocational/ 1.229 92.9 1.215 66.6 1.226 114.5 0.044 0.047
technical training
Whether having university education 1.740 168.2 1.758 118.4 1.768 156.3 0.148 0.194
Whether aged 20–24 years 0.133 15.9 0.139 12.4 0.135 5.5 0.178 0.162
239
Whether aged 24–49 years 0.374 50.3 0.354 36.2 0.350 12.6 0.642 0.677
Whether aged 50 years and over 0.571 47.7 0.498 30.1 0.495 15.3 0.095 0.099
Dummy for post-crisis period ⫺0.078 ⫺18.6 ⫺0.146 ⫺6.2 ⫺0.133 ⫺5.9 0 1
Crisis ⫻ Female 0.030 3.5 0.030 1.9
Crisis ⫻ Primary schooling 0.027 1.4
Crisis ⫻ Secondary schooling 0.006 0.3
Crisis ⫻ Vocational education 0.025 1.0
Crisis ⫻ University education ⫺0.033 ⫺1.7 ⫺0.055 ⫺8.5
Crisis ⫻ 20–24 years ⫺0.010 ⫺0.6
Crisis ⫻ 24–49 years 0.046 3.1 0.055 9.6
Crisis ⫻ 50 years and over 0.153 6.4 0.161 17.7
Intercept 2.372 198.6 2.403 144.7 2.397 316.1
Mean of dependent variable 3.269 3.289
R-square 0.505 0.506 0.506
F-statistic 8179 4371 17 570
Number of observations 223 484 223 484 223 484
8. Exchange rate or wage changes in
international adjustment? Japan and
China versus the United States
Ronald I. McKinnon1
Seldom have the pages of the financial press in Europe and America been
so full of grave editorializing on the need for a major depreciation of the
dollar to correct the ‘unsustainable’ current account and trade deficits of
the United States. Much of this international moralizing directs the high-
growth East Asian countries to stop pegging their currencies to the dollar –
or, in China’s case, to allow a large appreciation of the renminbi before
moving to unrestricted floating. The critics’ message has two facets.
First, in order to reduce East Asia’s large trade surpluses and thereby
reduce America’s even larger trade deficit, US and European critics suggest
that Asian governments should let their currencies appreciate discretely
against the dollar. For example, Fred Bergsten, Morris Goldstein, Nicholas
Lardy and Michael Mussa from the Institute for International Economics
(IIE) in Washington DC, all suggest that an immediate 20 to 25 percent
appreciation of the renminbi is warranted (Bergsten et al., 2005). However,
they provide no suitable conceptual model – let alone econometric evi-
dence – that this would significantly reduce China’s trade surplus with the
United States.
Second, because of the very high ongoing productivity growth in some
East Asian countries (notably China) relative to that in the United States
and Europe, critics contend that subsequent continual appreciation of the
renminbi (couched in terms of making its exchange rate more flexible) may
also be required to balance fairly international competitiveness. And many
outside critics see ‘smooth’ ongoing upward adjustments in the renminbi to
be best obtained by China’s government eventually allowing its currency to
float freely – instead of intervening heavily to hold it down as is now the case.
In this chapter, I contend that these critics are wrong in both respects.
Their ‘conventional wisdom’ is based on faulty, although unfortunately
widely accepted, theorizing that fails to come to grips with how the inter-
national dollar standard works.
240
Exchange rate or wage changes in international adjustment? 241
I will first discuss why a discrete appreciation by a dollar creditor country

of the United States such as China or Japan would have no predictable
effect on its trade surplus. I will then show how differential adjustment in
the rate of growth of money wages will more accurately reflect inter-
national differences in productivity growth, that is, international competi-
tiveness will be better balanced through time, when a peripheral country’s
dollar exchange rate is fixed. Finally, I will discuss the optimal degree of
flexibility in the foreign exchanges for China at the present time.
1 THE EXCHANGE RATE AND THE TRADE

BALANCE
Why is the common presumption that an exchange rate change by itself

would have a predictable effect on a country’s trade balance often incorrect?
This presumption has been canonized in a model called the elasticities
approach to the balance of trade that focuses on relative price effects. It is
intuitively plausible because an appreciating country’s exports obviously
become more expensive in world markets and fall, whereas imports become
cheaper and ‘should’ increase. The problem is that proponents of the elas-
ticities approach focus on these relative price effects of an exchange rate
change and either ignore the income (absorption) effects or believe them to
be small and controllable.
But under the world dollar standard where foreign trade and asset flows
are largely invoiced in dollars, a peripheral country will be exposed to major
income and wealth effects should its dollar exchange rate change. In par-
ticular, an appreciation against the dollar would have (is having) unaccept-
able worldwide macroeconomic consequences without correcting the US
trade and current account deficits.
Among financially open economies, sustained exchange rate changes
must reflect relative monetary policies expected in the future: relatively tight
money and deflation in the appreciating countries and relatively easy
money with inflation in the country whose currency depreciates. After a
sharp appreciation, multinational as well as national firms will see the
country as a less good (more expensive) place in which to invest so that
investment slumps. In creditor countries that have built up large dollar
claims on foreigners, this deflationary impact of an exchange appreciation
is further accentuated because these dollar assets lose value in terms of the
domestic currency: a negative wealth effect that reduces consumption as
well as investment. In summary, depressed domestic spending offsets the
relative price effect of an appreciation so as to leave the effect on the net
trade balance indeterminate.2
The high-saving countries in Asia and Europe (and including Canada),

all creditors of the low-saving United States, face the specter of a growth
slowdown or outright deflation should their currencies appreciate. For
example, the repeated appreciations of the yen from 1985 to 1995 created
the bubble in Japanese land and equity values from 1987 to 1990 and then,
with the inevitable collapse of the bubble, threw Japan into a deflationary
slump in the 1990s with a zero interest liquidity trap that made monetary
policy helpless to revive the Japanese economy. In 2003 and 2004, the
Japanese economy staged a partial recovery on the back of the China
boom. But the current rise of the yen toward 100 to the dollar could well
throw Japan, still in a liquidity trap, into a renewed deflationary slump in
2005. Similarly, with the 50 percent appreciation of the euro against the
dollar in 2002–04, continental Europe is facing slower economic growth –
although not yet as protracted as the earlier Japanese experience.
These exchange-rate-induced growth slowdowns or slumps in the appreci-
ating economies sharply reduce their demand for imports. At the same time,
their exports become more expensive in world markets. Because the fall in
exports is coupled with a fall in imports, the net effect on their trade balances
is unpredictable (McKinnon and Ohno, 1997, Chs 6 and 7). The ever-higher
yen from 1971 to 1995 led to even bigger Japanese trade surpluses. All that
is predictable is that the appreciating country will suffer deflationary pres-
sure. However, if outside commentators and government officials persist in
the mistaken belief that appreciation will reduce the trade surpluses of
America’s creditor countries, their ‘talk’ encourages hot money flows out of
dollars into the currencies of creditor countries in Asia and Europe – a par-
ticularly acute problem for China currently. These countries then find it more
difficult to avoid actual appreciation and unwanted deflation.
Conversely, if deflationary pressure in creditor countries is muted
because of orchestrated joint appreciations of their currencies against the
dollar, this would induce an outburst of unacceptably high inflation in the
United States. For example, President Nixon’s well-telegraphed deprecia-
tion of the dollar in August 1971 touched off a flight from dollar assets and
also the high and volatile US inflation of the 1970s.
If exchange rate changes are not the answer to American trade deficits
and Asian trade surpluses, what are the more fundamental causes? The
major ongoing and long-run distortion in the world’s financial system is
America’s saving deficiency, large fiscal deficits by the Federal Government
and meager household saving, coupled with a virtually unlimited dollar
line of credit on which to borrow from the rest of the world.3 In addition,
over the last two years, US monetary policy has also been too loose with
short-term interest rates well below the rate of inflation leading to excess
consumption – in part by inducing a bubble in housing prices.
The result is heavy US borrowing in international markets that is then

transferred in terms of real resources by foreign countries running trade
surpluses with the United States. The US current account deficit forces
Canada and countries in Asia, Europe and now even in Latin America, into
current account surpluses.
2 THE EXCHANGE RATE AND INTERNATIONAL

COMPETITIVENESS
At least some of the critics of Asian countries’ pegging to the dollar would
agree that low saving in the United States, rather than misaligned exchange
rates, is the root cause of the trade imbalance. However, suppose a country
with very high productivity growth such as China trades with countries
with much lower productivity growth. In the new millennium, Japan and
Europe have overall trade surpluses, and the United States has an overall
trade deficit. But all of these more mature industrial countries have much
lower productivity growth than China’s. Isn’t exchange rate flexibility with
ongoing appreciation of the renminbi more or less necessary to balance
international competitiveness by offsetting the productivity differential
between China and its slower growing trading partners? Indeed, because of
foreign unease, China has promised that the yuan/dollar exchange rate will
become more flexible in the future.
As long as the American price level remains stable, more flexibility in the
central exchange rate of 8.28 yuan per dollar is neither necessary nor desir-
able for balancing international competitiveness with China’s neighbors in
the long run. International adjustment occurs by money wages naturally
growing faster in the country with higher productivity growth. But this
mechanism of differential wage adjustment, with more rapid wage growth in
China than the United States, only works well when enterprises and workers
in China are confident that the central rate will remain fixed indefinitely, and
China’s inflation remains more or less aligned with that in the United States.
Then Chinese employers in the rapidly growing tradables sectors, largely
manufacturing, will vigorously bid for workers subject to the constraint of
having to remain internationally competitive at the fixed nominal exchange
rate. Money wages, particularly for the increasingly skilled workers, then rise
in line with the high-productivity growth. Similar wage growth than spreads
out through the rest of the economy, including non-tradable services.
In the 1950s and 1960s under the Bretton Woods system of fixed dollar
exchange rates, how differential wage growth became the principal mode of
international adjustment was first articulated for high-growth Scandinavia
when the Swedish, Norwegian and Danish currencies were all pegged to the
dollar. But very high productivity growth in post-war Japan relative to the
United States, when the yen/dollar rate was also convincingly fixed, pro-
vides an equally striking example of what is now known as the
‘Scandinavian Model’ of wage adjustment (Lindbeck, 1979).
2.1 Japan’s Dollar Exchange Rate in Historical Perspective
When the yen was fixed at 360 to the dollar from 1950 to 1971, the import-
ance of relative wage adjustment between Japan and the United States was
pronounced. Table 8.1 gives the summary statistics for this remarkable era
of very high Japanese growth in comparison to those of the wealthier, and
consequently more slowly growing, United States. From 1950 to 1971,
Japan’s annual growth in real output was 9.45 percent while industrial
production grew an even more astonishing 14.56 percent per year.
Unsurprisingly, the annual growth in Japanese labor productivity of 8.92
percent was far in excess of the 2.55 percent in the United States. However,
the balancing item was that average money wages grew at a robust rate of
10 percent per year in Japan and only 4.5 percent in the US. Figure 8.1
shows the dramatic rise of Japanese money wages relative to American
wages under the Bretton Woods system of fixed dollar exchange rates.
Table 8.1 Key economic indicators for Japan and the United States,
1950–71 (average annual percentage change)
Wholesale Money Wages Consumer Industrial

Prices Prices Production
US Japan US Japan US Japan US Japan
1.63 0.69a 4.52 10.00 2.53 5.01 4.40 14.56
Real GDP Nominal Narrow Money Labor
GDP Productivity
3.84 9.45a 6.79 14.52a 3.94 16.10b 2.55 8.92c
Notes:
a1952–71.
b1953–71.
c1951–71.
Sources: IMF, International Financial Statistics, CD-ROM, October 2003, unless

otherwise noted. Japanese real income data, wholesale price data, and labor productivity
data are from Japan Economic Yearbook, 1960–71, Economic Survey of Japan, 1954–59, and
OECD Economic Surveys: Japan, 1964–71. Labor productivity data for the US are obtained
from the index for the non-farm business sector as reported by the Bureau of Labor
Statistics. Note that labor productivity refers to the industrial sector.
800
Japan
USA
700
600
500
Index
400
245
300
200
100
195019511952195319541955 1956195719581959196019611962196319641965196619671968196919701971
Source: IMF: IFS.
Figure 8.1 Nominal manufacturing wage growth for US and Japan: 1950–71 (base year 1950 ⫽ 100)
Keeping the yen at 360 per dollar effectively anchored Japan’s price level
for tradable goods. In the 1950s and 1960s, the Japanese wholesale price
index (WPI) rose less than 1 percent per year whereas the American WPI
rose a bit more than 1 percent (Table 8.1). Because the bulk of world trade
was invoiced in dollars, fixing an exchange rate to the dollar was (is) a
stronger anchor for the price level than the size of Japanese bilateral trade
with the United States would suggest.
Employers in Japan’s manufacturing export sector, with its extremely
high growth in labor productivity, then bid vigorously for both skilled and
unskilled workers subject to remaining internationally competitive at the
fixed exchange rate. Wages rose rapidly in manufacturing so that workers
received the main fruits from the productivity growth there. But then, as in
the Scandinavian Model, these high wage settlements spread into the rest
of the economy, such as in non-tradable services, where productivity
growth was much lower. The result was that, within Japan, the price of ser-
vices rose relative to goods prices. For 1950–71, Table 8.1 shows that Japan’s
CPI, which includes services as well as goods, began to increase much faster
at 5 percent per year than its WPI, which contains only goods. But Japan’s
international competitiveness in its high-growth tradables sector remained
balanced with the United States.
In Japan’s bygone high-growth era, fashioning a purely domestic mone-
tary anchor would have been more difficult. As in China today, restrictions
on domestic interest rates proliferated; and the rate of growth in narrow
money was high and unpredictable – more than 16 percent per year from
1950 to 1971 as Japanese households rebuilt their financial assets after the
war. Thus having the Bank of Japan simply key on the dollar exchange rate
was the most convenient instrument for stabilizing Japan’s tradable goods
price level while promoting high growth in money wages.
By the end of the 1960s, however, American monetary policy became too
inflationary. The loss of America’s foreign competitiveness was too great
for the Bretton Woods system of fixed dollar exchange parities to survive.
President Nixon had to choose between disinflating at home and thus main-
taining the fixed rate system, or forcing a devaluation of the dollar against
other major currencies while continuing to inflate. He chose to devalue in
August 1971, and the United States suffered the great inflation of the 1970s.
2.1.1 The ‘floating’ yen: appreciation and deflation

For two decades after August 1971, productivity growth in Japan remained
high relative to that in the United States. Japanese exports made major
inroads into American markets for steel, autos, machine tools, semi-
conductors, and so on. In the midst of numerous trade disputes, the US
government reacted by continually trying to ‘talk’ or force the yen up on
the presumption that an appreciating yen would improve America’s exter-

nal competitiveness.4 Indeed, the yen did appreciate all the way from 360 in
1971, to touch 80 to the dollar in April 1995, and threw Japan into its
deflationary slump of the 1990s with a zero interest liquidity trap that lasts
to the present day (McKinnon, 2005).
The deflation also reduced growth in Japanese money wages. It essen-
tially destroyed the natural wage-adjustment mechanism for balancing
international competitiveness that had held when the exchange rate was
fixed. Once the yen began to appreciate, Figure 8.2 shows the breakdown
in relative wage adjustment – albeit with a lag. Before 1975, money wage
growth in Japan remained much higher than in the United States.
Subsequently, as relative deflation in Japan set in (particularly from the
sharp rise in the yen in 1977–78), Japan’s money wage growth slowed
sharply. From the 1980s into the new millennium, it became even lower
than that in the United States. So, besides damaging the Japanese economy
in a macroeconomic sense while failing to reduce its trade surplus, the errat-
ically appreciating yen undermined the natural process of relative wage
adjustment for balancing international competitiveness.
Although the yen has not appreciated on net balance since 1995, it has
fluctuated widely against the dollar. Without the assurance of a fixed
exchange rate anchor, re-establishing growth in Japanese money wages to
reflect (potential) productivity growth accurately remains problematic. In
2004, annual money wage growth in Japan was close to zero; whereas in the
US it was 2.4 percent.
2.2 Wage Adjustment in China under a Fixed Exchange Rate
Unlike Japan, China has kept its exchange rate stable since 1994 – and did
not have the earlier misfortune of being pushed into a deflationary slump
from an appreciating currency. Table 8.2 provides the key summary stat-
istics comparing China to the United States. From 1994 through 2003,
money wages in manufacturing increased by about 13 percent in China and
by just 3 percent in the United States. This ten percentage-point wage-
growth differential approximately reflected the differential growth of labor
productivity: about 12.3 percent in China5 versus 2.7 percent in the United
States since 1994. Under the fixed yuan/dollar exchange rate, the appropri-
ate wage-adjustment mechanism for balancing international competitive-
ness seems to be alive and well.
Figure 8.3 shows China’s dramatically higher growth in money wages in
manufacturing relative to the United States over the past decade. Within
China, Figure 8.4 shows that wages in all sectors were rising fast – with wage
growth in manufacturing about the median for the economy as a whole.
20 380
Wage differential
Inflation differential
15 Yen/dollar
300
10
220
Yen/dollar
Percent
5
248
140
0
50
53
56
59
62
65
68
71
74
77
80
83
86
89
92
95
98
01
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
20
60
–5
–10 –20
Note: Positive values indicate higher inflation and higher wage increases in Japan.
Source: IMF: IFS.
Figure 8.2 Inflation and wage differential between Japan and US, and yen/dollar rate, 1950 to 2004
Table 8.2 Key economic indicators for China and the United States,
1994–2003 (average annual percentage change)
Wholesale Money Wages Consumer Industrial

Prices (mfg) Prices Production
US China US China US China US China
1.53 1.26a 3.03 13.04b 2.43 2.84 3.00 12.17c
Real GDP Nominal Narrow Money Labor
GDP Productivity
3.17 8.55 5.03 10.74 4.16 17.88 2.70 12.32d
9.48e
Notes:
a Ex-factory price index.
b 2003 data on manufacturing wages are projected from overall average wages from
1997–2003.
c 1994–2002.
d 1994–2001. Zhang and Tan (2004).
e 1994–2002. R. Fernholz (2004).
Source: IMF, International Financial Statistics CD-ROM, Nov. 2004, unless otherwise
noted. Chinese CPI, manufacturing wage data, labor productivity data, real income data
and wholesale price data are from China Statistical Yearbook, 2004. Labor productivity data
for the US are obtained from the index for the non-farm business sector as reported by the
Bureau of Labor Statistics. The China labor productivity data refer only to the industrial
sector.
Much of this reflects the upgrading of skills and greater work experience
of the labor force. True, at the margin, the wages of unskilled migrant
workers may be lagging – and many of these seem to be absorbed into con-
struction activities where average wages show the slowest rate of growth in
Figure 8.4.
China’s exchange rate stabilization in 1994 followed a major depreciation
of the renminbi associated with the unification of the official exchange
rate at the much higher ‘free-market’ swap rate. Figure 8.5 shows that the
official rate jumped from 5.5 to 8.7 yuan per dollar. Because much of
China’s trade – particularly in manufacturing – had been transacted at the
higher swap rate, this jump in the official rate overstates the effective deval-
uation. Nevertheless, because of a temporary burst of domestic inflation
from 1993 to 1996 as shown in Figures 8.5 and 8.6, the ‘real’ devaluation was
negligible. But the nominal devaluation certainly exacerbated the inflation.
By 1996, the renminbi had appreciated slightly to 8.28 to the dollar where it
remained until 2005. Chinese price inflation then settled down after 1996
and seems have converged close to the American level. In 2004, China’s CPI
350
China
USA
300
250
Index
250
200
150
100
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
Source: IMF: IFS; Chinese manufacturing wage data are obtained from China Statistical Yearbook, 2004.
Figure 8.3 Nominal manufacturing wage growth for US and China, 1994–2003 (base year 1994 ⫽ 100)
300
280
260
240
Farm
Mine
220
Mfg
Index
200 ElecGas
Const
251
180 Trans
Retail
160 SocServ
140
120
100
1994 1995 1996 1997 1998 1999 2000 2001 2002
Source: China Statistical Yearbook, 2003.
Figure 8.4 China: nominal wages across different sectors, 1994–2002 (base year 1994 ⫽ 100)
30 10
25 Yuan/USD Exchange
Rate 9
(RHS)
20
8
15
Percent
10 7
CPI Inflation Differential:
China–US
252
(LHS)
5
6
0
1993M1 1994M1 1995M1 1996M1 1997M1 1998M1 1999M1 2000M1 2001M1 2002M1 2003M1 2004M1
5
–5
–10 4
Note: M1 ⫽ January.
Source: EIU.
Figure 8.5 China–US inflation differential and exchange rate, 1993–2005

135
130
125
120 China CPI

Index
USA CPI
China WPI
115
253
USA WPI
110
105
100
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
Source: IMF: IFS; Chinese ex-factory price data are obtained from China Statistical Yearbook, 2004.
Figure 8.6 CPI and WPI for US and China, 1994–2003

rose 3.8 percent while that in the United States rose 3.3 percent. The fixed
rate regime served China as a nominal anchor for its price level much like
the fixed yen/dollar rate served Japan in the 1950s and 1960s.
One might argue that, in 2004, inflation was too high in both China and
the United States. But under the international dollar standard, only the
center country can exercise monetary policy independently. Thus the onus
is on the United States to disinflate. Fortunately, in 2005 and 2006, the
Federal Reserve was committed to steadily increasing short-term interest
rates back to more normal levels after its unprecedented monetary easing
(low federal funds rates) in 2003 and 2004. Thus, inflation should calm
down in both countries. China should have less trouble with inflows of ‘hot’
money, and even less trouble if outside commentators and government
officials stop talking about the ‘need’ to appreciate the RMB.
More generally for the increasingly integrated East Asian economies,
China’s fix at 8.28 yuan per dollar became the key to intra East Asian
exchange stability in the new millennium. All the others more or less peg to
the dollar and thus to each other. If this fixed rate system continues, adjust-
ment in relative wage growth in the other East Asian economies becomes
the main vehicle for balancing international competitiveness.
In the context of the old Scandinavian model of wage adjustment,
Table 8.3 below compares the 1950–71 Japanese and 1994–2005 Chinese
experiences under fixed dollar exchange rates (the first row) with the post-
1971 Japanese experience with a floating and erratically appreciating yen
(the second row).
Under a fixed dollar exchange rate, the system converges to relative pur-
chasing power parity (PPP): the rate of inflation in tradables in the peripheral
country converges to that in the center country. Subject to the fixed exchange
rate constraint, wage bargaining is determinate in the sense that employers
bid up money wages to reflect fully the growth of labor productivity in the
rapidly growing export sector. International competitiveness is balanced.
Under a floating rate that moves randomly or is hectored into appreciat-
ing as with the Japanese yen from 1971 to 1995, bargaining over money
Table 8.3 Wage bargaining
Relative PPP Wage Bargaining Money Wage

Process Growth
Fixed dollar Yes Determinate Reflects productivity
exchange rate growth in tradables
Floating No Indeterminate ?
exchange rate
wage growth becomes indeterminate. Because risk-averse employers can no

longer judge the future course of the (erratically appreciating) exchange
rate, they hesitate to bid money wages up by the full amount of productiv-
ity growth in the tradables sector. This could be characterized as a negative
risk premium in wage bargaining that parallels the negative risk premium
in domestic interest rates arising out of fear of appreciation, which
has driven short-term interest rates in Japan to zero (McKinnon, 2005,
Chapter 4). When full-scale deflation sets in, money wage growth can also
approach zero but, unlike interest rates, could even become negative.6
External pressure aside, the weakest link in the current East Asian mon-
etary cum exchange rate system still seems to be Japan. The yen/dollar rate
has not been credibly fixed within a narrow range despite massive inter-
ventions by the Bank of Japan to prevent the yen from appreciating. The
fear of future yen appreciation and further deflation is still rife – and
money wages are not adjusting properly. So the first order of business is
for the Bank of Japan is to come up with a more credible dollar fix for the
yen that would better stabilize intra East Asian exchange rates while ame-
liorating deflationary fears in Japan itself (McKinnon, 2005; McKinnon
and Ohno, 1997).
However, China now faces more external pressure to appreciate, and the
threat of trade sanctions if it does not – much like the ‘Japan bashing’ of
the 1970s to the mid-1990s. The potential threat from being pushed into
continual currency appreciation, as did Japan after 1971, is a consequent
slowdown in China’s economic growth and eventually outright deflation.
3 LIMITED FOREIGN EXCHANGE FLEXIBILITY

FOR CHINA
Partly responding to foreign pressure but also because China could benefit
from a more open foreign exchange market with decentralized transacting,
it is important to pin down what the Chinese government should mean by
greater foreign exchange ‘flexibility’. This involves both the pace of liber-
alizing and rationalizing of capital controls (relaxing administrative con-
straints on foreign exchange transacting) and the optimal degree of
flexibility in the exchange rate itself. Let us discuss each in turn.
In one respect China has been, and remains, very open to foreign capital
flows. Inward foreign direct investment (FDI) at over US$40 billion per
year since 2000 has been higher than China’s multilateral trade surplus of
about 3 percent of GDP. FDI has been an important vehicle for introduc-
ing the modern technology underlying China’s rapid industrial transfor-
mation. It has also been a major contributor to the build up of China’s
liquid dollar assets, both held privately and increasingly as official exchange
reserves – about US$660 billion as of March 2005. And outward FDI may
be beginning in a significant way, as with Lenovo’s purchase of IBM’s PC
division in 2005, and with large numbers of less publicized investments in
African infrastructure projects.
However, there are two distortions in official Chinese policy that unduly
amplify the financial magnitudes of inward FDI. First, the tax treatment
of foreigners investing in China, particularly in special economic zones, is
still more favorable than that accorded domestic investment – although
much equalization has occurred. While entirely rational at the beginning of
China’s opening to international trade in the 1980s, such favoritism for
foreign FDI is now counterproductive for encouraging domestic entrepre-
neurship. The second distortion is to limit foreign firms borrowing from
domestic Chinese banks to help fund their operations in China. China
wants and needs foreign technical expertise, but with its high domestic
saving, it does not need foreign finance. Thus allowing foreign firms to
borrow domestically would reduce unwanted financial inflows.
In foreign trade more generally, China has followed, and is following, the
optimal order of (gradual) economic liberalization (McKinnon, 1993). In
the mid-1990s, China consolidated its exchange rate system and achieved
full current account convertibility for the renminbi in the sense of satisfy-
ing the IMF’s Article VIII. In the new millennium, China is rapidly satis-
fying its WTO obligations by eliminating quota restrictions and drastically
cutting tariffs. On capital account, it has liberalized relatively illiquid
FDI flows before eliminating restrictions on shorter-term and more liquid
financial flows. All this is according to what is now received textbook
theory.
But this last stage, that of liberalizing liquid international financial flows,
which can all too easily become ‘hot’ money, is best done very carefully in
conjunction with an appropriate regulatory framework – or not at all. The
most important principle is to contain (latent) moral hazard in financial
institutions, implicitly or explicitly insured by the Chinese government, by
limiting their ability to assume risk. In particular, undue foreign exchange
exposure in banks and other financial institutions can imperil both them-
selves and the economy. And foreign financial institutions should always be
subject to the same stringent regulatory constraints as domestic ones.
What is the best way to proceed with this delicate last stage in the liberal-
ization process? One way is to assign foreign exchange trading exclusively
to authorized banks that must keep their exposure in any foreign currency
against renminbi within well defined limits. The State Administration
for Foreign Exchange (SAFE) would then continuously monitor the net
foreign exchange positions of these authorized banks relative to their
capital positions. Only after this interbank foreign exchange market

between renminbi and dollars is well established would further liberaliza-
tion be considered. For example, free foreign exchange trading between
major foreign currencies, such as yen against euros, would remain prohib-
ited until a later stage.
Assuming that these prudential regulatory restraints are in place, what
should be the range of variation of the yuan/dollar rate itself ? In the 1960s
during the old Bretton Woods system of fixed dollar parities for Western
European countries and Japan, market rates varied within a 2 percent band
around their central rates. Indeed, the foreign exchange margins (1 percent
on either side of the official parity rate) were officially announced and cor-
responded to Article IV of the original IMF Articles of Agreement. On
most trading days, their central banks did not have to intervene and all the
clearing of international payments devolved to the authorized commercial
banks. But the system was punctuated by occasional crises when official
intervention became necessary.
In China today, allowing a similar 2 percent band around the central
rate of 8.28 yuan per dollar, within which the market rate could fluctuate
freely daily or weekly, would efficiently decentralize the foreign exchange
market. Indeed, as capital controls are replaced by careful prudential regu-
lation over the net foreign exchange exposures of authorized commercial
banks, a widening of the band to, say, 1 percent on either side of the
‘parity’ rate of 8.28 yuan per dollar would make the foreign exchange
market more flexible. The current margins are only about 0.3 percent on
either side of 8.28, and unduly limit the private profitability of foreign
exchange transacting.
If the 2 percent band is fully credible, in non-crisis periods, the People’s
Bank of China need seldom intervene to maintain it. Thus having the PBC
officially announce what the band limits are reinforces regressive expect-
ations: any deviation of the market rate away from 8.28 is expected to be
followed by a move back toward 8.28. When the rate moved toward the top
of the band, say, 8.4 yuan/dollar, then authorized banks would voluntarily
intervene to drive it back down. When these banks know that the PBC is
prepared to intervene decisively at either the upper or lower boundaries of
the band, they will act as dealers (stabilizing speculators) to nudge the rate
toward the middle of the band – without the PBC actually intervening. This
ensures that the PBC need not itself intervene much in practice. Then, the
clearing of most international payments would devolve from the People’s
Bank of China (PBC) to the commercial banks or other authorized
financial institutions. Subject to constraints on their net foreign exchange
position, they would be free to make hedging markets in foreign exchange
futures and options for their non-bank customers.
However, if there is a crisis, as when foreigners put heavy political pres-

sure on China to appreciate leading to hot money inflows, then the PBC has
little choice but to intervene by however much is necessary to preserve the
central rate.
Changes in the market exchange rate within such a narrow band would
not significantly affect – or be intended to affect – a country’s competitive-
ness in international markets for goods or services. It is simply a device for
providing flexibility in decentralizing the foreign exchange market. As dis-
cussed in section 2 above, under a securely fixed central exchange rate, high
growth in domestic money wages – reflecting China’s very high growth in
labor productivity – would remain the dominant mechanism for balancing
international competitiveness.
NOTES
1. I would like to thank Hong Qiao and Ricardo Fernholz of Stanford University, and
Gunther Schnabl of the University of Tubingen, for their great help in preparing this
chapter.
2. In dollar debtor countries facing the threat of having their currencies depreciate
against the dollar, the negative wealth effect tends to reinforce the relative price effect
of an actual devaluation. Their trade balances could improve sharply from devaluation
as domestic consumption (and imports) slumps even as their now cheaper exports
expand into world markets. This was the case for Indonesia, Korea, Malaysia,
Philippines and Thailand after the great Asian Crisis of 1997–98. Their current
accounts went from being sharply negative before the crisis to positive immediately
afterwards.
3. In a provocative paper, Ben Bernanke (2005) argues that the problem is more one of excess
saving in other countries, particularly in Asia, than a saving deficiency in the United
States. Either way, however, this international saving imbalance cannot be corrected by
exchange rate changes.
4. As discussed in detail in Chapter 4 of McKinnon and Ohno (1997).
5. This estimate of productivity growth is not official, and was taken from Zhang and Tan
(2004). In both countries, how best to measure labor productivity growth is controversial.
Estimates for China can vary.
6. I am greatly indebted to Ms Hong Qiao for pointing out this interesting parallel between
a negative foreign exchange risk premium in domestic interest rates and in bargaining over
growth in domestic money wages.
REFERENCES
Bergsten, C. Fred, and the Institute for International Economics (2005), The
United States and the World Economy, Institute for International Economics,
Washington DC.
Bernanke, Ben (2005), ‘The Global Saving Glut and the US Current Account
Deficit’, At the Sandridge Lecture, Virginia Association of Economics,
Richmond, Virginia, 10 March.
Fernholz, Ricardo (2004), ‘China, Japan, and Maintaining International Competitive

Balance under a Fixed Exchange Rate’, Senior Honors Thesis, Stanford University,
2004.
Lindbeck, Assar (1979), Inflation and Unemployment in Open Economies,
Amsterdam, North Holland, 1979.
McKinnon, Ronald (1993), The Order of Economic Liberalization: Financial
Control in the Transition to a Market Economy, Johns Hopkins University Press,
2nd Edition. (Translations: Chinese, Romanian and Arabic, 1995.)
McKinnon, Ronald (2005), Exchange Rates under the East Asian Dollar Standard:
Living with Conflicted Virtue, MIT Press. (Translations: Chinese 2005, Japanese
forthcoming 2007.)
McKinnon, Ronald and Kenichi Ohno (1997), Dollar and Yen: Resolving Economic
Conflict between the United States and Japan, MIT Press. (Translations: Japanese
1998, Chinese 1999.)
Zhang, Xiaobo and Kong-Yam Tan (2004), ‘Blunt to Sharpened Razor:
Incremental Reform and Distortions in the Product and Capital Markets in
China’, International Food Policy Research Institute, DSGD Discussion Paper
No. 13, August.
9. Adjustment to China’s CPI-based
inflation rate to account for the
‘true’ cost of living, 1993–2004
Lawrence R. Klein, Huiqing Gao and
Liping Tao
INTRODUCTION
Much discussion has taken place during the past two or three years con-
cerning China’s growth rate, often dealing with the appraisal of statistical
reliability and possible bias in the published figures. This discussion
has taken place during a period when many people have seen for them-
selves the unfolding of a Chinese expansion that has been almost unbe-
lievable in the speed, scope and depth of movement of the largest
economy in the world, measured in terms of population. Most of the
scholarly discussion and publication has dealt with the unbelievable
numerical size and speed of the expansion, drawing the frequent conclu-
sion that the magnitudes are overstated.1 However, that is not the
appraisal that comes from some of the authors of this chapter, who used
quite different methods of measurement that seem to validate the impres-
sions formed by the non- (economic) professional visitors who have actu-
ally experienced what has been taking place since 1978, especially on a
year-after-year basis, so that the evolution of new developments are
clearly exposed and China’s stabilizing role in the crisis of 1997–98 is
properly understood.
The main purpose of this chapter is to examine a crucial part of the issue
at hand by concentrating on a neglected important part, namely, the rate of
price level change, that is, the inflation rate that is needed to convert the
nominal direct measures of GDP to real GDP. When it is claimed that
China’s GDP has roughly quadrupled between 1980 and 2000 or that the
national growth rate since reform in 1978 has been about 9 percent, the
statements implicitly mean GDP adjusted for inflation. There has been rela-
tively little attention paid to the evaluation of the price deflator used to
convert nominal into real values. In this chapter, we are going to argue that
260
Adjustment to China’s CPI-based inflation rate 261
the magnitude of inflation has been overstated, because it does not take
account of quality change or lifestyle change, in a broader sense.
It should be noted at the outset that the United States went through a
period of significant change in the measurement of the economy’s rate of
inflation when some leading economic officials and scholars decided that
the major price indexes for the US overstated the magnitudes. This was
undertaken by the Boskin Committee, who examined such things as
product quality, new products, place of purchase and other lifestyle fea-
tures.2 They concluded that the prevailing price indexes overstated US
inflation by approximately as much as one full percentage point and that
about half a point was due to quality improvement alone.
Without using the Boskin Committee’s methods of analysis, which are
not easily transferable to China because of unavailability of comparable
data, we are going to estimate an adjustment to China’s CPI by quite
different methods and determine how to estimate the real growth rate
of GDP.
The approach that we are using for China is through estimation of the
‘true’ economic cost of living index. During World War II, some British
economists challenged the government’s allowances for civilians, because
the officials did not allow for enforced substitution of rationed for unra-
tioned goods, thus lowering the quality of the available market basket.
Following that debate, one of the authors of this chapter helped to produce
a research paper in 1947 to show how one might specify an equation system
that could be estimated from available data. This system is known as the
linear expenditure system (LES).3
The relevance of the LES in the present context is that it provides a
readily computable equation system that can be determined from Chinese
data, and the formula for the ‘true’ cost of living for China for the period
under dispute can be calculated to see how much the inflation rate might
be lowered (and the growth rate correspondingly increased) for the
economy as a whole. It should be borne in mind that the US potential
growth rate, which used to be considered to be roughly 3 percent was
raised to 4 percent or more by the work of the Boskin Committee. If the
US can claim a higher rate of expansion by revising the price index, why
not make a similar revision for China? It is quite evident that the Chinese
economy has gone through a much faster and larger change in lifestyle
since 1978 than has the US economy. The diet is better; clothing is
immensely better; transportation is superior in many ways; retailing is
better; services are better; tourism is far better; and so on and so on. Given
these changes in living conditions, the economic goal for all progressive
societies, surely one should try to take these changes into account when
evaluating economic growth.
THE LINEAR EXPENDITURE SYSTEM
The LES equations are not unique, but just any comprehensive demand
system is not suitable – it should have some economic-theoretical proper-
ties. Of course it need not be linear, but this property can be tested. The the-
oretical basis for the LES is that it satisfies the following properties from
consumer theory:
1. Demand for each type of real expenditure should be homogeneous of

degree zero in prices and incomes, that is to say, real demand should
depend on relative prices and real income (that is, the ratio of income
to price).
2. The equations should satisfy the Slutsky–Hicks symmetry conditions,
that is, the partial derivative of the i-th demand with respect to the
j-th price should equal the derivative of the j-th demand with respect
to the i-th price, both derivatives evaluated along a given indifference
curve.
3. The sum of individual expenditure amounts should equal total income,
including future spending, that is, saving (adding up).
The LES can be written as
n n
(pitqit ⫺ pit␥i ) ⫽ ␤i (rt ⫺ 兺pjt␥j);兺
j⫽1 i
␤i ⫽ 1
⫽1

expenditure on the minimum total total minimum ␤i ⫽marginal
i-the category subsistence income subsistence propensity to
expenditure expenditure consume
on the i-th on all out of
category categories supernumerary
income
This equation is expressed in current price values (nominal). It can also be

transformed into real (constant price values). All statistical equations
(except some definitions) should be expressed with explicit allowance for
error. In this version, we have proceeded as though there is additive error
for each equation as specified above in order to use properties of random
variation for evaluation. It is written in such a way as to express intuitive
behavior, that is, supernumerary expenditures are linear functions of super-
numerary income.
The LES is easy to understand and provides some empirical information

without making an elaborate effort, but there are drawbacks. In the first place
the LES does not permit inferior goods/services. An inferior good or service
is one whose consumption varies inversely with total income. In a typical
family budget there are, indeed, many examples of inferior goods and, if the
implicit equation system is to be integrated to the associated utility function,
it becomes immediately obvious that inferior goods are not acceptable.
Nevertheless, this defect can be dealt with by grouping. There are hundreds
or thousands of goods and services in a typical family budget, but if they are
grouped as we have used them in the present investigation, into only eight
classes (Food; Clothing; Household Facilities, Articles and Services;
Medicine and Medical Services; Transport, Post and Communication
Services; Educational, Cultural and Recreation Services; Residence;
Miscellaneous Commodities and Services), then we can treat each whole
group as a non-inferior good. There can be much substitution between finely
specified inferior and non-inferior goods in the process of given calculations.
The Engel curves that separately relate expenditure on each group of
goods and services to total income should be linear in this model, given the
implied utility function, and they are for 1993 or 1994, but are linear only
up to the highest income class in 2002, where the curves turn up. The part
of the Engel curves that we are using for parameter estimation is in a linear
range. We would not be using Engel curves for 2002 in the present analysis,
were it not for the matter of data availability. Our procedures treat urban
and rural consumers separately, and the Engel curve database is not avail-
able for rural consumers at an earlier date.
This brings us to the technique for parameter estimation for determin-
ation of sample values for ␥i and ␤i. The ␥ and ␤ values range across the
grouping of goods and services. There are eight values of each parameter,
but since the ␤ values are constrained to sum to unity, there are only seven
free values of ␤i.
There are two sources of variation in the sample. One source refers to
cross-section variation from family budgets. One group of budgets refers to
the urban population and another to the rural population, as their eco-
nomic positions and choices are known to be quite different. The other
source of variation is across time periods. For reasons of data availability
and reliability, we have limited our samples to the period 1993–2002,
although the effects of changes in lifestyle were also important in the earlier
period 1978–93. Graphs and tables of Engel curve data from family budget
data are given in Figures 9.1–9.8 and the corresponding Tables 9.1–9.8.
The parameters ␥i represent necessary levels of consumption in each group-
ing of goods and services. In order to measure supernumerary expenditures
for each group of goods and services we need to have estimates of the minimal
1600
1400
1200
Expenditure (yuan p. a.)

1000
800
600
400
264
200
0
1239.35 1359.87 1718.63 2041.67 1453.88 1985.88 3626.66 4905.77
Income (yuan p. a.)
Food Clothing
Household Facilities, Articles and Services Medicine and Medical Services
Transport, Post and Communication Services Education, Cultural and Recreation Services
Residence Miscellaneous Commodities and Services
Figure 9.1 Engel curve for 1993 (per capita, urban)

Table 9.1 Engel curve for 1993 (per capita, urban)
Poor Lowest Low Lower Middle Middle Upper Middle High Highest
Income (yuan p. a) 1239.35 1359.87 1718.63 2041.67 1453.88 1985.88 3626.66 4905.77
Expenditure (yuan p. a) 1183.15 1261.36 1528.68 1770.17 2055.72 2404.13 2810.32 3533.49
Food 700.33 744.58 869.27 957.57 1064.65 1169.40 1284.57 1464.65
Clothing 127.01 142.50 195.79 246.43 303.32 367.20 418.16 492.62
Household Facilities, 60.00 66.51 94.42 120.06 163.10 220.27 314.16 446.75
Articles and Services
265
Medicine and Medical 36.02 36.05 41.92 50.09 53.59 61.52 74.58 100.34
Services
Transport, Post and 24.18 28.34 35.69 51.04 68.72 99.60 127.27 212.16
Communication Services
Education, Cultural and 101.43 113.29 136.40 158.43 182.91 221.37 256.21 356.23
Recreation Services
Residence 92.88 93.28 103.74 119.33 132.98 150.50 187.61 240.87
Miscellaneous Commodities 34.69 36.81 51.45 67.21 86.45 114.26 147.76 219.88
and Services
4500
4000
3500
3000
2500
2000
1500
1000
500
266
0
4759.10 5153.91 6221.44 7064.18 4812.34 6275.38 11061.31 13834.27
Income (yuan p. a.)
Food Clothing
Figure 9.2 Engel curve for 1993 (family, urban)

Table 9.2 Engel curve for 1993 (family, urban)
Size of family (persons) 3.84 3.79 3.62 3.46 3.31 3.16 3.05 2.82
Income (yuan p.a.) 4759.10 5153.91 6221.44 7064.18 4812.34 6275.38 11 061.31 13 834.27
Expenditure (yuan p.a.) 4543.30 4780.55 5533.82 6124.79 6804.43 7597.05 8571.48 9964.44
Food 2689.27 2821.96 3146.76 3313.19 3523.99 3695.30 3917.94 4130.31
Clothing 487.72 540.08 708.76 852.65 1003.99 1160.35 1275.39 1389.19
267

Services
Recreation Services
Residence 356.66 353.53 375.54 412.88 440.16 475.58 572.21 679.25
and Services
2500
2000
Expenditure (yuan p. a.) 1500
1000
500
268
0
1566.33 1734.57 2238.37 2721.15 3303.66 4079.07 5007.24 6837.81
Income (yuan p. a.)
Food Clothing
Figure 9.3 Engel curves for 1994 (per capita, urban)

Table 9.3 Engel curves for 1994 (per capita, urban)
Income (yuan/p. a.) 1566.33 1734.57 2238.37 2721.15 3303.66 4079.07 5007.24 6837.81
Expenditure (yuan/p. a.) 1512.70 1644.56 2028.80 2351.56 2798.12 3252.70 3880.91 4799.83
Food 932.90 1005.76 1169.37 1305.41 1431.04 1563.81 1727.97 1921.61
Clothing 144.79 167.99 235.54 309.39 390.89 477.17 569.08 682.33
269
Services
Recreation Services
Residence 116.18 121.54 143.23 161.58 183.24 211.41 259.70 335.98
and Services
6000
5000

4000
3000
2000
1000
270
0
5920.73 6452.60 7990.98 92 24.70 10 836.00 12 889.86 15 071.79 19 214.25
Income (yuan p. a.)
Food Clothing
Figure 9.4 Engel curves for 1994 (family, urban)

Table 9.4 Engel curves for 1994 (family, urban)
Income (yuan p. a.) 5920.73 6452.60 7990.98 9224.70 10 836.00 12 889.86 15 071.79 19 214.25
Expenditure (yuan p. a.) 5718.01 6117.76 7242.82 7971.79 9177.83 10 278.53 11 681.54 13 487.52
Food 3526.36 3741.43 4174.65 4425.34 4693.81 4941.64 5201.19 5399.72
Clothing 547.31 624.92 840.88 1048.83 1282.12 1507.86 1712.93 1917.35
271

Services
Recreation Services
Residence 439.16 452.13 511.33 547.76 601.03 668.06 781.70 944.10
and Services
4500
4000
3500

3000
2500
2000
1500
1000
272
500
0
2063.95 2527.68 3833.01 5209.18 7061.37 9437.99 12555.07 20208.43
Income (yuan p. a.)
Food Clothing
Figure 9.5 Engel curves for 2002 (per capita, urban)

Table 9.5 Engel curves for 2002 (per capita, urban)
Income (yuan p. a.) 2063.95 2527.68 3833.01 5209.18 7061.37 9437.99 12 555.07 20 208.43
Expenditure (yuan p. a.) 2079.52 2387.91 3259.59 4205.97 5452.94 6939.95 8919.94 13 040.69
Food 988.19 1127.41 1457.87 1772.88 2140.34 2596.95 3171.36 4100.79
Clothing 152.38 193.09 309.49 438.38 571.19 737.20 866.38 1103.16
273
Services
Recreation Services
Residence 261.45 282.74 355.12 421.90 563.31 643.15 906.67 1485.72
and Services
12000

10000
8000
6000
4000
2000
274
0
7203.19 8669.94 12802.25 16669.38 21466.56 27464.55 34903.09 54158.59
Income (yuan p. a.)
Food Clothing
Figure 9.6 Engel curves for 2002 (family, urban)

Table 9.6 Engel curves for 2002 (family, urban)
Income (yuan p. a.) 7203.19 8669.94 12 802.25 16 669.38 21 466.56 27 464.55 34 903.09 54 158.59
Expenditure (yuan p. a.) 7257.52 8190.53 10 887.03 13 459.10 16 576.94 20 195.25 24 797.43 34 949.05
Food 3448.78 3867.02 4869.29 5673.22 6506.63 7557.12 8816.38 10 990.12
Clothing 531.81 662.30 1033.70 1402.82 1736.42 2145.25 2408.54 2956.47
275

Services
Recreation Services
Residence 912.46 969.80 1186.10 1350.08 1712.46 1871.57 2520.54 3981.73
and Services
1600
1400

1200
1000
800
600
400
276
200
0
1551.79 2288.34 3025.17 4075.60 7567.22
Income (yuan p. a.)
Food Clothing
Figure 9.7 Engel curves for 2002 (per capita, rural)

1600
1400

1200
1000
800
600
400
277
200
0
1551.79 2288.34 3025.17 4075.60 7567.22
Income (yuan p. a.)
Food Clothing
Figure 9.8 Engel curves for 2002 (family, rural)

Table 9.7 Engel curves for 2002 (per capita, rural)
Low Lower Middle Upper High

Middle Middle
Income (yuan p. a.) 1551.79 2288.34 3025.17 4075.60 7567.22
Expenditure (yuan p. a.) 1006.36 1310.34 1645.04 2086.72 3500.09
Food 562.37 686.76 808.99 949.52 1354.31
Clothing 56.75 74.23 93.98 122.86 198.89
Household Facilities, 38.95 52.83 67.77 93.32 168.51
Medicine and Medical 57.54 74.88 90.73 116.49 201.72
Services
Transport, Post and 41.76 67.65 100.50 153.64 321.40
Communication
Services
Education, Cultural and 97.36 145.50 193.28 246.91 416.88
Recreation Services
Residence 127.56 174.75 244.14 336.15 704.75
Miscellaneous 24.07 33.74 45.65 67.83 133.63
Commodities and
Services
Table 9.8 Engel curves for 2002 (family, rural)
Low Lower Middle Middle Upper Middle High
Average size of 4.66 4.36 4.15 3.92 3.56

family (persons)
Income (yuan/p. a.) 7231.34 9977.16 12 554.46 15 976.35 26 939.30
Expenditure (yuan/p. a.) 4689.64 5713.08 6826.92 8179.94 12 460.32
Food 2620.64 2994.27 3357.31 3722.12 4821.34
Clothing 264.46 323.64 390.02 481.61 708.05
Medicine and Medical 268.14 326.48 376.53 456.64 718.12
Services
Transport, Post and 194.60 294.95 417.08 602.27 1144.18
Communication
Services
Education, Cultural and 453.70 634.38 802.11 967.89 1484.09
Recreation Services
Residence 594.43 761.91 1013.18 1317.71 2508.91
Miscellaneous 112.17 147.11 189.45 265.89 475.72
Commodities
and Services
or necessary amounts. In order to estimate marginal propensities to consume,

the ␤i, we need estimates of supernumerary income, that is, income over and
above that needed to buy the minimum in all groups, together.
The non-linearity in parameters could be handled by using some itera-
tive technique for simultaneous estimation or in two-step procedures. In
combinations of time series and cross-section samples it has been custom-
ary to estimate spending propensities from cross-section data (as Engel
curve slopes, for example) and price coefficients from time series of market
data. We have reversed the process by first estimating the ␥i (necessary or
minimal amounts consumed) directly from the cross-section data and the
marginal propensities from time series aggregates. This procedure is used,
in order to make use of external information. Also, limited data are avail-
able for the rural group over the time period selected for analysis.
At best, it would be desirable to have expertise on the composition of
minimum subsistence income, say, from specialists on social behavior and
needs – many countries have such expert information. Instead we used
information from Huiqing Gao and Liping Tao, who are confident that
they know plausible values for the minimal subsistence levels of family
budgets for base years, such as 1993, 1994 or 2002. The second income level
from the bottom (‘Lowest’) was chosen as the minimum subsistence budget
for urban consumers, while the bottom level (‘Low’) was chosen for rural
consumers. In each case, the average amount spent in each of the selected
income classes for each of the individual groupings (eight urban and eight
rural) served as direct estimates of ␥i. If ␥i are estimated in this simple,
straightforward way, then the remaining parameters can be computed from
the linear relationship between supernumerary expenditures and supernu-
merary income. The constraint on the sum of ␤i can readily be met if we
use total expenditure rather than total income in the regressor variable.
This is sensible, because total expenditure is to be preferred to total income,
since total expenditure is generally felt to be a better indicator of long-run
income, rather than direct measures of current period income, itself.
Having the ␥i and ␤i estimated, we then evaluate the ‘true’ cost of living
index from
n n
C ⌸ p␤it ⫺ 兺 ␥i pit
ii
ri i⫽1 i⫽1
r0 ⫽ n n
C ⌸
i⫽1
p␤i0ii ⫺ 兺␥i pi0
i⫽1
The Engel curves provide estimates of ␥i (for a minimal subsistence value

of total income) and regressions of supernumerary expenditure on super-
numerary income provide estimates of ␤i. The constants of integration are
Table 9.9 Percentage change, cost of living and official CPI
Urban Rural
Per capita Household Per capita Household
Cost of living 4.55 4.56 3.96 3.42
CPI 5.47 5.47 5.13 5.13
determined by substitution in the above expression for all values of vari-

ables designated at t ⫽0 (the base) and at the next adjacent year; that pro-
vides a formula for C. In other words, we find the value of C that makes the
above expression hold for an adjacent year, t ⫽⫾1, given our estimates of
␥i and ␤i. For example, in estimating values of the cost-of-living formula
for base year 1993, we evaluate C such that the formula holds for
8 8
r94
C ⌸
i⫽1
p␤i ⫺
i94 兺␥i pi94
i⫽1
r93 ⫽ 8 8
C ⌸ p␤ ⫺ 兺 ␥i pi93
i
i93
i⫽1 i⫽1
For rural as well as urban consumers in 2002, we compute estimates of

␥i for that year from the family budget table and then proceed to estimate
supernumerary Engel curves for 1993–2002. The formula for ri /r2002 is used
to compute C (at 2001), for rural consumers only and then rebase the whole
price series back to t ⫽1993, as for the ‘true’ cost-of-living index base year.4
The results can be summarized, as follows in Table 9.9:
The lower inflation rate on the basis of ‘true’ cost of living translates into
higher growth rates for real consumption. If the nominal consumption
figures were to be adjusted for a lower CPI inflation rate we would estimate
on a per capita basis higher real consumption growth (1993–2002) of 9.32
percent versus 7.65 percent. On a household basis the figures would be 9.57
percent versus 7.65 percent. These results combine both urban and rural
consumption together; so slightly less than two percentage points could be
added to the estimate of the country’s growth rate of real consumption.
This is not an estimate in a strict sense for real GDP growth but simply an
example that shows the underestimate of a large GDP complement on the
demand side.
As a crude check on these estimates, we make another estimate, urban
areas alone, on a per capita and a family basis (see Tables 9.10 and 9.11),
using the 2002 urban Engel curves, rebasing the time series of prices to
1993. For urban areas, we find the spread between the rate of growth of the
Table 9.10 Consumption expenditure (by per capita, 100 million yuan)
Year Deflated by CPI Deflated by ‘True’ Cost of Living

Total Urban Rural Total Urban Rural
consumption consumption consumption consumption consumption consumption
1993 15 682.40 7815.20 7867.20 15 682.40 7815.20 7867.20
1994 16 754.77 8401.20 8353.57 17 372.18 8748.44 8623.75
1995 18 518.03 9381.78 9136.25 19 849.04 10 157.92 9691.11
281
1996 20 399.18 9917.85 10 481.33 22 098.14 10 895.25 11 202.89

1997 21 508.85 10 635.38 10 873.48 23 599.33 11 887.34 11 711.99
1998 22 955.91 11 827.47 11 128.44 25 459.99 13 403.01 12 056.98
1999 24 791.37 13 186.25 11 605.11 27 875.81 15 198.75 12 677.07
2000 26 920.97 14 632.55 12 288.41 30 868.45 17 301.17 13 567.28
2001 28 587.02 15 690.91 12 896.10 32 826.60 18 566.56 14 260.04
2002 30 445.98 16 944.71 13 501.26 34 979.35 20 050.14 14 929.21
Average growth rate (%) 7.65 8.98 6.18 9.32 11.04 7.38
Table 9.11 Consumption expenditure (by family, 100 million yuan)
Year Deflated by CPI Deflated by ‘True’ Cost of Living

Total Urban Rural Total Urban Rural
consumption consumption consumption consumption consumption consumption
1993 15 682.40 7815.20 7867.20 15 682.40 7815.20 7867.20
1994 16 754.77 8401.20 8353.57 17 646.35 8753.97 8892.38
1995 18 518.03 9381.78 9136.25 20 528.49 10 167.29 10 361.20
282
1996 20 399.18 9917.85 10 481.33 22 941.82 10 904.00 12 037.83

1997 21 508.85 10 635.38 10 873.48 24 386.49 11 894.48 12 492.01
1998 22 955.91 11 827.47 11 128.44 26 195.16 13 408.07 12 787.09
1999 24 791.37 13 186.25 11 605.11 28 556.79 15 201.49 13 355.30
2000 26 920.97 14 632.55 12 288.41 31 521.56 17 299.18 14 222.38
2001 28 587.02 15 690.91 12 896.10 33 501.58 18 562.52 14 939.06
2002 30 445.98 16 944.71 13 501.26 35 685.86 20 045.78 15 640.08
Average growth rate (%) 7.65 8.98 6.18 9.57 11.03 7.93
CPI at 5.47 percent to be almost two percentage points higher than that of
the ‘true’ cost of living at 3.51 percent per capita and 3.52 percent family,
1993–2002, using the Engel curve for 2002 but re-basing the associated cost
of living index to 1993 (the initialization point for valuing the constant of
integration). These results are shown in Figures 9.9–9.14 and Tables
9.12–9.18. The larger spread for this validation-checking calculation would
imply a larger spread in the real growth rate of consumption.
Our procedures for estimating a bias in the consumer price index for
China lack some data series that would be needed in order to make a
definitive estimate of the increment to China’s economic growth rate, but
we can see that persistent overestimates of the price deflator for consump-
tion is indicated by our various approximations. We come to the conclusion
that one or two percentage points should be added to the growth rate.
Table 9.12 Cost of living and official CPI (urban, per capita, initialized at
1993, Engel curve at 2002)
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Average
Growth
Rate
CPI 1.00 1.25 1.46 1.59 1.64 1.63 1.61 1.62 1.63 1.61 5.47%
‘True’ 1.00 1.20 1.35 1.45 1.47 1.44 1.39 1.37 1.38 1.36 3.51%
cost of
living
1.70
1.60
1.50
Price index
1.40
1.30
1.20
1.10
CPI ‘True’ cost of living
1.00
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002
Figure 9.9 Cost of living and official CPI (urban, per capita, initialized at
Table 9.13 Cost of living and official CPI (urban, family, initialized at
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Average
Growth
Rate
CPI 1.00 1.25 1.46 1.59 1.64 1.63 1.61 1.62 1.63 1.61 5.47%
‘True’ 1.00 1.20 1.35 1.44 1.46 1.44 1.39 1.37 1.38 1.36 3.52%
cost of
living
1.70
1.60
1.50
Price index
1.40
1.30
1.20
1.10
1.00
1993199419951996199719981999200020012002
Figure 9.10 Cost of living and official CPI (urban, family, initialized at
Although we determine this for consumption, the same order of magnitude

ought to be considered close for GDP as well.
SOME COMMENTS ON THE LINEAR EXPENDITURE

SYSTEM
The linear expenditure system (LES) is, without doubt, a very special case
and, as Samuelson pointed out in 1947, not a general case of linearity,
such as in a first-order Taylor approximation. Yet the linear expenditure
system is not linear in unknown parameters nor in individual prices,
quantities, and income (total expenditures). Linearity depends on how the
non-stochastic specification is transformed and how stochastic properties
Table 9.14 Cost of living and official CPI (rural, per capita, initialized at
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Average
Growth
Rate
CPI 1.00 1.23 1.45 1.56 1.60 1.59 1.56 1.56 1.57 1.57 5.13%
‘True’ 1.00 1.20 1.37 1.46 1.49 1.47 1.43 1.41 1.42 1.42 3.96%
cost of
living
1.70
1.60
1.50
Price index
1.40
1.30
1.20
1.10
1.00
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002
Figure 9.11 Cost of living and official CPI (rural, per capita, initialized at
are introduced. It is linear in nominal supernumerary expenditures and

income.
It must be emphasized that our present interest in use of the LES is for
determining growth, and, in the important Chinese case, growth means
development since 1978; so the meaningful annual data samples are limited
in scope, and as China is very much a transition economy it means that
available data are not plentiful at a high degree of accuracy, although the
situation is rapidly changing for the better.
In this system, the LES will be examined in a related, but somewhat
different, setting, namely in the context of a cross-country sample built and
maintained at the University of Pennsylvania by our colleagues Robert
Summers, Alan Heston and (now deceased) Irving Kravis. In their 1985
volume on purchasing power parity (PPP) prices for evaluating GDP
across countries in comparable world price values, they made an extremely
Table 9.15 Cost of living and official CPI (rural, family, initialized at
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Average
Growth
Rate
CPI 1.00 1.23 1.45 1.56 1.60 1.59 1.56 1.56 1.57 1.57 5.13%
‘True’ 1.00 1.16 1.28 1.36 1.40 1.38 1.36 1.35 1.36 1.35 3.42%
cost of
living
1.70
1.60
1.50
Price index
1.40
1.30
1.20
1.10
1.00
93
95
97
99
01
19
19
19
19
20
Figure 9.12 Cost of living and official CPI (rural, family, initialized at
interesting application of LES. They started this analysis with the provoca-
tive question, ‘Are tastes the same the world round?’ This question is akin,
but not identical, to our preoccupation with measuring price changes in the
context of a constant level of utility over time. Their within-sample variation
comes from variability from country to country, while ours comes from vari-
ability from time period to time period within a given country – China. In
the Summers–Heston–Kravis study, expenditures (or consumption volumes,
in an aggregate sense) vary from country to country because incomes and
prices vary, while tastes are assumed to be fixed. They show that LES can be
written in three ways: (1) the dependent variable being expenditures, (2) the
dependent variable being quantity consumed, (3) the dependent variable
being budget share. They chose two forms (2) and (3), while we chose (1) for
our analysis. They used four demand categories (Food, Clothing, Shelter, All
others), while we chose eight demand categories.
Table 9.16 Cost of living and official CPI (urban, per capita, initialized
at 1993, Engel curve at 1993)
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Average
Growth
Rate
CPI 1.00 1.25 1.46 1.59 1.64 1.63 1.61 1.62 1.63 1.61 5.47%
‘True’ 1.00 1.25 1.46 1.58 1.60 1.57 1.52 1.50 1.50 1.49 4.55%
cost of
living
1.70
1.60
1.50
1.40
Price index
1.30
1.20
1.10
1.00
0.90 CPI ‘True’ cost of living
0.80
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002
Figure 9.13 Cost of living and official CPI (urban, per capita, initialized
at 1993, Engel curve at 1993)
For purposes of parameter estimation, either in a pure cross-section

sample or a pure time series sample, it is straightforward to carry out the
steps of statistical inference by non-linear means, often in iterated linear
steps, until convergence is attained. That is the approach used by
Summers, Heston and Kravis. By contrast, we have estimated LES by
step-wise linear methods, without iteration. We achieve this by use of a
priori information. In our case, the parameters obtained as estimates of
essential minimal consumption expenditure in a given year are based on
basic needs of a typical household. It is like the process used in many
countries for establishing subsistence budgets, by expenditure category.
Instead of using specialists’ estimates of a minimal budget, we used the
judgment of Gao and Tao (joint authors for this study) on the basis of
their knowledge of norms that they estimate at the State Information
Center. These are essentially microeconomic estimates in association with
Table 9.17 Cost of living and official CPI (urban, family, initialized at
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Average
Growth
Rate
CPI 1.00 1.25 1.46 1.59 1.64 1.63 1.61 1.62 1.63 1.61 5.47%
‘True’ 1.00 1.25 1.46 1.58 1.60 1.57 1.52 1.51 1.50 1.49 4.56%
cost of
living
1.70
1.60
1.50
1.40
Price index
1.30
1.20
1.10
1.00
0.90 CPI ‘True’ cost of living
0.80
93
94
95
96
97
98
99
00
01
02
19
19
19
19
19
19
19
20
20
20
Figure 9.14 Cost of living and official CPI (urban, family, initialized at
family budget data samples. The other parameters are then estimated,
given the minimum subsistence amounts, from the regression of supernu-
merary expenditure for each category on supernumerary income, which is
taken to be total expenditure, serving as a proxy for long-run income. It
seems reasonable to assume that the first set of parameter estimates does
not depend on the second set, which uses time series of macro concepts –
supernumerary expenditure, by categories of goods and services, in rela-
tion to total supernumerary income, from time series estimation. This is
a linear estimator, given the estimates of minimum essential expenditure,
on average, in the different groupings, over time periods in our sample of
family budget data.
A remarkable property of the Summers–Heston–Kravis estimates is that
the LES fits the data of the consumption estimate and their associated price
indexes on a cross-country sample. An interesting application of the ana-
Table 9.18 Urban consumption expenditure (1993 price, 100 million yuan)
Year Deflated by CPI Deflated by Deflated by

‘True’ Cost of ‘True’ Cost of
Living (per capita) Living (family)
1993 7815.20 7815.20 7815.20
1994 8401.20 8401.20 8401.20
1995 9381.78 9362.54 9368.95
1996 9917.85 9958.47 9971.08
1997 10 635.38 10 858.98 10 872.53
1998 11 827.47 12 255.82 12 271.45
1999 13 186.25 13 929.52 13 947.86
2000 14 632.55 15 809.67 15 830.79
2001 15 690.91 17 003.79 17 026.41
2002 16 944.71 18 324.98 18 349.56
Average growth 8.98 9.93 9.95
rate (%)
logue to the constant-utility property of the LES system in the context of

a cross-country analysis is the use of the estimated LES consumption
values to compute individual consumption values for each country. If the
system is used to rank per capita consumption levels, country by country
relative to the base country (the USA), the values listed in Table 9.19 below
refer to how much the expenditures of the various countries would have to
be increased in order to raise the level of satisfaction of the individual rep-
resentative consumers to that of the representative US consumer. In place
of maintenance of a given utility level over time variation, the results for
the PPP consumption estimates maintain the taste levels across country
variation.
The 34 countries in the sample used by Heston, Summers and Kravis
show remarkable correspondence between their PPP rankings (Geary-
Khamis) of per capita consumption and estimates of the same statistic
computed from an empirical linear expenditure system from the cross-
section sample of the same countries. The final results are all expressed in
relation to per capita consumption in the US.
The sample used for the above table does not include all the countries
chosen for the consumption tables (PPP) because some necessary data are
lacking for a number of countries. The layout in Table 9.19 groups coun-
tries roughly by stage of development, and is, in a sense, an example of the
usefulness of the LES system, which receives far less attention than does
the PPP system of country comparisons of GDP and its components. In
this case the component is consumption.
Table 9.19 Comparison of estimates of real per capita consumption

expenditures using Geary–Khamis formulas and LES formulas
(percentage of US per capita consumption 1975)
Geary-Khamis LES
Malawi 4.76 4.49
Kenya 6.09 6.67
India 5.86 6.51
Pakistan 7.81 8.31
Sri Lanka 8.75 10.50
Zambia 7.25 7.46
Thailand 12.79 11.52
Philippines 12.14 14.09
Korea 18.69 17.95
Malaysia 17.14 18.33
Colombia 22.21 22.23
Jamaica 23.92 26.70
Syria 23.66 23.22
Brazil 22.21 22.82
Romania 26.00 27.89
Mexico 33.66 37.04
Yugoslavia 31.78 34.68
Iran 24.64 25.12
Uruguay 39.92 36.83
Ireland 43.70 44.38
Hungary 42.81 44.09
Poland 39.63 40.12
Italy 49.05 53.19
Spain 57.70 58.90
UK 61.03 62.70
Japan 55.34 55.96
Austria 72.01 74.95
Netherlands 66.11 68.35
Belgium 71.90 73.99
France 71.58 74.37
Luxembourg 71.69 77.71
Denmark 75.48 74.16
Germany 72.98 74.98
US 100.00 100.00
It is also interesting that R.C. Geary, who is a co-author of the

Geary–Khamis method of making country comparisons for growth analy-
sis, also made an independent derivation of the underlying utility function
for the LES system. R.C. Geary derived the same utility function from
Richard Stone’s independent discovery of the LES system that P.A.
Samuelson obtained from the Klein–Rubin version.
In conclusion, we state the final comments of Summers, Heston and
Kravis as a piece of information supporting our use of the LES for evalu-
ation of China’s CPI; it too is a piece of understanding for the economic
growth of China. ‘The stochastic multilateral approach to international
comparisons embodied in the LES differs from the Geary–Khamis multi-
lateral approach, but still produces remarkably similar estimates’.5
ANOTHER APPROACH FOR ADJUSTMENT: VIA

LIFESTYLE CHANGE
There are different ways of making adjustments to the official published

price indexes. The changes made by the Boskin Committee in the United
States to recognize product quality improvement is known as the method of
hedonic price indexes. This has been introduced or investigated on a limited
scale for specific goods and services such as optional equipment devices for
motorcars, operational features of computers, or health care techniques.
The specific technical detail for a large fraction of consumer spending was
not available for our use in the present Chinese investigation, so we did not
attempt to corroborate our studies with hedonic calculations, but we did
attempt to measure aspects of lifestyle to see if we could estimate an index
adjustment by referring to a variety of indicators of life quality or lifestyle
in a wider sense. That is the objective of this section. In practice, hedonic
index adjustments focus on quality improvements, but there are obvious
cases of deterioration; that is a reason for not following that method.
In a sense we are looking for indicators of human development. The
United Nations Development Programme (UNDP) provides an annual
index made up of a real income measure, a literacy factor, longevity, and
educational enrollment. In ordinary price indexes, unit prices are weighted
by their associated quantities of goods and services. In a sense, these
weighting factors, once they are put into appropriate units reflect the
volume of economic activity, that is, their price valuations enable them to
aggregate into real values of national income or national expenditure, or
national production. They do not show the details of the composition of
production and, in particular, quality improvement, over time, or lifestyle
changes. As a single, uni-dimensional magnitude, life expectancy comes
close to telling us about quality of life. We, however, did not want such a
concentrated measure and looked for a variety of indicators, thus spread-
ing the risk of poor judgment.
In a series of research papers, Nagar and Basu, have made cross-country
estimates using the method of principal components to construct an index
of human development.6 Nagar and Basu recommend the construction of
an index in which each principal component (which is a linear combination
of indicators) is weighted by its eigenvalues.
Following the path laid out by Nagar and Basu, but using the time
dimension in a time series sample for China rather than country designa-
tion in a cross-country sample Suleyman Özmucur and Lawrence Klein
estimated five principal components from five variables:
● CO2 emissions (metric tons per capita) inverted;

● secondary school enrollments (%);
● physicians (per 1000 people);
● infant survival rate (per 100 live births);
● life expectancy at birth (years).
These data were prepared (1980–2000), separately, for USA and China,
interpolating values for specific years in which quantitative reports were not
published and the results showed growth of the weighted aggregates to be
0.33 percent for USA and 0.53 percent for China.
For comparison without cost of living price deflators for China (and the
USA for comparison) leads us to conclude that this source of variation shows
that quality improvement grew about 60 percent (20 percentage points) faster
in China than in the USA. This is about growth and not about levels.
There are some very important indicators for China that have not yet been
investigated. For the USA, however, some of these indicators have been
taken about as high as country performance could go. For example, illiter-
acy, in a practical sense, has been banished in the USA. Professor Teh-wei
Hu of the University of California School of Public Health has suggested
that we expand our list by adding data on access to tap water and public san-
itation facilities for insight into improvement of China’s quality of life.
In this chapter, we estimate the contributions of the following additional
variables to China’s principal component analysis: SO2 Emissions; Urban
Access to Tap Water; Urban Garbage Disposal; Urban Excrement and
Urine Disposal (tons per capita), and Graduates of Higher Educational
Institutions (per 1000 people). These variables are all measured for urban
China. In the case of the USA, they are not necessarily relevant. They are
year after year already at virtual saturation levels in the USA. So we have
made an analysis of China alone, to be compared with the previous values
for China, to see how much improvement, in an overall sense, has been
taking place between the two countries.
We have, however, introduced a particular variable, namely SO2 Emissions.
The variable CO2 Emissions, in an earlier study, measured possible environ-
mental deterioration, but not directly a cause of individual human health
problems; rather a cause of atmospheric conditions that could cause prob-
lems to existing cities or towns. CO2 Emissions are generally associated with
such problems as global warming, which could eventually cause human
suffering. SO2 Emissions are more likely to have direct effects on the health of
exposed people so we have experimented with the use of this variable for
China, either with or without simultaneous consideration of a CO2 variable.
Below, we list ten indicators, including the original five variables used in
the paper with Suleyman Özmucur (Klein and Özmucur, 2003):
1. Life Expectancy at Birth (years);

2. Infant Survival Rate (per 100 live births);
3. Physicians (per 1000 people);
4. Secondary School Enrollment (per 1000 people);
5. SO2 Emissions (tons per capita);
6. Urban Access to Tap Water (%);
7. Urban Garbage Disposal (per 1000 people);
8. Urban Excrement and Urine Disposal (per 1000 people);
9. Graduates of Higher Education (per 1000 people);
10. CO2 Emissions (tons per capita).
These are indicators of urban lifestyle change since data are not available
for rural persons and areas.
The sample covers 1980–2002. In Table 9.20, we show the end results for
annual growth, for an index computed as a ten-element weighted average,
which grows from 100 in 1980 to 115.47 in 2002. The 22-year increase com-
puted from
115.47 ⫽ 100.00 (1 ⫹ r) 22
indicates that the estimated rate of growth is
r ⫽ 0.65596%
This growth rate is nearly double that for the USA, 1980–2000. Thus the
adjustment to China’s growth rate could be as high as 1 percent or more,
because the quality adjustment for the USA case has been estimated at 0.32
in a confidence interval of –0.1 to 0.8 percentage points.
Table 9.20 Index of lifestyle: China (ten indicators)
1980 100.00 1992 109.28

1981 100.53 1993 110.06
1982 102.11 1994 111.26
1983 102.12 1995 112.87
1984 103.06 1996 113.74
1985 103.55 1997 113.62
1986 104.50 1998 113.82
1987 105.35 1999 114.21
1988 106.51 2000 114.99
1989 107.14 2001 115.88
1990 108.11 2002 115.47
1991 108.58 0.65596% p. a. growth rate
Table 9.21 Index of lifestyle, China (seven indicators)
1980 100.00 1992 106.67

1981 100.33 1993 107.00
1982 102.02 1994 107.92
1983 101.35 1995 109.04
1984 102.24 1996 109.78
1985 102.85 1997 110.15
1986 103.54 1998 110.67
1987 104.18 1999 111.66
1988 104.78 2000 112.71
1989 105.20 2001 113.87
1990 105.68 2002 114.41
1991 106.14 0.6138% p. a. growth rate
While this calculation confirms our cost-of-living degree of superiority

for China’s growth rate, it is not entirely satisfactory. At least three of the
variables considered (CO2, SO2 and urban excrement and urine disposal)
have some problematic characteristics. In the first three components, the
marginal values for CO2 and SO2 emissions show large positive coeffi-
cients, suggesting that higher values for CO2 and SO2 raise the index. It
may require some years for the effects of these emissions to have their full
negative effects on the environment and on personal health. All the indi-
cator variables have complicated direct and indirect effects, but the large
positive contributions leave us uneasy. Similarly, the disposal of urban
excrement and urine have significant, but not overwhelming, negative
effects.
If we delete these three indicators, and make a principal compo-

nent analysis of the seven remaining indicators, we find sensible pos-
itive values for all the vector elements in the first principal component,
with as much as 85 percent of the total variance of the seven remaining
indicators explained by the first component. The index in this case seems
to be more defensible. The results in Table 9.21 suggest a growth rate of
0.6138%, which is quite close to the growth rate of all ten indicators.
NOTES
1. See Klein and Özmucur (2002/03).

2. Advisory Committee to Study the CPI (2006).
3. Klein and Rubin (1947/48). Richard Stone (1954) independently discovered the same
system at a later date. Both R. Geary (1950) and P.A. Samuelson (1947/48) pointed out
the associated utility specifications.
4. It is to be hoped that in future research it will be possible to build up a whole time sequence
of Engel curves for rural consumers.
5. Summers et al. (1982).
6. Nagar and Basu (1999).
REFERENCES
Advisory Committee to Study the CPI (1996), Final Report of the Advisory
Committee to Study the Consumer Price Index, Chairman Michael J. Boskin,
Washington, DC, Government Printing Office. http://www.ssa.gov/history/
reports/boskinrpt.html.
Geary, R.C. (1950), ‘A Note on a Constant-utility Index of the Cost of Living’,
Review of Economic Studies, 18, 65–6.
Klein, L.R. and S. Özmucur (2002/03), ‘The Estimation of China’s Growth Rate’,
Journal of Economic and Social Measurement, 28, 187–202.
Klein, L.R. and H. Rubin (1947/48), ‘A Constant-utility Index of the Cost of
Living’, Review of Economic Studies, 15, 84–7.
Klein, L.R. and S. Özmucur (2003), ‘Estimates of Changes in Chinese Lifestyle’,
unpublished research paper, University of Pennsylvania.
Nagar, A.L. and Sudip Ranjan Basu (1999), ‘Weighting Socio-economic Indicators
of Human Development (A Latent Variable Approach)’, National Institute of
Public Finance and Policy, New Delhi.
Samuelson, P.A. (1947/48), ‘Some Implications of “Linearity” ’, Review of
Economic Studies, 15, 88–90.
Stone, Richard (1954), ‘Linear Expenditure Systems and Demand Analysis: An
Application to the Pattern of British Demand’, Economic Journal, 64, 511–27,
September.
Summers, Robert, Alan Heston and Irving Kravis (1982), ‘Interspatial Demand
Analysis’, World Product and Income, Baltimore, MD: Johns Hopkins University
Press, Ch. 9, p. 374.
10. Estimating China’s core
inflation rate*
Deming Wu
1 INTRODUCTION
The objective of this chapter is to measure the core inflation rate in China
between 1997 and 2002. An accurate and timely measurement of inflation
is a critical prerequisite for successful monetary policies. The conventional
price indices, however, may not accurately capture the underlying inflation
from a monetary policy point of view. Taking the consumer price index
(CPI) as an example, it measures inflation as experienced by consumers in
their day-to-day living expenses. However, the inflation from the con-
sumers’ perspectives may be different from the underlying inflation that
concerns monetary policy-makers for reasons discussed below.
First, the prices of certain goods, such as agricultural products and oil,
depend on factors such as weather and OPEC’s oil production and pricing
decisions, which are beyond the control of the central bank. Monetary
policy as a response to inflation or deflation due to these kinds of price
changes may be ineffective.
Second, some goods and services have very volatile prices whose move-
ments typically reverse direction quickly. It is inappropriate for monetary
policy to respond to these short-run fluctuations since it typically takes
between 12 and 18 months, based on the US experience, for a monetary
policy to begin to affect inflation significantly.
Third, some price changes are due to changes in the policy environment,
such as those in the tax and tariff rates, or due to the increased competition
in the distribution sector as a result of price reforms and trade liberaliza-
tion. It is not necessary to respond to these one-time changes as if they are
persistent.
Inflation as measured by the consumer price index is the sum of these
different kinds of price changes. However, not all movements in the general
price level are equally important from a monetary policy point of view.
Instead, the central bank should focus on the underlying, or the core inflation,
that is, the portion of inflation that can be affected through monetary policy.
296
Estimating China’s core inflation rate 297
Therefore, the focus of this research is to construct a measurement of the

core inflation rate for monetary policy purposes. This measurement of
the inflation rate should exclude the price changes due to factors beyond
the central bank’s control, temporary or one-time price changes, and cycli-
cal price changes.
This research was also motivated by widespread concerns about the
deflation threat among some economists and research organizations during
the period of 1997–2002 (Bernanke, 2002; De Long, 1999; Erceg et al.,
2002; Krugman, 1999a, 1999b, 2003; MF, 2003; Lau, 2003). In addition,
China has been criticized as not being a full stabilization factor in Asia
during the 1997–98 crisis. By measuring China’s core inflation rate,
however, I found that the expansionary monetary and fiscal policies in
China did play a stabilization role during the crisis.
This chapter is organized as follows. In section 2, I briefly review the
definitions of core inflation rate. Section 3 is a brief description of China’s
consumer price index by category. I also report the weights of individual
categories in the CPI basket. In section 4, I estimate China’s core inflation
rate by excluding the price changes of agricultural products and energy.
Section 5 estimates the impact of energy price changes. Section 6 reports an
alternative estimation of the core inflation rate using a structural condi-
tional error correction model with an unobserved variable. Section 7 con-
cludes.
2 DEFINITION OF CORE INFLATION
In this chapter, the core inflation is defined as the portion of inflation that
depends on monetary policy. This definition is based on the rationale that
not all movements in the general price level are due to changes in monetary
policy. Therefore, the central bank should focus on the part it can poten-
tially control. Based on this rationale, the core inflation rate is measured by
excluding agricultural products and energy, whose prices cannot be easily
controlled by the central bank.
It should be noted that there are different interpretations of core
inflation, which typically lead to different approaches to the measurement
of the core inflation rate (Wynne, 1999). One interpretation, proposed by
Okun (1970) and Flemming (1976), is that the observed inflation can be
decomposed into two components: the general price change and the rela-
tive price change due to supply-side shocks. Core inflation can be thought
of as the general price change that reflects the monetary policy. Therefore,
the measurement of the core inflation rate is to remove the relative price
changes from the observed inflation.
An alternative interpretation by Eckstein (1981) and Quah and Vahey

(1995) defines core inflation as a persistent component of the headline infl-
ation. According to this definition, the observed inflation includes a per-
sistent component and a temporary or transitory component. Therefore,
the measurement of the core inflation rate is to subtract the transitory
component from the observed inflation. For example, Eckstein (1981)
defines core inflation as a trend of production factor cost, which would
reflect the long-run inflation expectation. Quah and Vahey (1995) define
core inflation as the inflation that does not affect production in the long
run. They propose a two-variable SVAR (structural vector auto-regres-
sion) model to estimate the core inflation rate. Their estimation method
uses the identification condition that the core inflation shock does not have
long-run effect on the growth rate. Bjornland (2000) uses a three-variable
SVAR with oil price, GDP and CPI inflation to estimate the core inflation
rate with the identification conditions that the oil price is influenced only
by oil price shock, and the core inflation shock does not affect the growth
rate in the long run.
The most commonly used estimation methods for core inflation are the
methods of exclusion, which exclude some items from the basket of CPI.
These methods attempt to exclude temporary non-monetary changes of
prices from the observed inflation. On the other hand, there are different
model approaches that attempt to extract the unobserved component of
core inflation using different time-series-based estimation methods such as
SVAR and Kalman filter. For example, Lee et al. (2003) propose a struc-
tural conditional error correction model for CPI and core CPI (VECM
[vector error correction model]). In this model the core CPI indicator is an
unobserved variable that is cointegrated with the headline CPI. They
propose a state-space representation of this model, which is later estimated
using a Kalman filter.
It is important to remember that the definitions of core inflation in
different approaches are different. As a result, they do not measure the
same thing. Therefore, it is difficult to compare the measurement of core
inflation rate of one approach with another.
3 A DESCRIPTION OF CHINA’S CONSUMER PRICE

INDEX BY CATEGORY
The composition of China’s consumer price index underwent two major

revisions in 1994 and 2001. The revision in 2001 introduced a new cate-
gory called ‘Tobacco, Liquor and Articles’, and removed an existing cat-
egory called ‘Services’. As a result, the category definitions in the CPI
market basket from 1994 to 2000 are different from those after January
2001. Table 10.1 and Table 10.2 list the categories in the CPI basket before
and after 2001 respectively. The weights for calculating the consumer
price index are determined according to the composition of the con-
sumption expenditures of more than 90 000 urban and rural households.
Table 10.3 and Table 10.4 report the composition of per capita consump-
tion expenditures for urban and rural households respectively from 1997
to 2001.
As can be seen from these tables, in China, food-related costs comprise
more than 40 percent of the per capita consumption expenditures. As a
result, the weight for food-related costs in the CPI was above 40 percent
before 2001. Because of this, the majority of the fluctuation in CPI was due
to fluctuations of food prices. Changes in food prices in China depend on
natural factors, such as the weather and the government restrictions on
Table 10.1 CPI categories before 2001
Category 2000 Weights

1. Food 0.4011
2. Clothing 0.1136
3. Household Facilities and Articles 0.0857
4. Medicine and Medical Articles 0.0769
5. Transportation and Communication 0.0652
6. Recreation, Education and Culture Articles 0.1196
7. Residence 0.097
8. Services 0.05
Source: National Bureau of Statistics of China.
Table 10.2 CPI categories after 2001
Category 2002 Weights

1. Food 0.3359
2. Tobacco, Liquor and Articles 0.0447
3. Clothing 0.0875
4. Household Facilities and Articles 0.065
5. Medicine and Medical Articles 0.0912
6. Transportation and Communication 0.0814
7. Recreation, Education and Culture Articles 0.1436
8. Residences 0.1507
Source: National Bureau of Statistics of China.

Table 10.3 Composition of per capita expenditures for urban households
Item 1997 (%) 1998 (%) 1999 (%) 2000 (%) 2001 (%)
Total living expenditures 100.00 100.00 100.00 100.00 100.00
1. Food 46.41 44.48 41.86 39.18 37.94
1.1 Grain 5.69 5.24 4.67 3.77 3.54
1.2 Meat, Poultry and 10.98 9.96 8.85 8.23 7.79
Related Products
1.3 Eggs 1.76 1.55 1.42 1.13 1.07
1.4 Aquatic Products 3.37 3.29 3.12 2.87 2.86
1.5 Milk and Dairy 0.99 1.11 1.22 1.37 1.51
Products
2. Clothing 12.45 11.10 10.45 10.01 10.05
2.1 Garments 7.84 7.18 6.94 6.75 6.86
3. Household Facilities, 7.57 8.24 8.57 8.79 8.27
3.1 Durable Consumer 4.14 4.65 5.01 5.18 4.72
Goods
4. Medicine and 4.29 4.74 5.32 6.36 6.47
Medical Services
5. Transportation, Post and 5.56 5.94 6.73 7.90 8.61
5.1 Transportation 2.66 2.65 2.96 3.25 3.31
5.2 Communication 2.90 3.29 3.76 4.66 5.30
6. Recreation, Education 10.71 11.53 12.28 12.56 13.00
and Cultural Services
6.1 Durable Consumer Goods 2.69 2.91 2.93 2.94 2.62
for Recreational Use
6.2 Education 5.68 6.35 7.00 7.28 8.07
6.3 Recreation 2.35 2.27 2.35 2.34 2.30
7. Residence 8.57 9.43 9.84 10.01 10.32
7.1 Housing 3.55 3.99 4.24 4.03 4.09
7.2 Water, Electricity, 5.02 5.43 5.59 5.98 6.23
Fuel and Others
8. Miscellaneous 4.44 4.55 4.96 5.17 5.35
Commodities and Services
8.1 Personal Consumption 3.51 3.62 4.01 4.21 4.40
8.2 Other Commodities 0.43 0.41 0.43 0.41 0.41
8.3 Other Services 0.50 0.52 0.51 0.55 0.54
Source: China’s Statistical Yearbook.
food imports. Because these factors are beyond the central bank’s control,
in order to construct a measurement of the core inflation rate that is more
relevant from the central bank’s perspective, we need to exclude food prices
from CPI, as shown in Figure 10.1.
Table 10.4 Composition of per capita expenditure for rural households
Item 1997 (%) 1998 (%) 1999 (%) 2000 (%) 2001 (%)
Total living expenditure 100.00 100.00 100.00 100.00 100.00
Food 55.05 53.43 52.56 49.13 47.71
Clothing 6.77 6.17 5.83 5.75 5.67
Medicines and Medical 3.86 4.28 4.44 5.24 5.55
Services
Transportation, Post and 3.33 3.82 4.36 5.58 6.32
Telecommunication
Services
Cultural, Educational and 9.16 10.02 10.67 11.18 11.06
Recreational Articles
and Services
Residence 14.42 15.07 14.75 15.47 16.03
Other Commodities 2.12 2.07 2.18 3.14 3.24
and Services
Source: China’s Statistical Yearbook.
4 ESTIMATING CHINA’S CORE INFLATION RATE
4.1 Estimation of Core CPI Inflation Rate
The core CPI inflation rate is defined as the CPI inflation rate excluding food
and energy. In China’s CPI basket, there is no explicit category for energy.
There is a subcategory of ‘Water, Electricity and Fuel’ under the major cate-
gory of ‘Residences’. The weight for ‘Residence’ in the CPI is about 10 percent;
the weight for ‘Water, Electricity and Fuel’ is about 5 percent of the total CPI
calculation. Because of the lack of published data for the price index of the
‘Water, Fuel and Electricity’ subcategory in the CPI, I use the price index of
the ‘Fuel’ subcategory under the RPI (retail price index) basket as a proxy.
I also have to estimate the weight by using the composition of per capita
expenditure for urban and rural households. The expenditure shares of this
category for urban households were 5.02 percent in 1997 and 6.23 percent
in 2001; the expenditure share of this category for rural households was not
available. It can be imputed by assuming the ratio of this category to its
parent category is the same as that of the expenditure of urban households.
In the research, I used the expenditure share for urban households as the
approximation for this weight.
The inflation rate of core CPI is calculated using the following equations:
CPI ⫽ CPIfood * wfood ⫹ CPIenergy * wenergy

⫹ CPIcore*(1 ⫺ wfood ⫺ wenergy ) (10.1)
CPI ⫺ CPIfood*wfood ⫺ CPIenergy*wenergy

CPIcore ⫽ 1 ⫺ wfood ⫺ wenergy (10.2)
Table 10.5 and Figure 10.2 report the overall CPI inflation, the inflation
of the food price index, the calculated CPI inflation excluding food, and the
CPI inflation excluding food and energy from 1980 to 2003. As can be seen
from the data, for 1998 and 1999, the overall CPI inflation rate was nega-
tive, but this was due to the large decline of food prices. If food prices were
excluded from the basket, the core CPI inflation rate would have been
positive in 1998 and 1999. For the year 2002, even after excluding food
prices, the core CPI inflation rate was still negative. This indicates that the
CPI decline in 2002 was likely to have been caused by factors other than
food prices.
In Table 10.6 and Figure 10.3 I report the decomposition of the monthly
CPI inflation rate between 1994 and 2004. I am taking advantage of the
availability of monthly data because the annual data are somewhat mis-
leading in that they are averages of the monthly data; so that the monthly
positive and negative changes offset each other. As can be seen from Table
10.6 and Figure 10.3 the inflation rate of core CPI had been mostly posi-
tive until February 2002.
The decline of the consumer price index in 2002 was likely due to the
effects of China’s accession into the WTO and the reduction of the tariff
rate in 2002. According to the WTO rules, and the commitments China
made, China lowered the tariff rates of 5300 items in 2002, accounting for
73 percent of the total tariff items. The average overall tariff rate was cut
down from 15.3 percent to 12 percent; the average tariff rate of industrial
products was lowered to 11.6 percent; the average tariff rate of agricultural
products, excluding aquatic products, was lowered to 15.8 percent; the
average tariff rate of aquatic products was lowered to 14.3 percent. Lower
tariff rates caused a large price reduction in imported goods, adding to the
downward pressure on domestic price levels. The large increase of imported
goods has also intensified the competition between imported and domestic-
ally produced goods. Domestic enterprises had to reduce their prices and
costs in order to preserve their market shares.
The slowdown of economic growth in 2001 might also have con-
tributed to the downward pressure on prices in 2002. Price changes
35.0
30.0
25.0
20.0
15.0
Percent
10.0
303
5.0
0.0
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
00
01
02
03
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
20
20
20
20
–5.0
–10.0
Year
CPI Inflation, all items CPI Inflation, all items less food CPI inflation, food
Figure 10.1 Decomposition of annual CPI inflation in China, 1980–2003

Table 10.5 Decomposition of annual CPI inflation 1980–2003
Year CPI CPI CPI CPI CPI

Inflation, Inflation, Inflation, All Inflation, Inflation,
All Items All Items Items Less Food Food Energy
Less Food and Energy
1980 6.0 2.1 2.2 10.5 0.7
1981 2.4 1.3 1.3 3.7 0.6
1982 1.9 1.1 1.2 2.8 0.8
1983 1.5 0.7 0.7 2.4 1.0
1984 2.8 3.0 3.1 2.6 2.2
1985 8.8 4.0 3.9 14.4 4.0
1986 6.5 5.7 5.9 7.4 3.6
1987 7.3 4.9 5.0 10.1 3.9
1988 18.8 15.2 15.1 23.0 16.1
1989 18.0 19.6 18.7 16.2 27.4
1990 3.1 5.5 5.2 0.3 8.2
1991 3.4 3.5 2.2 3.3 15.6
1992 6.4 5.3 4.3 7.7 14.6
1993 14.7 15.0 13.0 14.3 35.0
1994 24.1 17.4 17.7 31.8 15.1
1995 17.1 12.1 12.6 22.9 7.4
1996 8.3 8.9 9.3 7.6 5.0
1997 2.8 5.3 5.1 ⫺0.1 7.3
1998 ⫺0.8 1.3 1.8 ⫺3.2 ⫺3.9
1999 ⫺1.4 1.0 1.1 ⫺4.2 0.4
2000 0.4 3.0 1.5 ⫺2.6 17.7
2001 0.7 1.1 0.9 0.0 2.4
2002 ⫺0.8 ⫺0.9 ⫺1.2 ⫺0.6 2.0
2003 1.2 0.1 ⫺0.9 3.4 9.5
Weights (1997), 1.0000 0.5359 0.4857 0.4641 0.0502
1980–2000
Weights (2002), 1.0000 0.6641 0.6018 0.3359 0.0623
2001–2003
typically lag the GDP from half one year to one year. The GDP growth
rates for the four quarters in 2001 were 8.1 percent, 7.8 percent, 7.0
percent, and 6.6 percent consecutively, declining quarter by quarter.
This slowdown was reflected in price movement in 2002. In addition,
there was a substantial decline of ex-factor prices and the purchase price
of raw materials in 2001, which had a strong lagged influence on
retail prices in 2002. Finally, the decline of stock prices in 2002
might have also contributed to the decline in the price of investment and
consumer goods.
40.0
35.0
30.0
25.0
20.0
Percent
15.0
305
10.0
5.0
0.0
0
3
8
0
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
20
20
20
20
–5.0
–10.0
Year
CPI Inflation, all items CPI inflation, all items less food and energy CPI inflation, food CPI inflation, energy
Figure 10.2 CPI Inflation, all items less food and energy, 1980–2003
Table 10.6 Decomposition of monthly CPI inflation, China,

01/1994–04/2004
Month CPI CPI CPI CPI CPI

Inflation, Inflation, All Inflation, Inflation, Inflation,
All Items Items Less Food All Items Food Energy
and Energy Less Food
Jan-94 23.2 22.1 22.0 24.6 20.9
Feb-94 25.9 21.1 20.9 31.7 19.1
Mar-94 24.5 20.9 20.6 29.0 17.4
Apr-94 23.2 20.6 20.3 26.6 16.5
May-94 21.8 19.3 18.9 25.1 15.2
Jun-94 22.7 19.1 18.3 27.8 10.1
Jul-94 24.2 18.1 17.5 31.9 12.1
Aug-94 27.1 18.3 17.8 37.8 13.4
Sep-94 27.5 18.5 18.0 38.5 13.0
Oct-94 27.0 18.3 17.8 37.6 13.0
Nov-94 26.4 18.1 17.7 36.5 13.4
Dec-94 24.1 17.5 17.0 32.3 12.4
Jan-95 22.5 14.5 14.0 32.3 9.1
Feb-95 20.4 13.5 13.1 28.8 9.4
Mar-95 19.4 12.6 12.2 27.7 8.5
Apr-95 19.7 12.8 12.4 28.1 8.4
May-95 19.7 13.0 12.5 28.0 7.7
Jun-95 18.1 13.0 12.5 24.6 7.3
Jul-95 16.7 13.0 12.5 21.6 7.1
Aug-95 14.1 12.4 11.8 16.8 5.7
Sep-95 12.9 10.6 10.1 16.1 5.9
Oct-95 11.9 10.4 10.0 14.1 5.8
Nov-95 12.1 10.9 10.4 14.1 5.1
Dec-95 11.1 11.1 10.5 11.8 4.7
Jan-96 9.0 10.2 9.7 8.2 4.6
Feb-96 9.3 9.8 9.3 9.3 4.3
Mar-96 9.8 10.2 9.6 10.0 4.3
Apr-96 9.7 9.9 9.4 10.0 4.9
May-96 8.9 9.7 9.2 8.5 4.4
Jun-96 8.6 9.7 9.2 7.9 4.4
Jul-96 8.3 9.2 8.7 7.8 4.1
Aug-96 8.1 9.0 8.5 7.6 3.9
Sep-96 7.4 9.2 8.7 5.9 4.3
Oct-96 7.0 8.5 8.1 5.7 4.9
Nov-96 6.9 8.4 8.2 5.4 6.3
Dec-96 7.0 8.7 8.7 5.0 9.4
Jan-97 5.9 7.5 7.8 3.7 10.6
Feb-97 5.6 7.1 7.4 3.5 10.6
Mar-97 4.0 6.1 6.4 1.2 9.6
Apr-97 3.2 5.6 5.9 0.1 8.4

May-97 2.8 5.2 5.5 ⫺0.3 8.0
Jun-97 2.8 5.0 5.2 0.0 7.3
Jul-97 2.7 4.8 5.0 0.0 7.5
Aug-97 1.9 4.5 4.8 ⫺1.4 7.4
Sep-97 1.8 4.2 4.4 ⫺1.2 6.7
Oct-97 1.5 4.2 4.4 ⫺1.8 6.2
Nov-97 1.1 3.9 4.0 ⫺2.2 4.4
Dec-97 0.4 3.5 3.3 ⫺2.9 0.8
Jan-98 0.3 3.0 2.6 ⫺2.3 ⫺1.5
Feb-98 ⫺0.1 2.7 2.3 ⫺2.9 ⫺1.7
Mar-98 0.7 2.4 2.0 ⫺0.8 ⫺2.0
Apr-98 ⫺0.3 2.3 1.9 ⫺2.8 ⫺2.8
May-98 ⫺1.0 2.1 1.5 ⫺3.9 ⫺3.9
Jun-98 ⫺1.3 1.8 1.2 ⫺4.2 ⫺4.3
Jul-98 ⫺1.4 1.7 1.0 ⫺4.2 ⫺5.3
Aug-98 ⫺1.4 1.6 0.9 ⫺4.1 ⫺5.7
Sep-98 ⫺1.5 1.2 0.6 ⫺3.9 ⫺5.5
Oct-98 ⫺1.1 1.3 0.7 ⫺3.2 ⫺5.2
Nov-98 ⫺1.2 1.1 0.6 ⫺3.3 ⫺4.3
Dec-98 ⫺1.0 1.1 0.6 ⫺2.8 ⫺4.2
Jan-99 ⫺1.2 1.4 0.9 ⫺3.6 ⫺4.2
Feb-99 ⫺1.3 1.3 0.8 ⫺3.7 ⫺4.5
Mar-99 ⫺1.8 1.5 0.9 ⫺4.9 ⫺4.6
Apr-99 ⫺2.2 1.1 0.7 ⫺5.5 ⫺3.9
May-99 ⫺2.2 1.0 0.6 ⫺5.4 ⫺3.1
Jun-99 ⫺2.1 0.8 0.5 ⫺5.1 ⫺2.4
Jul-99 ⫺1.4 0.5 0.4 ⫺3.5 ⫺0.1
Aug-99 ⫺1.3 0.2 0.5 ⫺3.4 3.7
Sep-99 ⫺0.8 1.3 1.7 ⫺3.7 5.3
Oct-99 ⫺0.6 1.4 1.7 ⫺3.3 5.2
Nov-99 ⫺0.9 1.3 1.7 ⫺3.9 5.6
Dec-99 ⫺1.0 1.3 1.9 ⫺4.4 7.9
Jan-00 ⫺0.2 1.5 2.3 ⫺3.1 10.3
Feb-00 0.7 1.7 2.5 ⫺1.4 10.9
Mar-00 ⫺0.2 1.4 2.5 ⫺3.3 12.5
Apr-00 ⫺0.3 1.7 2.8 ⫺3.9 13.9
May-00 0.1 1.3 2.9 ⫺3.1 17.9
Jun-00 0.5 1.0 3.0 ⫺2.4 22.3
Jul-00 0.5 0.9 3.0 ⫺2.4 23.4
Aug-00 0.3 0.8 2.8 ⫺2.6 22.0

Sep-00 0.0 1.2 3.2 ⫺3.7 22.6
Oct-00 0.0 1.3 3.2 ⫺3.7 21.3
Nov-00 1.3 1.6 3.3 ⫺1.0 19.4
Dec-00 1.5 1.5 2.9 ⫺0.1 16.2
Jan-01 1.2 0.8 2.0 ⫺0.4 13.6
Feb-01 0.0 0.7 1.7 ⫺3.3 10.9
Mar-01 0.8 0.7 1.7 ⫺1.0 11.5
Apr-01 1.6 1.0 1.8 1.3 8.7
May-01 1.7 1.3 1.7 1.7 5.7
Jun-01 1.4 1.4 1.6 1.0 3.3
Jul-01 1.5 1.9 1.5 1.5 ⫺2.1
Aug-01 1.0 2.1 1.4 0.2 ⫺5.2
Sep-01 ⫺0.1 0.2 ⫺0.3 0.3 ⫺5.5
Oct-01 0.2 0.3 ⫺0.1 0.8 ⫺3.7
Nov-01 ⫺0.3 0.3 0.0 ⫺0.8 ⫺3.3
Dec-01 ⫺0.3 0.5 0.0 ⫺0.8 ⫺5.0
Jan-02 ⫺1.0 0.1 ⫺0.5 ⫺1.9 ⫺6.8
Feb-02 0.0 0.3 ⫺0.3 0.6 ⫺6.2
Mar-02 ⫺0.8 ⫺0.4 ⫺0.9 ⫺0.6 ⫺5.8
Apr-02 ⫺1.3 ⫺0.9 ⫺0.9 ⫺2.0 ⫺1.6
May-02 ⫺1.1 ⫺1.0 ⫺0.9 ⫺1.5 0.3
Jun-02 ⫺0.8 ⫺1.4 ⫺1.1 ⫺0.2 1.4
Jul-02 ⫺0.9 ⫺1.3 ⫺0.9 ⫺0.9 3.3
Aug-02 ⫺0.7 ⫺1.5 ⫺0.9 ⫺0.4 5.6
Sep-02 ⫺0.7 ⫺1.8 ⫺1.0 ⫺0.2 6.8
Oct-02 ⫺0.8 ⫺1.8 ⫺1.0 ⫺0.5 7.6
Nov-02 ⫺0.7 ⫺2.0 ⫺1.0 ⫺0.1 8.4
Dec-02 ⫺0.4 ⫺2.1 ⫺0.9 0.5 11.6
Jan-03 0.4 ⫺2.6 ⫺0.6 2.4 18.1
Feb-03 0.2 ⫺2.7 ⫺0.6 1.8 19.1
Mar-03 0.9 ⫺2.3 ⫺0.3 3.2 19.3
Apr-03 1.0 ⫺1.6 ⫺0.1 3.2 14.4
May-03 0.7 ⫺0.7 0.1 1.9 8.1
Jun-03 0.3 ⫺0.5 0.2 0.4 7.5
Jul-03 0.5 ⫺0.5 0.2 1.0 7.0
Aug-03 0.9 ⫺0.4 0.2 2.2 6.2
Sep-03 1.1 ⫺0.4 0.0 3.2 4.6
Oct-03 1.8 0.0 0.1 5.1 1.6
Nov-03 3.0 0.2 0.4 8.1 2.5
Dec-03 3.2 ⫺0.1 0.5 8.6 5.6
Jan-04 3.2 ⫺0.6 0.8 8.0 5.6

Feb-04 2.1 0.3 0.3 5.6 3.0
Mar-04 3.0 0.2 0.5 7.9 3.2
Apr-04 3.8 0.1 0.6 10.2 4.9
Weights (1997), 1 0.4857 0.5359 0.4641 0.0502
1994–2000
Weights (2002), 1 0.6018 0.6641 0.3359 0.0623
2001–2004
4.2 Decomposition of RPI
Using the monthly retail price index (RPI) by category, I estimate the core
rate of inflation as measured by the RPI by excluding the food and fuel cat-
egories. Because of the lack of published data for the weights of categories
in the RPI baskets, I regress the overall RPI against the indices of subcat-
egories to estimate the weights. To resolve the multicollinearity problem of
highly intercorrelated indices of subcategories, I have to remove some cat-
egories from the right-hand side in order to get plausible estimated results.
Table 10.7 reports the estimated weights of the major categories in the RPI
basket for different periods. Using the estimated weights for the period
between 04/1994 and 05/2003, Table 10.8 reports the decomposition of the
inflation rate of the RPI by excluding food and energy.
As can be seen from Figures 10.3 and 10.4, one striking difference
between the decomposition of RPI and that of CPI is that the inflation rate
of RPI excluding food and energy has been consistently negative since
October 1997. One possible explanation for the decline of the retail price
index is the increased competition in the distribution sector, which has
squeezed the gross margins of the wholesalers and retailers, and thus has
put downward pressure on retail prices (Wu, 2004).
5 ESTIMATING THE IMPACT OF CHANGES IN

PETROLEUM PRICES
In this section, I will assess the impact of changes in petroleum prices on

China’s domestic price levels. Although the weight of energy in the CPI
25.0
20.0
15.0
10.0
Percent
5.0
310
0.0
97 97 97 97 98 98 98 98 99 99 99 99 00 00 00 00 01 01 01 01 02 02 02 02 03 03 03 03 04 04
n- pr- ul- ct- an- pr- ul- ct- an- pr- ul- ct- an- pr- ul- ct- an- pr- ul- ct- an- pr- ul- ct- an- pr- ul- ct- an- pr-
Ja A J O J A J O J A J O J A J O J A J O J A J O J A J O J A
–5.0
–10.0
Month
CPI inflation, all items CPI inflation, all item less food and energy CPI inflation, food CPI inflation, energy
Figure 10.3 Decomposition of monthly CPI inflation, 01/1997–04/2004, China

Table 10.7 Estimated weights for categories in RPI basket
Category 01/94– 01/94– 01/97– 01/99– 01/00–

12/00 (%) 05/03 (%) 05/03 (%) 05/03 (%) 05/03 (%)
1. Food 41.09 40.69 36.58 37.92 37.82
3. Clothes, Shoes 8.86 9.27 12.57 14.44 7.46
and Hats
5. Medicines 0.19 2.31 4.38 4.38 5.91
7. Books and Magazines 1.91 1.91 3.25 2.37 2.34
8. Culture and Sports 7.23 7.77 1.70 1.70 6.21
9. Daily Used Articles 16.83 13.78 20.45 20.50 15.12
10. Household Electrical 10.54 5.19 3.64 3.33 1.60
Products
12. Fuel 3.60 3.34 2.86 2.72 3.04
13. Construction and 2.97 3.94 3.29 4.28 7.97
Decorative Materials
14. Mechanical and 4.30 5.93 6.67 5.87 4.50
Electrical Products
basket is small, the indirect impact of oil price on domestic price levels
could be large because energy cost accounts for a significant portion of
the production cost in every industrial sector. Therefore, the simple index
accounting approach used in the previous sections may underestimate the
impact of petroleum price changes. In light of this limitation, I construct
an input–output price model to estimate both the direct and indirect
impact of changes in petroleum prices on the price levels.
5.1 An Input–Output Price Model for Petroleum Price Changes
In this section I derive a modified version of the Leontief input–output

price model (Leontief, 1986). The derivation of this model is similar to that
of Melvin’s (1979). In the discussion below, upper-case variables denote the
price level, and their corresponding lower-case variables denote price
changes in percentages. Superscript T denotes the transpose of a matrix or
vector. By default, vectors are represented as column-vectors, and row-
vectors are represented using transpose.
Writing the condition that total cost equals total revenue, we have, for the
j-th industry,
Xj Pj ⫽ X1jP1 ⫹ X2jP2 ⫹ . . . ⫹ XnjPn ⫹ Vj (10.3)

Table 10.8 Decomposition of the monthly RPI inflation, 1996–2002, China
Month RPI Inflation, RPI Inflation, RPI Inflation, RPI Inflation,

Overall Excluding Food Fuel
Food and Fuel
Jan-96 7.6 6.9 8.8 4.6
Feb-96 7.7 6.4 9.8 4.3
Mar-96 7.7 5.8 10.6 4.3
Apr-96 7.4 5.4 10.4 4.9
May-96 6.5 5.2 8.4 4.4
Jun-96 5.9 4.8 7.6 4.4
Jul-96 5.8 4.5 7.7 4.1
Aug-96 5.8 4.5 7.8 3.9
Sep-96 5 4.4 5.9 4.3
Oct-96 4.7 4.0 5.6 4.9
Nov-96 4.6 4.1 5.2 6.3
Dec-96 4.4 3.8 4.8 9.4
Jan-97 3.3 2.7 3.5 10.6
Feb-97 2.9 2.2 3.2 10.6
Mar-97 1.7 2.0 0.7 9.6
Apr-97 1.1 1.7 ⫺0.3 8.4
May-97 0.8 1.4 ⫺0.6 8
Jun-97 0.8 1.1 ⫺0.1 7.3
Jul-97 0.6 0.6 0 7.5
Aug-97 1.1 2.3 ⫺1.1 7.4
Sep-97 0 0.3 ⫺1 6.7
Oct-97 ⫺0.4 0.2 ⫺1.7 6.2
Nov-97 ⫺0.8 ⫺0.2 ⫺2.1 4.4
Dec-97 ⫺1.2 ⫺0.2 ⫺2.8 0.8
Jan-98 ⫺1.5 ⫺0.9 ⫺2.3 ⫺1.5
Feb-98 ⫺1.9 ⫺1.2 ⫺2.9 ⫺1.7
Mar-98 ⫺1.2 ⫺1.6 ⫺0.6 ⫺2
Apr-98 ⫺2.1 ⫺1.5 ⫺2.8 ⫺2.8
May-98 ⫺2.7 ⫺1.8 ⫺3.8 ⫺3.9
Jun-98 ⫺3 ⫺2.1 ⫺4.1 ⫺4.3
Jul-98 ⫺3.2 ⫺2.1 ⫺4.5 ⫺5.3
Aug-98 ⫺3.3 ⫺2.4 ⫺4.4 ⫺5.7
Sep-98 ⫺3.3 ⫺2.6 ⫺4.1 ⫺5.5
Oct-98 ⫺2.9 ⫺2.4 ⫺3.4 ⫺5.2
Nov-98 ⫺2.8 ⫺2.3 ⫺3.3 ⫺4.3
Dec-98 ⫺2.7 ⫺2.6 ⫺2.7 ⫺4.2
Jan-99 ⫺2.8 ⫺2.2 ⫺3.5 ⫺4.2
Feb-99 ⫺2.8 ⫺2.1 ⫺3.6 ⫺4.5
Mar-99 ⫺3.2 ⫺2.0 ⫺4.8 ⫺4.6

Food and Fuel
Apr-99 ⫺3.5 ⫺2.1 ⫺5.4 ⫺3.9
May-99 ⫺3.5 ⫺2.1 ⫺5.5 ⫺3.1
Jun-99 ⫺3.4 ⫺2.2 ⫺5.1 ⫺2.4
Jul-99 ⫺2.6 ⫺2.4 ⫺3.1 ⫺0.1
Aug-99 ⫺2.6 ⫺2.7 ⫺3 3.7
Sep-99 ⫺2.8 ⫺2.5 ⫺3.9 5.3
Oct-99 ⫺2.6 ⫺2.6 ⫺3.3 5.2
Nov-99 ⫺2.8 ⫺2.4 ⫺4 5.6
Dec-99 ⫺3 ⫺2.4 ⫺4.7 7.9
Jan-00 ⫺2.1 ⫺2.3 ⫺2.9 10.3
Feb-00 ⫺1.4 ⫺2.5 ⫺0.9 10.9
Mar-00 ⫺2.1 ⫺2.2 ⫺3.1 12.5
Apr-00 ⫺2.4 ⫺2.2 ⫺4 13.9
May-00 ⫺1.9 ⫺2.1 ⫺3.3 17.9
Jun-00 ⫺1.4 ⫺2.1 ⫺2.4 22.3
Jul-00 ⫺1.2 ⫺1.7 ⫺2.5 23.4
Aug-00 ⫺1.3 ⫺1.8 ⫺2.5 22
Sep-00 ⫺1.5 ⫺1.3 ⫺3.7 22.6
Oct-00 ⫺1.7 ⫺1.3 ⫺4.1 21.3
Nov-00 ⫺0.6 ⫺1.7 ⫺0.7 19.4
Dec-00 ⫺0.4 ⫺1.8 0.2 16.2
Jan-01 ⫺0.5 ⫺1.6 ⫺0.1 13.6
Feb-01 ⫺1.7 ⫺1.6 ⫺2.9 10.9
Mar-01 ⫺0.9 ⫺1.9 ⫺0.6 11.5
Apr-01 0.1 ⫺2.0 2.3 8.7
May-01 0.1 ⫺2.1 2.7 5.7
Jun-01 ⫺0.1 ⫺2.0 2.2 3.3
Jul-01 ⫺0.2 ⫺2.2 2.7 ⫺2.1
Aug-01 ⫺1.2 ⫺2.3 0.6 ⫺5.2
Sep-01 ⫺1.3 ⫺2.6 0.8 ⫺5.5
Oct-01 ⫺0.8 ⫺2.4 1.6 ⫺3.7
Nov-01 ⫺1.7 ⫺2.3 ⫺0.8 ⫺3.3
Dec-01 ⫺1.8 ⫺2.3 ⫺0.8 ⫺5
Jan-02 ⫺2 ⫺2.2 ⫺1.3 ⫺6.8
Feb-02 ⫺1.1 ⫺2.5 1.3 ⫺6.2
Mar-02 ⫺1.8 ⫺2.4 ⫺0.6 ⫺5.8
Apr-02 ⫺2.1 ⫺2.1 ⫺2.1 ⫺1.6
May-02 ⫺1.8 ⫺2.2 ⫺1.4 0.3
Jun-02 ⫺0.8 ⫺2.5 1.3 1.4

Food and Fuel
Jul-02 ⫺1.1 ⫺2.4 0.3 3.3
Aug-02 ⫺1.1 ⫺2.2 ⫺0.1 5.6
Sep-02 ⫺1 ⫺2.3 0.1 6.8
Oct-02 ⫺1.2 ⫺2.2 ⫺0.6 7.6
Nov-02 ⫺0.9 ⫺2.4 0.4 8.4
Dec-02 ⫺0.3 ⫺2.2 1.3 11.6
Where Pj is the price of commodity j; andVj is the value-added of com-

modity j. We can divide this equation by Xj Pj to obtain
1 ⫽ a1, j ⫹ a2, j ⫹ . . . ⫹ an, j ⫹ vj (10.4)
Here, a1, j, a2, j, . . . , an, j are the input–output coefficients, and vj is the share
of valued-added in commodity j.
Now assume sector n is the petroleum industry. Suppose the price of
petroleum changes by p⬘n , the direct impact of petroleum price changes on
the price of commodity j is
1 ⫹ p⬘j ⫽ a1, j ⫹ a2, j ⫹ . . . ⫹ an, j (1 ⫹ p⬘n ) ⫹ vj (10.5)
That is,
p⬘j ⫽ an, j p⬘n (10.6)
Here, we use the lower-case p⬘j to represent the percentage of price changes.
However, because prices of other inputs also change as a result of the
change in petroleum prices, we need to consider both the direct and the
indirect effects, therefore, we have
1 ⫹ pj ⫽ a1, j (1 ⫹ p1 ) ⫹ a2, j (1 ⫹ p2 ) ⫹ . . . ⫹ an, j (1 ⫹ pn ) ⫹ vj (10.7)
And the final price changes are
pj ⫽ a1, j p1 ⫹ a2, j p2 ⫹ . . . ⫹ an, j pn (10.8)

25
20
15
10
Percent
5
315
0
96 96 96 96 97 97 97 97 98 98 98 98 99 99 99 99 00 00 00 00 01 01 01 01 02 02 02 02
n- pr- ul- ct- an- pr- ul- ct- an- pr- ul- ct- an- pr- ul- ct- an- pr- ul- ct- an- pr- ul- ct- an- pr- ul- ct-
Ja A J O J A J O J A J O J A J O J A J O J A J O J A J O
–5
–10
Month
RPI inflation, all items RPI inflation, all items less food and fuel RPI inflation, food RPI inflation, fuel
Figure 10.4 Decomposition of monthly RPI inflation, 1996–2002, China

Because there are similar equations for other sectors, we can write this
total system in matrix notation to obtain
p ⫽ ATn⫺1p ⫹ Anpn (10.9)
Here we have
冢冣
a1,1 ... an ⫺1,1
ATn⫺1 ⫽ (10.10)
...
...
..
.
a1,n ⫺1 ... an ⫺1,n ⫺1
and
冢冣
an, 1
an, 2
An ⫽ (10.11)
...
an, n ⫺1
and
冢冣
p1
p2
p⫽ (10.12)
...
pn ⫺1
Here, An⫺1 is the upper-left (n⫺1) ⫻ (n⫺1) sub-matrix of the input–output

coefficient matrix of the intermediate input, A. And An is a (n⫺1) ⫻ 1
vector, derived from the first n⫺1 elements of n-th column vector of AT.
Solving this system gives
p ⫽ [I ⫺ ATn⫺1 ] ⫺1Anpn (10.13)
5.2 Estimation Results
Table 10.9 reports the estimated impact of a 1 percent change in petroleum

prices using a 40-sector input–output table of China in 1997. As the table
shows, a 1 percent decline in petroleum prices will cause the consumer price
index to decline by 0.06 percent. In 2001, the world market oil price
declined by 14 percent as shown in Table 10.10. Because there is time lag
Table 10.9 Estimated impact of a 1 percent change in petroleum price
Code Name 2001–2002

Change in petroleum prices, % ⫺1 ⫺14.02
CPI CPI ⫺0.06 ⫺0.82
GDPI GDP deflator ⫺0.08 ⫺1.09
01 Agriculture ⫺0.02 ⫺0.33
02 Coal mining and processing ⫺0.04 ⫺0.53
03 Crude petroleum and natural gas products ⫺1.00 ⫺14.02
04 Metal ore mining ⫺0.05 ⫺0.73
05 Non-ferrous mineral mining ⫺0.07 ⫺0.95
06 Manufacture of food products and ⫺0.02 ⫺0.33
tobacco processing
07 Textile goods ⫺0.03 ⫺0.36
08 Wearing apparel, leather, furs, down and ⫺0.02 ⫺0.33
related products
09 Sawmills and furniture ⫺0.03 ⫺0.47
10 Paper and products, printing and record ⫺0.03 ⫺0.45
media reproduction
11 Petroleum processing and coking ⫺0.07 ⫺1.01
12 Chemicals ⫺1.00 ⫺14.02
13 Non-metal mineral products ⫺0.06 ⫺0.91
14 Metal smelting and pressing ⫺0.09 ⫺1.24
15 Metal products ⫺0.06 ⫺0.79
16 Machinery and equipment ⫺0.05 ⫺0.67
17 Transportation equipment ⫺0.05 ⫺0.65
18 Electric equipment and machinery ⫺0.06 ⫺0.78
19 Electronic and telecommunication ⫺0.04 ⫺0.53
equipment
20 Instruments, meters, cultural and ⫺0.04 ⫺0.56
office machinery
21 Maintenance and repair of machine ⫺0.04 ⫺0.58
and equipment
22 Other manufacturing products ⫺0.04 ⫺0.55
23 Scrap and waste 0.00 0.00
24 Electricity, steam and hot water ⫺0.21 ⫺2.88
production and supply
25 Gas production and supply ⫺0.02 ⫺0.27
26 Water production and supply ⫺1.00 ⫺14.02
27 Construction ⫺0.07 ⫺0.97
28 Transportation and warehousing ⫺0.12 ⫺1.72
29 Post and telecommunication ⫺0.03 ⫺0.37
30 Wholesale and retail trade ⫺0.04 ⫺0.49
31 Eating and drinking places ⫺0.02 ⫺0.32
32 Passenger transportation ⫺0.14 ⫺1.95
33 Finance and insurance ⫺0.02 ⫺0.27
34 Real estate ⫺0.01 ⫺0.20
Code Name 2001–2002

35 Social services ⫺0.06 ⫺0.79
36 Health services, sports and social welfare ⫺0.04 ⫺0.60
37 Education, culture and arts, radio, film ⫺0.03 ⫺0.38
and television
38 Scientific research ⫺0.04 ⫺0.50
39 General technical service ⫺0.03 ⫺0.39
40 Public administration and other sectors ⫺0.04 ⫺0.59
Table 10.10 The world market price of crude oil and China’s average
tariff rate
Year TARIFF OILP OILPI

China’s Change Average Change Average Crude Change
Average in % Crude Price, in % Price Index, in %
Tariff Rate US$/Barrel 1995⫽100
1980 49.5 36.68 213.2
1981 49.5 0.00 35.27 ⫺3.84 205 ⫺3.85
1982 49.5 0.00 32.45 ⫺8.00 188.6 ⫺8.00
1983 46.65 ⫺2.85 29.64 ⫺8.66 172.3 ⫺8.64
1984 43.8 ⫺2.85 28.55 ⫺3.68 165.9 ⫺3.71
1985 40.95 ⫺2.85 27.37 ⫺4.13 159.1 ⫺4.10
1986 38.1 ⫺2.85 14.17 ⫺48.23 82.4 ⫺48.21
1987 39.5 1.40 18.2 28.44 105.8 28.40
1988 39.7 0.20 14.77 ⫺18.85 85.8 ⫺18.90
1989 39.9 0.20 17.91 21.26 104.1 21.33
1990 40.3 0.40 22.99 28.36 133.6 28.34
1991 42.9 2.60 19.37 ⫺15.75 112.6 ⫺15.72
1992 41 ⫺1.90 19.04 ⫺1.70 110.6 ⫺1.78
1993 39.9 ⫺1.10 16.79 ⫺11.82 97.6 ⫺11.75
1994 36.3 ⫺3.60 15.95 ⫺5.00 92.7 ⫺5.02
1995 35 ⫺1.30 17.2 7.84 100 7.87
1996 23.6 ⫺11.40 20.37 18.43 118.4 18.40
1997 17.6 ⫺6.00 19.27 ⫺5.40 112 ⫺5.41
1998 16.8 ⫺0.80 13.07 ⫺32.17 76 ⫺32.14
1999 16.55 ⫺0.25 17.98 37.57 104.5 37.50
2000 16.3 ⫺0.25 28.24 57.06 164.2 57.13
2001 15.3 ⫺1.00 24.28 ⫺14.02 141.1 ⫺14.07
2002 12 ⫺3.30 24.96 2.80 145.1 2.83
Source: The World Bank, IMF’s IFS.

between the oil price changes and the price changes in the final market, this
14 percent decline in oil prices could have contributed to 0.82 percent
decline of the consumer price index in 2002.
6 AN ALTERNATIVE MEASURE OF CORE

INFLATION
In this section, I report an alternative measure of the core inflation rate

based on a structural conditional error correction model. This model is
similar to the ones suggested by Marques et al. (2002) and Lee et al.
(2003). In this model, the core inflation is an unobserved I (1) variable,
which is cointegrated with the observed headline inflation. This model
can be represented as a state-space model and be estimated by a Kalman
filter.
The state-space model (Kalman filter) can be represented as follows:
Transition equation (or state equation)
␰t ⫹1 ⫽ T␰t ⫹ ␷t (10.14)
Observation equation (or signal equation)
yt ⫽ A⬘xt ⫹ H⬘␰t ⫹ ␻t (10.15)
Where ␰t is the state variable vector, yt is the observed variable vector, xt is

the exogenous variable, T is the transition matrix, and H is the observation
matrix.
Below is the description of the model I estimated. I define the following
variables:
␲t ⫽ CPI inflation
gt ⫽ Real GDP growth
ˇ
␲core
t ⫽ Core CPI inflation
The observed variable vector is
yt ⫽
冤冥
␲t
gt
The state variable vector is

冤冥
␲t
gt
␲core
t
␰t ⫽
⌬␲t
⌬gt
⌬␲core
t
Where ⌬ xt ⫽ xt ⫺ xt⫺1.
The observation equation is defined as
冤冥
␲t
gt
yt ⫽ 冤冥冤
␲t
gt
⫽
10 0 0 0 0
01 0 0 0 0 冥 ␲core
t
⌬␲t
⌬gt
⌬␲core
t
The transition equation is defined as follows
␲t ⫽ ␲t⫺1 ⫹ ⌬␲t
gt ⫽ gt⫺1 ⫹ ⌬gt
␲core
t ⫽ ␲core core
t⫺1 ⫹ ⌬ ␲t
⌬ ␲t ⫽ ␣1 (␲t⫺1 ⫺ ␲core
t⫺1 ) ⫹ ␤1 ⌬␲t⫺1 ⫹ ␤2 ⌬gt⫺1 ⫹ ␤3 ⌬␲t⫺1 ⫹ ␧1t
core
⌬gt ⫽ ␤4 ⌬␲t⫺1 ⫹ ␤5 ⌬gt⫺1 ⫹ ␤6 ⌬␲core

t⫺1 ⫹ ␧2t
⌬ ␲core
t ⫽ ␤7 ⌬␲core
t⫺1 ⫹ ␧3t
Or
冤冥冤冥冤冥冤冥
␲t 1 0 0 1 0 0 ␲t⫺1 0
gt 0 1 0 0 1 0 gt⫺1 0
␲core
t 0 0 1 0 0 1 ␲core
t⫺1 0
␰t ⫽ ⫽
⌬␲t ␣1 0 ⫺ ␣1 ␤1 ␤2 ␤3 ⌬␲t⫺1 ␧1t
⌬gt 0 0 0 ␤4 ␤5 ␤6 ⌬gt⫺1 ␧2t
⌬␲t core 0 0 0 0 0 ␤7 ⌬␲core
t⫺1 ␧2t
Where var(␧1t ) ⫽ ␴21, var(␧2t ) ⫽ ␴22, var(␧3t ) ⫽ ␴23.

Table 10.11 reports the estimation results; Table 10.12 reports the esti-
mated core CPI inflation; Figure 10.5 plots the estimated core CPI inflation
along with the overall CPI inflation. It can be seen from the chart that the
core CPI inflation rates were all positive over the estimation interval.
It should be noted that the estimated core inflation rates based on
this model are very sensitive to the initial conditions used in the estima-
tion. For example, the estimated results for sample period 1980–2002 are
significantly different from the estimated results for sample period
1978–2002. It is likely that there are multiple time series that are cointe-
grated with the headline inflation. As a result, the initial conditions will
determine which time series get picked. As the core inflation is defined as
Table 10.11 Estimation results by Kalman filter
Coefficient Std. Error z-statistic Prob.

␣1 0.099733 0.030546 3.264962 0.001100
␤1 0.536499 0.091571 5.858822 0.000000
␤2 ⫺0.008897 0.066462 ⫺0.133870 0.893500
␤3 0.222716 0.152332 1.462038 0.143700
␴1 1.859187 0.124553 14.926920 0.000000
␤4 ⫺0.265674 0.298732 ⫺0.889339 0.373800
␤5 ⫺0.324210 1.039805 ⫺0.311799 0.755200
␤6 3.333721 1.158122 2.878559 0.004000
␴2 0.055450 79.019400 0.000702 0.999400
␤7 ⫺0.269216 1.086708 ⫺0.247736 0.804300
␴3 1.958858 0.638306 3.068840 0.002100
Final State Root MSE z-statistic Prob.
␲t 0.139150 1.919737 0.072484 0.942200
gt 8.128168 6.530537 1.244640 0.213300
␲core
t 5.461912 3.123347 1.748737 0.080300
⌬␲t 0.999049 2.190233 0.456138 0.648300
⌬gt ⫺0.268281 7.669686 ⫺0.034979 0.972100
⌬␲core
t 0.000874 2.033565 0.000430 0.999700
Log likelihood ⫺528.1941 Akaike info criterion 11.72161
Parameters 11 Schwarz criterion 12.02313
Diffuse priors 6 Hannan-Quinn criterion 11.84331
Notes:
Method: Maximum likelihood (Marquardt).
Date: 04/25/04 Time: 11:04.
Sample: 1980:1 2002:4.
Included observations: 92.
Table 10.12 CPI inflation and estimated core inflation using Kalman filter,
1980–2002
OBS CPI Inflation Core CPI Inflation

1980:1 9.10 5.99
1980:2 7.60 5.87
1980:3 7.10 5.83
1980:4 6.60 4.85
1981:1 0.80 4.09
1981:2 0.80 4.09
1981:3 2.10 4.19
1981:4 3.60 5.53
1982:1 2.40 5.82
1982:2 1.90 5.78
1982:3 1.80 5.78
1982:4 1.70 3.01
1983:1 0.60 7.15
1983:2 1.00 5.54
1983:3 1.11 7.76
1983:4 2.10 6.92
1984:1 1.10 6.43
1984:2 1.70 7.09
1984:3 2.40 9.59
1984:4 1.77 10.06
1985:1 4.83 9.33
1985:2 9.37 8.36
1985:3 11.40 4.95
1985:4 11.73 4.77
1986:1 8.17 5.18
1986:2 4.43 5.51
1986:3 4.17 7.09
1986:4 5.93 7.17
1987:1 5.40 7.04
1987:2 7.40 7.19
1987:3 7.90 7.09
1987:4 8.20 6.67
1988:1 10.30 7.07
1988:2 14.20 7.67
1988:3 23.10 7.86
1988:4 27.30 8.07
1989:1 27.60 7.23
1989:2 24.70 5.69
1989:3 15.30 4.51
1989:4 7.50 2.98

1990:1 4.00 2.32
1990:2 2.30 3.43
1990:3 2.20 6.49
1990:4 3.70 6.46
1991:1 1.90 6.47
1991:2 3.10 6.00
1991:3 4.70 3.99
1991:4 4.60 5.56
1992:1 5.40 6.32
1992:2 5.50 6.45
1992:3 6.20 7.48
1992:4 8.30 11.66
1993:1 11.10 3.96
1993:2 13.90 4.36
1993:3 16.10 10.88
1993:4 17.20 10.60
1994:1 22.20 9.59
1994:2 21.90 9.37
1994:3 25.70 10.75
1994:4 26.90 4.29
1995:1 22.60 12.64
1995:2 20.00 12.08
1995:3 14.80 10.05
1995:4 11.10 11.85
1996:1 9.40 10.83
1996:2 9.10 7.74
1996:3 7.90 3.38
1996:4 7.00 5.75
1997:1 5.20 6.71
1997:2 2.90 6.90
1997:3 2.10 5.29
1997:4 1.00 5.93
1998:1 0.00 4.20
1998:2 ⫺0.90 5.69
1998:3 ⫺1.40 6.38
1998:4 ⫺1.10 5.66
1999:1 ⫺1.40 4.88
1999:2 ⫺2.10 5.11
1999:3 ⫺1.30 3.20
1999:4 ⫺0.80 4.43

2000:1 0.10 5.54
2000:2 0.10 5.67
2000:3 0.30 5.19
2000:4 0.60 5.74
2001:1 0.10 5.55
2001:2 1.10 5.28
2001:3 0.80 4.60
2001:4 ⫺0.10 4.80
2002:1 ⫺0.60 5.11
2002:2 ⫺1.10 5.50
2002:3 ⫺0.80 5.45
2002:4 ⫺0.60 5.47
an unobservable variable in this model, by nature it is very difficult to

find out which series is the real core inflation. Because of this limitation,
this measurement of core inflation will not be used for the rest of this
chapter.
7 CONCLUSION
In this chapter, I have shown that although the overall CPI inflation rate
was negative in the years 1998 and 1999, the core CPI inflation rate after
excluding the price of food was positive in these two years. The core CPI
inflation rate for 2002 was negative after excluding the food prices. This
indicates that the CPI decline in 2002 might have been caused by factors
other than food prices. It was highly likely that the decline of the consumer
price index in 2002 was due to the effects of China’s accession into the WTO
and the reduction of the tariff rate in 2002. According to the WTO rules
and the commitments China made, China lowered the tariff rates of 5300
items in 2002, accounting for 73 percent of the total tariff items. The overall
tariff level was cut from 15.3 percent to 12 percent.
I have also decomposed the inflation rate of the RPI, and have found that
a striking difference between the movement of the RPI and the CPI is that
the inflation rate of the RPI, excluding food and energy, has been consis-
tently negative since November 1997. A potential explanation (Wu, 2004)
is that the increased competition in the distribution sector squeezed the
gross margins of the distributors, hence adding to the downward pressure
on retail prices.
30.00
25.00
20.00
15.00
Percent
10.00
325
5.00
0.00
:1 :4 :3 :2 :1 :4 :3 :2 :1 :4 :3 :2 :1 :4 :3 :2 :1 :4 :3 :2 :1 :4 :3 :2 :1 :4 :3 :2 :1 :4 :3
80 80 81 82 83 83 84 85 86 86 87 88 89 89 90 91 92 92 93 94 95 95 96 97 98 98 99 00 01 01 02
19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 20 20 20 20
–5.00
Quarter
CPI Inflation Core CPI Inflation
Figure 10.5 CPI inflation and estimated core inflation by Kalman filter, 1980–2002
NOTE
* The views expressed in this chapter are those of the author, and do not necessarily reflect
those of the Office of the Comptroller of the Currency. This chapter is based on Chapter 2
of my doctoral dissertation at Stanford University. I am grateful to Professor Lawrence
Lau, Professor Ronald McKinnon and Professor Takeshi Amemiya for their invaluable
guidance.
REFERENCES
Bernanke, Ben (2002), ‘Deflation – Making Sure “It” Doesn’t Happen Here’,
Speech by Ben S. Bernanke, Member of the Board of Governors of the US
Federal Reserve System, before the National Economists Club, Washington, DC,
21 November.
Bjornland, H.C. (2000), ‘Identifying Domestic and Imported Core Inflation’, IMF
Working Paper.
De Long, Bradford J. (1999), ‘Should We Fear Deflation?’, Brookings Papers on
Economic Activity, 1, 225–41.
Eckstein, O. (1981), Core Inflation, Engelwood Cliffs, N.J: Prentice-Hall.
Erceg, Christopher, Jon Faust, Luca Guerrieri, Carter Hemphill, Linda Kole,
Jennifer Roush, John Rogers, Nathan Sheets and Jonathan Wright (2002),
‘Preventing Deflation: Lessons from Japan’s Experience in the 1990s’,
International Finance Discussion Paper, Board of Governors of the Federal
Reserve System.
Flemming, J. (1976), Inflation, Oxford University Press.
IMF (2003), ‘Deflation: Determinants, Risks, and Policy Options – Findings of an
Interdepartmental Task Force’, IMF working paper.
Krugman, Paul R. (1999a), ‘Can Deflation Be Prevented?’, Krugman website,
http://web.mit.edu/krugman/www/.
Krugman, Paul R. (1999b), ‘Deflationary Spirals’, Krugman website, http://web.
mit.edu/krugman/www/.
Krugman, Paul R. (2003), ‘Fear of a Quagmire’, New York Times, 24 May,
Section A, Page 15, Column 5.
Lau, Lawrence J. (2003), ‘The World Economy and Deflation’, working paper pre-
sented at the Fifth Meeting of the International Advisory Council of the China
Development Bank, Beijing, 7 November 2003, Stanford University.
Lee, Jong Wong, Sang Don Lee and Ki-Ho Kim (2003), ‘New Approach to
Estimation of the Core Inflation’, Journal of the Korean Economy, 4 (2), 315–43.
Leontief, Wassily (1986), ‘Input–output Economics’, 2nd Edition, Oxford
University Press.
Marques, C.R., P.D. Nerves and A.G. Da Silva (2002), ‘Why Should Central Banks
Avoid the Use of the Underlying Inflation Indicator’, Economics Letters, 75,
17–23.
Melvin, James R. (1979), ‘Short-run Price Effects of the Corporation Income Tax
and Implications for International Trade’, The American Economic Review,
69 (5), 765–74.
National Bureau of Statistics of China (1999), ‘Input–output Table of China’,
China Statistics Press, China, Beijing.
Okun, Arthur M. (1970), The Political Economy of Prosperity, Brookings

Institution, Washington DC.
Quah, Danny and Shaun P. Vahey (1995), ‘Measuring Core Inflation’, The
Economic Journal, 105 (432), 1130–44.
Wu, Deming (2004), ‘Was there Deflation in China between 1997 and 2002? An
Empirical Study of Price Movement in China’, PhD dissertation, Stanford
University.
Wynne, Mark A. (1999), ‘Core Inflation: A Review of Some Conceptual Issues’,
working paper, European Central Bank.
Index
Abiad, A.G. 27, 28, 74–5 impact on growth and productivity
accountability, IMF 125–6, 128 13–15, 213–15
accounting standards 163, 181, 184, 187 new Asian miracle 53–5
Adams, F. Gerard 13–15, 197–218 overview 8, 23–4
adverse selection problem 155 predictability of crisis 44–53
age and wage rates 229–34 questions raised by 2–8
Akamatsu, K. 198 secular aftermath 5–6
Alba, Pedro 184 societal impacts 55–65
Allen, Franklin 148 Thai wage rates during 15–17
Alper, C.E. 27, 31, 134, 136 and third-generation crisis models
Alvarez-Plata, Patricia 31 72–3
APEC (Asia-Pacific Economic vulnerabilities of affected countries
Cooperation) 149 32–44
Argentina see also economic development;
bankruptcy infrastructure 183 economic growth; Thailand,
crisis (2002) 31 wage rates
IMF rescue package 125, 128 Asian Policy Forum 187–9
poverty 57 Association of Southeast Asian
Arrow, Kenneth 155 Nations (ASEAN), see ASEAN
artificial serial correlations 27–8 Australia
Aryah, Gosah 222, 225 borrowings 177
ASEAN (Association of Southeast corporate borrowing 173
Asian Nations) 44 economic information quality
Asia 167
in 21st century 53–5 financial infrastructure 164
bond markets 148–9 judicial system 166
new developments 6–8 Austria, domestic consumption 290
Asia before and after financial crisis
(1997–98) balance of payments, Malaysia 144
currency crisis theories 24–32 Baldacci, Emanuele 56
new Asian miracle 53–5 bank deposits/loans 158–9
overview 8, 23–4 Bank for International Settlements 3
predictability of crisis 44–53 bank lending, reliance on 185–6
societal impacts 55–65 Bank Negara 137
vulnerabilities of affected countries Bank of Japan 246, 255
32–44 Bank of Thailand (BOT) 9, 69, 80, 81,
Asia-Pacific Economic Cooperation 149, 185–6
(APEC), see APEC bankruptcy infrastructure 183
Asian Bond Fund 126, 189 banks
Asian Development Bank 181, 201 and debt financing 172–3
Asian Financial Crisis (1997–98) implications of absence of bond
currency crisis theories 24–32 markets 13, 174–5
329
330 Index
Barro, Robert 160 review of studies 135–7

Basel Capital Accord 121 simulations with model 138–45
Basu, Sudip Ranjan 292 capital flight, Thailand 32
Behrman, Jere R. 15–17, 221–35 capital growth 205
Belgium, domestic consumption 290 capital inputs 208–11
benchmark yield curve 177–8, 180, Capital Management Crisis (1998) 124
186 capital market
Bergsten, Fred 240 China 7
Bernanke, Ben 297 reforms 4–5, 8
Betcherman, G. 222 capital mix 4
Bjornland, H.C. 298 capital mobility 123
Black, Bernard 172–3 Caprio, Gerard, Jr 172
Black, Fischer 169, 170 carry trade 73
bond markets cartel pricing 174
absence of 169–76 Central and Eastern Europe, current
development of 126 account deficits 130–31
failure of 167–8 Chatusripitak, Nathporn 5, 12–13,
financial sector overview/flow of 148–90
funds analysis 150–62 Chile, foreign borrowing 127
new flotations 122 China
overview 12–13, 148–50 in 21st century 53–5
role of financial capital growth 205
infrastructure/efficient financial domestic saving 206
markets 162–76 economic growth 121–2
role of government as issuer 176–80 exchange rates 126, 247–58
Thailand 5, 180–90 export growth 207–8
borrowers GDP growth 200–201
monitoring of 174 inputs 210–11
preferences 158 investment 203–4
Boskin Committee 261, 291 labor productivity 202–3
Bosworth, Barry P. 198, 209 new developments 6–7
Brady Plan 121 petroleum price increases 10
Brazil total factor productivity 212–13,
domestic consumption 290 217–18
exchange rates 123 wage adjustment under a fixed
Bretton Woods Institutions 11, 121, exchange rate 247–55
127–8, 243–6, 257 China, core inflation rate
Brimble, Peter 225, 232 alternative measure of core inflation
Brunetti, C. 31 319–24
business cycle 199, 209 definition of core inflation 297–8
description of consumer price index
Calvo, G. 72, 134 by category 298–301
Camdessus, Michel 121 estimation of core inflation rate
Canada, bond market 149, 189 301–9
capital account indicators 85, 86 impact of changes in petroleum
capital account variables 83 prices 309–19
capital controls overview 19–20, 296–7
main characteristics of China, cost of living
macroeconometric model 137–8 comments on linear expenditure
overview 11–12, 134–5 system 284–91
Index 331
lifestyle change 291–5 cost of living, China

linear expenditure system (LES) comments on linear expenditure
262–84 system 284–91
overview 17–19, 260–61 lifestyle change 291–5
Chuppe, Terry 148 linear expenditure system (LES)
Claessens, Stijn 156, 184 262–84
classification errors 28 overview 17–19, 260–61
clothing expenditure, China 263–78 Crafts, Nicholas 198, 209
CLSA Asia-Pacific Markets 187 credit culture, creation of 126
collective action clauses (CACs), debt credit risk, pricing 170–71
securities 125 creditor countries 241–2
Collins, Susan M. 198, 209 creditor rights 162–3, 181
Colombia, domestic consumption 290 creditworthiness information 157,
commodities expenditure, China 163
263–78 crises
consumer price index (CPI), China aftermath 5–6
by category 298–301 cures 3–5
comments on linear expenditure impending 10–11
system 284–91 lessons learned 3
core inflation rate estimation 301–9 nature and dynamics 2
lifestyle change 291–5 crony capitalism 13
linear expenditure system (LES) cross-border financial transactions 122,
262–84 161–2
overview 17–19, 260–61 cultural services expenditure, China
US 253 263–78
consumer price index (CPI), Japan 246 currency crisis theories
consumer prices early warning systems 26–8
China/US 249 Markov-switching approach 28–32
Japan/US 244 overview 24–6
consumption expenditure, China currency crises, vulnerability
262–91 indicators/analysis
contagion 25–6, 32, 57, 72, 105, 136 definition of currency
Copeland, Morris 150 crises/methodologies for
core inflation rate, China identifying 80–84
alternative measure of core inflation empirical results 85–108
319–24 overview 9–10, 69–71
definition of core inflation 297–8 review of selected studies 71–80
description of consumer price index summary 108–12
by category 298–301 currency devaluation 53
estimation of core inflation rate current account (CA)
301–9 Indonesia 39, 40
impact of changes in petroleum Korea 37, 38
prices 309–19 Malaysia 42
overview 19–20, 296–7 Philippines 45, 46
corporate borrowing 172–4 relative 48
corporate governance reforms 5, Thailand 33, 34, 83, 85–9, 91,
127–30 98–100, 104, 106, 107, 110,
corporate performance, Thailand 111
184 current account deficits 121–2
Corsetti, G. 197 current account imbalances 4
332 Index
Da Silva, A.G. 319 economic development 197–9

Dalla, Ismail 148 economic fundamentals, long-term
De Long, Bradford J. 297 deterioration 32, 33–44
de Mello, Luiz 56 economic growth
dealer markets 179 China 7, 18, 53–5
debt claims 167–8 East Asia 13–15
debt financing, dominance of banks India 7–8, 53–5
172–3, 174–5 economic growth during Asian
decomposition analysis 209 Financial Crisis
default 163, 168 overview 197–9
definitions growth record 199–208
core inflation 297–8 input/total factor productivity
currency crisis 81–2 approach 208–18
deflation 242 economic indicators
China 297 China/US 249
Japan 246–7 Japan/US 244
delivery versus payment (DVP) system, Edison, H.J. 134, 135, 136, 142
Thailand 186 education and wage rates, Thailand
Demirgüç-Kunt, Aslï 26, 172 229–34
Denison, Edward F. 208 education expenditure, China 263–78
Denmark, domestic consumption 290 education indicators 62–4
Deolalikar, Anil B. 15–17, 221–35 Eichengreen, Barry 2, 3, 5, 10–11, 26,
deposit rates, Thailand 85–9, 96, 72, 121–31, 135
98–100 elasticities approach, balance of trade
Detragiache, E. 26 241
Diamond, Douglas 172 emerging crises 121–31
differential wage adjustment 243–4, 247 empirical results, currency crisis
direct financial claims 156–7, 159–60, indicators
162 assessment of future vulnerability
with no secondary market 153–6 105–8
disclosure standards 181, 187 composite indicator and probability
Djankov, Simeon 156, 184 of crisis from signals approach
Dodd, R. 73 100–101
Doing Business in 2004 (World Bank) crisis indicators from signals
181 approach 85–100
dollar standard 240–58 Markov-switching models 31–2
domestic credit, Thailand 85–9, 95, profit estimates 101–5
98–100, 110, 111 empirical studies, currency crisis
domestic saving 205, 206 indicators 73–80
employment
early warning systems Malaysia 137–8
currency crises 9, 26–8 Thailand 15–16, 222, 227
Thailand 69–70 energy prices, China 301–2, 304–19
East Asia Engel curves 18, 263–91
economic growth 13–15 Engwatana, Pipat 74, 77, 81
income inequality 57–62 equity markets 167–8, 171
new developments 6–7 Erceg, Christopher 297
Eckstein, O. 298 Esquivel, Gerardo 73, 75–6
econometric predictive models 8, 24, ethnic tensions 56
28–32 Europe, bond markets 189
Index 333
European Monetary System (EMS) external debt, Thailand 85–9, 94,

72 98–100, 110, 111
European Union external sector variables 111
bond markets 149, 180, 189
reform agenda 129–30 Fallon, Peter R. 222
exchange rate differentials, China/US family-controlled industrial
252 conglomerates 155–6, 184
exchange rate dynamics 29–32 Fan, Joseph P.H. 184
exchange rate mechanism (ERM) crisis Faust, Jon 297
(1992–93) 25 Federal Reserve (FED), US 10, 121,
exchange rate policy, appropriate 130, 254
choice of 3–4 Fernholz, Ricardo 249
exchange rates financial claims 153–6
China 7, 247–58 financial crises
and international competitiveness impact on income inequality 57–62
243–55 impact on living standards 62–4
Japan 244–7 impact on poverty 56–7
Malaysia 145 impending 10–11
Thailand 74, 78–80, 81–2 overview 55–6
and trade balance 241–3 policy implications 64
see also flexible exchange rates; review of studies 135–7
floating exchange rates, pegged financial development, importance of
exchange rates; real effective bond market
exchange rate financial sector overview/flow of
exchange rates in international funds analysis 150–62
adjustment overview 12–13, 148–50
exchange rates and international role of financial infrastructure/
competitiveness 243–55 efficient financial markets
exchange rates and trade balance 162–76
241–3 role of government as issuer 176–80
flexibility, China 255–8 Thailand as example 180–90
overview 17, 240–41 financial infrastructure, role of
exclusion windows 27–8 162–76
Executives’ Meeting of East Asia- financial intermediaries 156–9
Pacific Central Banks (EMEAP) financial markets
149, 189 efficiency in 162–76
export growth rates saving/investment with lack of
Asia 205, 207–8 150–53
Indonesia 39, 40 Thailand 180–90
Korea 37, 38 financial sector
Malaysia 42, 43 flows with direct financial claims but
Philippines 45, 46 no secondary market 153–6
relative 50 government and international sector
Thailand 33, 34, 85–9, 91, 98–100, 159–62
104, 107, 111 Malaysia 138
export-led growth savings and investment without
China 8 financial markets/institutions
East Asia 197–8 150–53
exports during Asian Financial Crisis with financial intermediaries 156–9
13–15, 215–18 financial variables 83, 85, 86, 111
334 Index
first generation currency crisis models GDP (gross domestic product)

24–5, 71, 72 China 249, 260–61
fiscal balance Japan/US 244
Indonesia 39, 40 Malaysia 140, 141
Korea 37, 38 Thailand 83–4, 85–9, 91, 92, 95,
Malaysia 42, 43 98–100, 104, 106, 107, 110, 111
Philippines 45, 46 US 249
relative 49 GDP growth rates, Asia 200–201
Thailand 33, 34, 85–9, 92, 98–100, Geary–Khamis method 289–91
104, 106, 107, 111 Geithner, Timothy 130
fiscal discipline 123–4 gender and wage rates, Thailand
fiscal policy, post-crisis 56 229–34
Fischer, Stanley 125 general price changes 297
fixed exchange rate, China 247–55 Gerlach, S. 72
Flatters, Frank 232 Germany
Flemming, J. 297 borrowing 177
flexible exchange rates 123, 126–7, domestic consumption 290
255–8 Gilson, Ronald 172–3
floating exchange rates 254–5 Gini coefficient 57–62
Japan 246–7 global economic growth 121–2
flow of funds analysis Global Financial Stability Report, IMF
flows with direct financial claims but 124
no secondary market 153–6 Goldfajn, Ilan 73, 75–6
government and international sector Goldsmith, Raymond 148, 150, 156
159–62 Goldstein, Morris 134, 240
savings and investment without goods consumption, China 263–78
financial markets/institutions governance problems, IMF 127–8
150–53 government as issuer of bonds 176–80
with financial intermediaries 156–9 government budget balance, Malaysia
‘flying geese pattern’ 198 143
food expenditure, China 263–78 government consumption, Malaysia
food prices, China 299–300, 302, 141
304–9, 310–16 government debt, Malaysia 143
foreign borrowing 127 government fixed investment, Malaysia
foreign capital flows, China 255–6 141
foreign debt, Thailand 74, 184–5 government policies, unsustainability
foreign direct investment (FDI) 14–15, of 24–5
135, 216, 217–18, 255–6 government sector, impacts on flow of
foreign reserves, accumulation of funds 159–62
123 government securities, Thailand
forward markets/contracts 169–70 180–81
France, domestic consumption 290 Greenspan, Alan 130
Fund of Bond Funds, Asia 189 gross domestic product (GDP), see GDP
futures markets/contracts 169–70 growth accounting framework 198–9
Guerrieri, Luca 297
Gale, Douglas 148 Gultekin, B.N. 27, 31, 74–5, 134, 136
Gao, Huiqing 17–19, 260–95 Guttentag, Jack M. 179
GARCH(1,1) model 136
Gaussian cumulative distribution Hakansson, Nils H. 171
function 29 Hamilton, James B. 30
Index 335
Hart, Oliver 172 income inequality, impacts of financial

Hausman, Jerry A. 160 crises 57–62
Headcount Index, East Asia 58–62 India
headline inflation 298 in 21st century 53–5
health indicators 62–4 domestic consumption 290
hedging 124, 170, 175 new developments 7–8
hedonic price indexes 291 indicators, currency crisis
Hemphill, Carter 297 empirical results 85–108
herding behavior 26 empirical studies 73–80
Herrick, Bruce 197 identifying 80–84
Herring, Richard J. 5, 12–13, 148–90 overview 82–4
Herstatt Crisis (1974) 121 indirect financial claims 156–9
heterogeneity across countries 28 Indonesia
Hong Kong bankruptcy infrastructure 183, 189
bond market 148–9, 180, 189 bond market 149
borrowings 177 borrowings 177
capital growth 205 capital controls 136–7
corporate borrowing 173 capital growth 205
economic information quality 167 corporate borrowing 173
export growth 207 domestic saving 206
financial infrastructure 164 economic information quality 167
GDP growth 201 export growth 207–8
inputs 211 financial infrastructure 164
investment 203 GDP growth 200–201
judicial system 166 income inequality 59–62
labor productivity 203 inputs 209–11
total factor productivity 213, 217 investment 203–4
Horton, S. 222, 232 judicial system 166
household facilities/services labor productivity 202–3
expenditure, China 263–78 living standards 63
household savings, increases in macroeconomic indicators 36, 39–41
160–61 poverty 57
household sector, sources and uses of real wage rates 222
funds 151–9 societal impact of crises 23–4
Hu, Teh-wei 292 total factor productivity 212–13, 217
human development indicators 291–5 vulnerability 53
Hungary industrial production
current account deficit 131 during Asian Financial Crisis
domestic consumption 290 215–18
China/US 249
IBM 256 Japan/US 244
import growth rate inferior goods/services 263
Indonesia 39 inflation differentials
Korea 37 China/US 252, 254
Malaysia 42 Japan/US 248
Philippines 45 inflation rate
relative 51 Indonesia 39, 40
Thailand 33, 85–9, 98–100 Korea 37, 38
Inada, Yoshihisa 53 Malaysia 42, 43
Inchauste, G. 56 Philippines 45, 46
336 Index
relative 52 Special Data Dissemination

Thailand 33, 34, 85–9, 95, 98–100, Standard (SDDS) 112, 124–5
110, 111 Special Drawing Rights 129
see also core inflation rate, China surveillance role 3, 124–6
information flows 163, 171 Thai package 9–10, 69–70, 78–80,
input–output price model, petroleum 136
price changes 311–16 international reserves, Thailand 9–10,
inputs 198–9, 208–11, 215–16 69, 78–80, 81–2, 85–9, 94, 98–100,
Institute for International Economics 110, 111
(IIE), US 240 international sector, impacts on flow of
Institute of Directors, Thailand 187 funds 159–62
Institute of International Finance 122 investment
Intaravitak, Chedtha 225, 232 bias in decisions 172
Inter-American Development Bank during Asian Financial Crisis 13–15,
127 203–4, 215–18
interest rates without financial markets/
government bonds 178 institutions 150–53
Malaysia 142, 145 investors
term structure of 169–71 implications of absence of bond
international adjustments markets 172–4
exchange rates and international portfolio allocation preferences
competitiveness 243–55 178–9
exchange rates and trade balance Iran, domestic consumption 290
241–3 Ireland, domestic consumption 290
foreign exchange flexibility, China Islam, R. 222
255–8 Italy, domestic consumption 290
overview 240–41 Ittisupornrat, Anotai 74, 77, 81
international competitiveness, effect of
exchange rate 243–55 Jamaica, domestic consumption 290
international dollar standard 240–58 Japan
international financial markets, bond market 149, 173, 189
structure of 127 borrowings 177
international financial system corporate borrowing 173
agenda for reforms 126–30 dollar exchange rate in historical
reforms 122–6 perspective 244–7
international issuance of bonds 173–4 domestic consumption 290
International Monetary Fund (IMF) economic information quality
Articles of Agreement 256, 257 167
Capital Markets Division 124 financial infrastructure 164
Financial Sector Assessment international adjustment 17
Program 124 judicial system 166
Independent Evaluation Office 129 new developments 7
International Monetary and Jeanneau, S. 73
Financial Committee (IMFC) judicial systems 163, 166, 168
129 Jun, Lwang 148
Korean package 136
reform agenda 122–30 Kahler, M. 134
Reviews of the Observance of Kakwani, N. 225
Standards and Codes (ROSCs) Kalman filter 298, 319–24, 325
124 Kaminsky, G.L. 9, 26, 27, 70–71, 73–4,
Index 337
75–6, 81, 82–3, 84, 99, 101, 104, Leontief, Wassily 311
105, 134 letter stocks 179
Kane, Edward J. 163, 166, 167 leverage 124, 172, 175, 184
Kaplan, E. 136 Levine, Ross 148
Kenya, domestic consumption 290 liberalization
Khatdhate, Deena 148 capital account 11, 135–7
Kim, Ki-Ho 298, 319 capital markets 4–5
Kindleberger, Charles P. 197 China 256–7
Kittiprapas, Sauwalak 225, 232, 233 Liberty Bonds, US 176–7
Klein, Lawrence R. 1–20, 23–65, lifestyle change 19, 261, 263, 291–5
134–46, 260–95 Lindbeck, Assar 244
Klein, M. 146 linear expenditure system (LES)
Knowles, James C. 232 262–84
Kojima, K. 198 comments on 284–91
Kole, Linda 297 overview 18–19, 260–61
Kondury, Kali 148 liquid secondary markets 178–80
Kravis, Irving 285–6, 287, 288–9, 291 liquidity shocks 175
Kruger, M. 73, 75–6 literature, real wage declines 224–5
Krugman, Paul 25, 53, 72, 73, 134, living standards, impacts of financial
198, 297 crises 62–4
Kumar, M. 26 Lizondo, S. 9, 71, 73–6, 81, 82–3, 84
loan contracts 159
La Porta, Rafael 162, 163, 166 logit model 26–8, 73, 75–6, 80–81, 84
Labor Force Surveys (LFSs) 224–5 London Interbank Offered Rate
labor inputs 208–11 (LIBOR) 85–9, 96, 98–100
labor market conditions 55 Lopez-de-Silanes, Florencio 162, 163,
labor market, Thailand 221 166
labor productivity Lucas, Robert E.B. 222
Asia 201, 202, 203 Luxembourg, domestic consumption
China/US 249 290
Japan/US 244
Lang, Larry H.P. 156, 184 M1 85–9, 97, 98–100
Lardy, Nicholas 240 M2 9, 29, 73, 85–9, 94, 97, 98–100, 104,
Larrain, Felipe 73, 75–6 106, 107, 110, 111, 136–7
Latin America macroeconometric model, Malaysia
current accounts 122 main characteristics 137–8
exchange rates 123, 127 overview 11–12, 134–5
fiscal discipline 123–4 simulations 138–45
income inequality/poverty 57 macroeconomic indicators 24
living standards 62–3, 64 Indonesia 36, 39–41
Lau, Lawrence J. 53, 199, 297 Korea 36, 37–8, 44
leading-indicator approach 26–8 Malaysia 41–4
Thailand 69–70, 80–84, 108–9, 111 overview 12
Leblang, D. 135 Philippines 44, 45–6
Lee, Jong Wong 298, 319 Thailand 33–6
Lee, Sang Don 298, 319 macroeconomic performance 135–7
legal frameworks, bond markets 162–3 Mahmood, Moazam 222, 225
legal infrastructure, Thailand 181 Makino, J. 53
lending rate, Thailand 85–9, 96, 98–100 Malawi, domestic consumption
Lenovo 256 290
338 Index
Malaysia medicine/medical services expenditure,

bankruptcy infrastructure 183 China 263–78
bond market 148, 149, 189 Melvin, James R. 311
borrowings 177 Merton, Robert 162
capital growth 205 Mexican Crises (1982/1994–95) 53, 73,
capital market reforms 5 121
corporate borrowing 173 Mexico
domestic consumption 290 bankruptcy infrastructure 183
domestic saving 206 domestic consumption 290
economic information quality 167 living standards 62–3
export growth 207–8 loans 127
financial infrastructure 164 poverty 56, 57
GDP growth 200–201 Micu, M. 73
income inequality 57–8, 60–62 minimum lending rate (MLR) 85–9,
inputs 209–11 96, 98–100
investment 203–4 monetary policy, post-crisis 56
judicial system 166 money markets, net flows 122
labor productivity 202–3 money wages
macroeconomic indicators 41–4 China/US 249
total factor productivity 212–13, 217 Japan/US 244
Malaysia, capital controls monitoring of borrowers 174
main characteristics of Moore, John 172
macroeconometric model 137–8 Moorthy, U. 26
overview 11–12, 134–5 moral hazard 155, 256
real wages 222 mortality rates 62–3
review of studies on capital controls Mussa, Michael 240
135–7
simulations with model 138–45 Nagar, A.L. 292
societal impact of crises 23–4 narrow money
management accountability, IMF 128 China/US 249
manufacturing wage growth Japan/US 244
China 247, 249 National Economic and Social
China/US 250 Development Board, Thailand
Japan/US 245–6 112
Marcos regime, Philippines 44 negative shocks 91–5, 98–100
Mariano, Roberto S. 4–5, 11–12, 27, Nerves, P.D. 319
28, 31, 74–5, 134–46 Netherlands, domestic consumption
market sentiment 26 290
Markov-switching models 28–31, 65, new developments
136 China 6–8
empirical results 31–2 East Asia 6
Marques, C.R. 319 India 7–8
Martinez-Peria, M.S. 28 Japan 7
Masson, P. 26 new issues, government bonds 178
Mazumdar, D. 222, 232 New York Stock Exchange (NYSE)
McKinnon, Ronald 4, 17, 148, 240–58 179
mean absolute error (MAE) 138 Nixon, Richard 242, 246
mean absolute percentage error nominal GDP
(MAPE) 138 China/US 249
mean consumption, East Asia 58–62 Japan/US 244
Index 339
non-inferior goods/services 263 capital growth 205

non-parametric approach, analysis of corporate borrowing 173
leading indicators 84 domestic consumption 290
non-performing loans (NPLs) 3, 71, domestic saving 206
137, 175 economic information quality 167
export growth 207–8
Obstfeld, M. 25, 72 financial infrastructure 164
OECD (Organization for Economic GDP growth 200–201
Co-operation and Development) income inequality 59–62
Code of Good Corporate inputs 210–11
Governance 187 investment 203–4
use of leading indicators 70 judicial system 166
Ohno, Kenichi 242, 255 labor productivity 202–3
Okun, Arthur M. 297 macroeconomic indicators 44, 45–6
OPEC (Organization of the Petroleum societal impact of crises 23–4
Exporting Countries) 296 total factor productivity 212–13, 217
option contracts 170 Poland, domestic consumption 290
Organization for Economic Co- policy implications, financial crises 64
operation and Development Pongsapich, Amara 225, 232
(OECD), see OECD Poonpatpibul, Chaipat 74, 77, 81
Organization of the Petroleum positive shocks 89, 95–7
Exporting Countries (OPEC), see post-crisis patterns of economic
OPEC fundamentals 33–44
Osakwe, P.N. 73, 75–6 Poterba, James M. 160
overinvestment, Thailand 184 poverty, impact of financial crises 56–7
Özmucur, Süleyman 4–5, 11–12, 27, 28, Pratt, Shannon P. 179
31, 134–46, 292, 293 pre-crisis patterns of economic
fundamentals 33–44
Page, J. 73, 75–6 predictability, Asian Financial Crisis 8,
Pakistan, domestic consumption 290 9, 44–53
Pan-Asia Bond Index Fund 189 predictive models, currency crises
parametric approach, analysis of 28–32
leading indicators 84 primary market, Thailand 185–6
pegged exchange rates 6, 25, 53; see principal–agent problem 155
also dollar standard private capital flows, Malaysia 142
People’s Bank of China (PBC) 257–8 private consumption, Malaysia 141
per capita expenditures, China 300–301 private fixed investment, Malaysia
Pérez, J. 26 141
Pernia, Ernesto M. 232 probit model 26–8, 73, 74, 75–7, 80–81,
Perraudin, W. 26 84, 101–8, 109, 111
Pesenti, P. 197 production function approach 199
petroleum price changes, impact in production sector, Malaysia 137–8
China productivity
estimation results 316–19 China 240, 243
input–output price model 311–16 during Asian Financial Crisis 13–15
overview 10, 298, 309–11 see also total factor productivity
Philippines (TFP)
bond market 149, 189 public sector, Malaysia 138
borrowings 177 purchasing power parity (PPP) 254,
capital controls 136, 137 285–6, 289
340 Index
pure loans 158–9 corporate governance 5

pure security 159 exchange rate policy 3–4
international financial system 3, 11,
Quah, Danny 298 122–6
quasi-financial systems 155–6 Reinhart, C.M. 9, 26, 27, 70–72, 73–4,
75–6, 81, 82–3, 84, 99, 105, 134,
Racelis, Mary 232 136, 142
Radelet, Steven 197 relative price change 55, 297
Rajan, Raghuram 176–7 representativeness, IMG 128–9
Rao, D.C. 148 Republic of Korea, see South Korea
real assets 150–56 residence expenditure, China 263–78
real effective exchange rate (REER) 29, retail price index (RPI), China
31, 136–7 309–16
Indonesia 39 review of studies
Korea 37 capital controls 135–7
Malaysia 42 currency crises 71–80
Philippines 45 Ricci, L. 134, 135
Thailand 33 risk
real exchange rate, Thailand 85–9, management of 3, 174–5
93, 98–100, 104, 106, 107, 110, settlement obligations 179–80
111 Rodrik, D. 134, 136
real exports, Malaysia 139, 140 Rogers, John 297
real GDP growth Romania
China/US 249 current account deficit 131
Indonesia 39, 40 domestic consumption 290
Japan/US 244 root mean square error (RMSE) 138
Korea 37, 38 Rose, A. 26, 72
Malaysia 42, 43, 139, 140 Roubini, N. 197
Philippines 45, 46 Roush, Jennifer 297
relative 47 rural household expenditure, China
Thailand 33, 34, 85–9, 92, 98–100, 263, 276–8, 279, 280–83, 285, 286,
104, 106, 111 301
real government consumption Russia, poverty 57
expenditures, Malaysia 139
real imports, Malaysia 140 Sachs, Jeffrey D. 197
real indicators, Malaysia 141 Salvatore, D. 134
real private consumption expenditures, Samuelson, P.A. 284, 291
Malaysia 139 savers
real sector variables 83, 85, 86, 111 implications of absence of bond
real wage rates, Thailand markets 171–2
overview 5–6, 15–17, 221–3 preferences 158
re-examination of real wage rate savings without financial
changes 226–34 markets/institutions 150–53
summary of previous literature Scandinavian model of wage
224–5 adjustment 243–4, 246, 254
recreation services expenditure, China Scholes, Myron 170
263–78 Schrooten, Mechthild 31
reforms Scotti, C. 31
capital markets 4–5 second generation currency crisis
China/India 54–5 models 25, 71, 72
Index 341
secondary markets Solow, Robert M. 208

direct financial claims with lack of Sonakul, M.R. 149
153–6 South Asia, new developments 7–8
nurturing 162, 178–80 South Korea
Thailand 181, 185–6 bankruptcy infrastructure 183
secular aftermath, Asian Financial bond market 149, 189
Crisis 5–6 borrowings 177
sensitivity analysis 209 capital controls 136, 137
services expenditure, China 263–78 capital growth 205
Shabbir, Tayyeb 1–20, 23–65, 74–5, corporate borrowing 173
134, 136, 197–218 domestic consumption 290
Shaw, Edward 148 domestic saving 206
Sheets, Nathan 297 economic information quality 167
Sheng, Andrew 149 export growth 207–8
Shleifer, Andrei 162, 163, 166 financial infrastructure 164
Siamwalla, Ammar 225 GDP growth 200–201
signals approach income inequality 58, 60, 61
composite indicator and probability inputs 210–11
of crisis 100–101, 111 investment 203–4
crisis indicators from 85–100, 105–8 judicial system 166
overview 9, 26–8, 73–4, 75–7, 80–81, labor productivity 202–3
84 macroeconomic indicators 36, 37–8,
significance testing 28 44
simultaneous equation macroeconomic real wage rates 222
model 11–12 societal impact of crises 23
Singapore total factor productivity 212–13, 217
bond market 149, 189 Spain, domestic consumption 290
borrowings 177 speculative attacks
capital growth 205 early warning systems 26–8
corporate borrowing 173 impact of 81
domestic saving 206 Markov-switching approach 28–32
economic information quality 167 vulnerability to 73
export growth 207 spillover effects 26, 32
financial infrastructure 164 Sri Lanka, domestic consumption 290
GDP growth 201 stabilization role, China 260, 297
inputs 211 staff surveillance, IMF 129–30
investment 203 stakeholder representation, IMF 128
judicial system 166 State Administration for Foreign
labor productivity 203 Exchange (SAFE), China 256–7
total factor productivity 213, 217, 218 State Information Center, China 287
Singh, A. 135 statistical analysis, determinants of
Slok, T. 134, 135 TFP 211–18
slumps, exchange-rate induced 242 step-wise linear methods 287–8
Slutsky–Hicks symmetry conditions Stiglitz, J.E. 135, 168
262 Stock Exchange of Thailand (SET)
Smets, F. 72 85–9, 90, 98–100, 104, 106, 107,
Smith, Clifford, Jr 169–70 110, 111, 184, 187
Smithson, Charles 169–70 Stone, Richard 291
societal impact of crises 2, 5–6, 23–4, structural conditional error correction
55–65 model 319
342 Index
structural problems, Thailand 181–4 bond market 5, 13, 149, 180–90

structural vector auto-regression borrowings 177
(SVAR) model 298 capital controls 136, 137, 146
Stulz, René 148 capital growth 205
subsistence income 279–80, 287 capital flight 32
Summers, L. 135 corporate borrowing 173
Summers, Robert 285–6, 287, 288–9, 291 domestic consumption 290
surveillance of financial systems 124–6 domestic saving 206
Sussangkarn, Chalongphob 232 economic information quality 167
swap contracts 169–70 export growth 207–8
swap operations 81 financial infrastructure 164
swap rates 249 GDP growth 200–201
Syria, domestic consumption 290 income inequality 57–8, 60–62
Szirmai, A. 199 inputs 209–11
international reserves 7
Taipei investment 203–4
capital growth 205 judicial system 166
export growth 207 labor productivity 202–3
GDP growth 201 macroeconomic indicators 33–6
inputs 210–11 poverty 57
investment 203 total factor productivity 212–13,
labor productivity 203 217
total factor productivity 212–13, vulnerability 53
217, 218 Thailand Securities Depository
Taiwan Corporation (TSD) 186
borrowings 177 Thailand, vulnerability to currency
corporate borrowing 173 crisis
economic information quality 167 definition of currency crisis/
exchange rates 127 methodologies for identifying
export growth 208 80–84
financial infrastructure 164 empirical results 85–108
judicial system 166 overview 9–10, 69–71
Tan, A.H.H. 28, 31, 74–5 selected studies on currency crises
Tan, Kong-Yam 249 71–80
Tao, Liping 17–19, 260–95 societal impact of crises 23–4
tariffs, China 302 summary 108–12
tax revenues, Malaysia 142, 143 Thailand, wage rates
taxation, China 256 overview 5–6, 15–17, 221–3
technology change 211, 213, 216 re-examination of real wage rate
technology-intensive production 198 changes 226–34
Tequila Crisis (1994–95) 105 real wage decline literature 224–5
Thai Bond Dealing Center (BDC) 181, theoretical explanation of currency
186 crisis, Thailand 71–3
Thai Central Bank 189 third generation currency crisis models
Thai Rating and Information Services 25–6, 72–3
Company (TRIS) 184 time-series-based estimation methods
Thai Securities and Exchange 298
Commission 186 Timmer, M.R. 199
Thailand Tinakorn, Pranee 2, 9–10, 15–17,
bankruptcy infrastructure 183 69–112, 221–35
Index 343
Tobin Tax 134 international adjustment 17

total factor productivity (TFP) 208–11 judicial system 166
overview 13–15, 198–9 price indicators 291–3
statistical analysis of determinants treasury bonds 123
211–18 US and international adjustment
trade balance, effect of exchange rate exchange flexibility, China 255–8
241–3 exchange rates and international
trade balance/growth, Thailand 85–9, competitiveness 243–55
90, 98–100, 110, 111 exchange rates and trade balance
trade surplus 4 241–3
trading volume, Thai bond markets 188 overview 240–41
transactions costs 157 US Treasury 128
transparency 3
IMF 11, 125 Vahey, Shaun P. 298
transport expenditure, China 263–78 Valdés, Rodrigo O. 73, 75–6
Turkey vector error correction model (VECM)
bankruptcy infrastructure 183 298
current account deficit 131 venture capital sector 173
fiscal discipline 124 Vishny, Robert 162, 163, 166
volatile prices 296
UK vulnerabilities
borrowings 177 competing hypotheses as
corporate borrowing 173 explanations of crisis 32–3
cost of living 261 early warning systems 26–32
domestic consumption 290 macroeconomic indicators 24
economic information quality 167 pre-crisis/post-crisis patterns of
financial infrastructure 164 fundamentals 33–44
judicial system 166 vulnerability, Thailand
unemployment 56 definition of currency crisis/
United Nations Development methodologies for identifying
Programme (UNDP) 291 80–84
University of California School of empirical results 85–108
Public Health 292 overview 9–10, 69–71
University of Pennsylvania 285 selected studies on currency crises
urban household expenditure, China 71–80
263, 264–75, 279, 280–83, 284, societal impact of crises 23–4
286, 287, 288, 289, 300, 301 summary 108–12
Uruguay, domestic consumption 290
US wage adjustment, China 247–55
bond market 149, 176–7, 180, 189 wage bargaining, China 254–5
borrowings 177 wage changes in international
corporate borrowing 173 adjustment 17
current account deficit 121–2, wage differentials, Japan/US 248
130–31 wage rates in international adjustment
dollar exchange rate 244–7 exchange flexibility, China 255–8
domestic consumption 290 exchange rates and international
economic information quality 167 competitiveness 243–55
financial infrastructure 164 exchange rates and trade balance
inflation rate 261 241–3
interest rates 10 overview 17, 240–41
344 Index
wage rates, China 251 declines (2000) 15–16, 221–3,

wage rates, Thailand 224, 225, 226–32
overview 5–6, 15–17, 221–3 study of Asian bond markets 148–9
re-examination of real wage rate World Development Report (2000–01),
changes 226–34 World Bank 57, 64
summary of previous literature Wright, Jonathan 297
224–5 WTO (World Trade Organization) 256,
wage recipients, Thailand 227–32 324
Wakeman, Lee 169–70 Wu, Deming 19–20, 296–325
Weinstein, D.E. 173 Wynne, Mark A. 297
Weiss, Andrew 168 Wyplosz, C. 72
wholesale price index (WPI), Japan/US
246 Yafeh, Y. 173
wholesale prices Yam, Joseph 149
China/US 249 Young, A. 198
Japan/US 244 Yugoslavia, domestic consumption 290
working hours, Thailand 227–8
World Bank Zambia, domestic consumption 290
lending 127 Zhang, Xiaobo 249
studies of Thai real wage rate Zingales, Luigi 176–7

Recent Financial Crises

Uploaded by

Copyright:

Available Formats

Recent Financial Crises

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Recent Financial Crises

Uploaded by

Copyright:

Available Formats

Introduction

Lawrence R. Klein and Tayyeb Shabbir

1 ABOUT THIS BOOK

This volume has been assembled to focus on specific analytical as well as

2 QUESTIONS OF IMPORTANCE RAISED BY THE

The last decade or so has been marked by a significant number of major

2.1 Nature and Dynamics of the Crises

2.2 Lessons Learned

2.3 ‘Cures’ for the Crises

2.3.1 Appropriate exchange rate policy

choice of an exchange rate system is not an open and shut case, as it is

2.3.2 Capital market reforms

study of Malaysia, which assesses the effects of capital controls, to help

2.3.3 Private business governance

2.4 Secular Aftermath of the Asian Financial Crisis 1997–98

in the affected countries pre- versus post-Asian Crisis. (Chapter 7 presents

2.5 Focus on New Developments in the Asian Region

We feel that in any discussion of the recent financial crises, in addition to

2.5.1 East Asian directly affected countries

2.5.2 China and Japan

dominant part of future economic prosperity of the region – a ‘New Asian

2.5.3 India (South Asia)

3 SALIENT POINTS OF THE CONTRIBUTIONS TO

Chapter 1: ‘Asia Before and After the Financial Crisis of 1997–98:

Chapter 2: ‘Indicators and Analysis of Vulnerability to Currency Crisis:

This chapter by Professor Tinakorn is a case study of Thailand, a country

the IMF, thus possibly ameliorating some of the extreme effects of an

Chapter 3: ‘The Next Financial Crisis’ by Barry Eichengreen

In this chapter, Barry Eichengreen contends that, as of spring 2005, the

1. Rising US interest rates as a result of the FED’s (Federal Reserve’s)

of the exchange rate mechanisms, greater fiscal ‘responsibility’ as well as

Chapter 4: ‘Capital Controls, Financial Crises and Cures: Simulations

Liberalization of capital account is generally favored as a desirable policy

variables. Such detail allows the authors to supplement a traditional macro-

absence of a well-functioning bond market, hedging in the derivatives

Chapter 6: ‘Investment, Growth and Productivity during the East Asian

did not show up with statistical significance. Unexplained TFP growth is

In their chapter, Behrman, Deolalikar and Tinakorn (BDT) explore what

some other groups typically characterized as more vulnerable fared rela-

Chapter 8: ‘Exchange Rate or Wage Changes in International Adjustment?

Chapter 9: ‘Adjustment to China’s CPI-based Inflation Rate to Account for

Chinese economic growth since 1978 has certainly been spectacular in

This equation is expressed in current value prices (nominal). Goods and

populations on the basis of ‘true’ cost of living, compared with standard

Chapter 10: ‘Estimating China’s Core Inflation Rate’ by Deming Wu

The premise of this chapter by Deming Wu is to ascertain the ‘core’

1. Kaminsky et al. (1998).

Kaminsky, G.L., S. Lizondo and C.M. Reinhart (1998), ‘Leading Indicators of

Analysis of Currency Crises

1. Review the different theories that motivate the specification of econo-

The chapter is arranged in the following manner. Section 2 presents an

2 CURRENCY CRISIS THEORIES1

The theoretical literature on currency crises can be categorized into three

the likelihood of a crisis occurring in other countries. Masson (1998)

2.1 Early Warning Systems

Recent efforts such as those of Demirgüç-Kunt and Detragiache (1998);

● percentage changes in nominal exchange rates;

Other possible determinants of relevance include current account balance,

point in time in the sample – based on whether each variable exceeds a

Crisis occurs within No crisis occurs within