c4pdf PDF

4
chapter
Financial Stress and Economic Downturns
This chapter examines why some episodes of financial vation and regulatory changes. Financial systems
stress lead to economic downturns and others have are inherently procyclical, with growth in credit,
only a limited impact on the overall economy. The leverage, and asset prices often reinforcing the
analysis indicates that episodes of financial turmoil underlying economic dynamic—and in some
characterized by banking sector distress are more likely cases leading to a buildup of financial imbal-
to be associated with severe and protracted downturns ances followed by a sharp correction (see Borio,
than episodes of stress centered mainly in securities or 2007; Goodhart, 1996; and Minsky, 1992).
foreign exchange markets. Countries with more arm’s- The impact of financial cycles on the real
length financial systems seem particularly vulnerable economy remains subject to debate in both
to sharp contractions in economic activity, because of academic and policy circles. One strand of
the greater procyclicality of leverage in their banking research emphasizes the role of the financial
systems. This chapter draws implications for economic accelerator in amplifying the effects of financial
prospects in the United States and the euro area and cycles on the real economy through its effect
considers how policy responses could help limit the on the value of collateral and thereby on the
output consequences of the current and future finan- willingness of the financial system to provide
cial crises. credit to the economy (Bernanke and Gertler,
T
1995; Bernanke, Gertler, and Gilchrist, 1999;
he financial turmoil that began in the and Kiyotaki and Moore, 1997). In this view,
summer of 2007 has mutated into a shocks that affect the creditworthiness of bor-
full-blown crisis, encompassing broader rowers tend to accentuate swings in output.
securities markets and the banking Another branch of inquiry focuses on lenders’
systems of several advanced economies. How will balance sheets and the relationship between
macroeconomic activity be affected, and what bank capital and aggregate credit, the so-called
can policymakers do both to reduce the eco- bank capital channel (Bernanke, Lown, and
nomic consequences of this crisis and to forestall Friedman, 1991; Kashyap and Stein, 1995; Peek
such crises in the future? Although past episodes and Rosengren, 1995; and Altunbas, Gamba-
of stress in banking, securities, and/or foreign corta, and Marqués, 2007). When bank capital
exchange markets have only sometimes been is eroded, banks become more reluctant to
associated with economic downturns (Figure 4.1, lend and may be forced to deleverage, lead-
top panel), these downturns have tended to be ing to sharper economic downturns. Another
more severe (Figure 4.1, bottom panel). area of analysis is the extent to which the role
An important concept in assessing the impact of the financial accelerator in the economy
of financial stress on economic activity is the varies with the type of financial system (World
role of financial cycles, which have been a con- Economic Outlook, September 2006; and Rajan
stant feature of the economic landscape despite and Zingales, 2003). The general trend toward
the evolution of financial systems through inno- greater reliance on arm’s-length financing and
less reliance on relationship-based lending may
have left economies better able to absorb finan-
The main authors of this chapter are Subir Lall, cial stress, as both households and firms can
Roberto Cardarelli, and Selim Elekdag, with support from now substitute away from banks to markets (and
Angela Espiritu and Gavin Asdorian. Hyun Song Shin
provided consultancy support. Jörg Decressin and Tim thus benefit from the so-called twin engines of
Lane were chapter supervisors. the financial system).
129
Chapter 4 Financial Stress and Economic Downturns
Against this background, this chapter

addresses the following questions: Why are some
periods of financial stress associated with slow-
downs, or even recessions, while others appear
to have little impact on the real economy? What
role is played by the size or location of finan-
Figure 4.1. Financial Stress and Output Loss cial imbalances or the state of households’ and
firms’ balance sheets? Has financial innovation
Only about half the episodes of financial stress identified in Chapter 4 for advanced reduced the role of banks in propagating shocks
economies over the past three decades were followed by economic downturns. from the financial system to the real economy?
Slowdowns and recessions preceded by financial stress tend to be longer and more
severe. To answer these questions, this chapter ana-
lyzes episodes of financial stress and economic
Number of Financial Stress Episodes 120 cycles in 17 advanced economies during the
1980s
1990s 100
past three decades. It draws lessons from these
2000s
episodes by differentiating among them on the
80
basis of the conditions in place at the start of
60 the financial stress episode, including the state
of households’ and firms’ balance sheets and
40
the dynamics of credit and asset prices in the
20 run-up to the stress episode; the type of finan-
0 cial stress involved (that is, whether related to
All Followed by Followed by Other
slowdowns recessions
banks, securities markets, or foreign exchange
markets); and the policy responses. Taken
Cumulative Output Loss during Slowdowns and Recessions1 together, these factors provide a comprehen-
(median; percent of GDP) sive window on the channels and mechanisms
0
through which financial stress affects economic
-1 cycles. While establishing the causality between
financial stress and economic downturns poses
-2
an inherently difficult challenge, the analysis in
-3
the chapter makes an attempt to address this
issue by explicitly accounting for the effects of
-4 the most common types of shocks studied in the
macroeconomic literature.
-5
Preceded Not Preceded Not The key findings of this chapter are the
by financial preceded by financial preceded
stress by financial stress by financial
following:
stress stress
Slowdowns Recessions
The countries included in this study are Australia, Aus-
Sources: Haver Analytics; OECD, Analytic Database; OECD, Economic Outlook (2008); tria, Belgium, Canada, Denmark, Finland, France, Ger-
and IMF staff calculations.
1Measured as the cumulative output loss when output was below the Hodrick-Prescott trend for many, Italy, Japan, Netherlands, Norway, Spain, Sweden,
slowdowns and cumulative output loss until recovery for recessions. Switzerland, United Kingdom, and United States.
This broader approach complements recent research
on the empirical relationship between asset prices—such

as equity and house prices and bond spreads—and the
dynamics of output during the course of the business
cycle (see Cihák and Koeva Brooks, forthcoming; and
Claessens, Kose, and Terrones, 2008) or between bank
capital, lending, and output (see Bayoumi and Melander,
2008; and Kashyap, Rajan, and Stein, 2008, in the context
of the United States).
130
Identifying Episodes of Financial Stress
• Episodes of financial turmoil characterized The rest of the chapter is structured as fol-
by banking distress are more often associated lows. The next section elaborates the concept of
with severe and protracted downturns than financial stress that is employed in this chapter,
episodes of stress centered mainly in securities and uses this concept to identify episodes of
or foreign exchange markets. financial turmoil during the past three decades.
• The likelihood that financial stress will be The chapter then analyzes the behavior of
followed by a downturn appears to be associ- economic cycles following these financial stress
ated with the extent to which house prices episodes. The section that follows discusses the
and aggregate credit rise in the period before factors that differentiate episodes that were
the financial stress. Moreover, greater reli- associated with economic downturns from those
ance on external financing by households and that were not. Following the macro-level analy-
nonfinancial firms is associated with sharper sis is a micro analysis, using bank-level data, of
downturns in the aftermath of financial stress. the procyclicality of investment banks’ and com-
• Countries with more-arm’s-length financial mercial banks’ leverage in both arm’s-length
systems appear to be vulnerable to sharper and relationship-based financial systems. The
contractions in economic activity in the wake chapter then focuses on six of the most well-
of banking stress, because leverage in the known episodes of banking-related financial
banking system appears to be more procy- stress and places the current financial turmoil
clical in countries characterized by greater in historical context. The concluding section
financial innovation. outlines some implications for policy.
• The importance of core financial intermedi-
aries in transmitting financial shocks to the
real economy suggests that policies that help Identifying Episodes of Financial Stress
restore the capital base of these institutions Financial systems—both financial institu-
within a strong framework of financial stability tions and the channels of intermediation—have
can help alleviate downturns. historically been prone to periods of rapid
• The patterns of asset prices and aggregate expansion followed by corrections. To better
credit in the United States during the current understand the impact of financial cycles on
episode of financial stress appear similar to the economy, it is useful to look for previous
those of previous episodes that were fol- episodes of financial stress that share common
lowed by recessions. In particular, changes in features with the current one.
the pattern of household net borrowing—a The current episode began in early 2007 as
measure of reliance on external financing— a generally orderly repricing of risk for assets
closely track the trajectory of past recessions. linked to U.S. subprime mortgages. By the
Nonfinancial firms entered the turmoil from summer, it had rapidly escalated into a severe
a relatively strong position. Combined with liquidity squeeze in the banking systems in the
the large losses sustained by core banking United States and western Europe and was caus-
institutions, these factors suggest that the ing serious dislocations in the interbank fund-
United States continues to face considerable
recession risks, even though real interest rates
are low by the standards of financial-stress-
driven recessions. In the euro area, house- See Kindleberger and Aliber (2005) for a history of
holds’ relatively strong balance sheets offer financial crises. A well-known exposition of this procy-
some protection against a sharp downturn, clical feature of financial systems is Minsky’s financial
despite the sizable increases of asset prices instability hypothesis (Minsky, 1992), which posits that
financial markets encourage businesses and individuals to
and credit ratios preceding the financial take on too much risk, generating ruinous boom-and-bust
turmoil. cycles.
131
ing market. The crisis mutated again more although studies for single countries are instruc-
recently, as heavy credit losses raised questions tive (Shiller, 1999).
about the capital strength of many banks. More- Although such narrative approaches to identi-
over, the stress has spread across various market fying financial crises provide a rich database of
segments in emerging as well as advanced episodes, they are less well suited to the pur-
economies, with these segments marked by a poses of this chapter for a number of reasons.
loss of liquidity, falling valuations, rising risk First, these are the episodes known to have
aversion, and heightened volatility. Foreign had large output consequences and/or to have
exchange markets have also been affected by required significant public intervention. Less
heightened uncertainty about the safety and attention has been given to “near misses”—epi-
soundness of global financial assets and the sodes of financial stress with little macroeco-
impact of financial stress on economic perfor- nomic impact—which could serve as useful
mance. Given these key features of the current counterfactuals. Second, the episodes identified
financial market turmoil, any characterization are typically of considerable length and involve
of previous financial stress episodes should take stresses of varying intensity, making it diffi-
into account conditions in the banking sector, cult to identify both when the financial stress
the state of nonbank intermediation through peaked and whether there was any meaningful
equities and bonds, and the behavior of foreign causal relationship to an economic downturn.
exchange markets. Finally, even the most comprehensive databases
The academic literature on financial crises focus on banking crises and currency crises and
has relied largely on historical narratives of pay little attention to pure securities market
well-known systemic banking crises, when bank stresses or liquidity squeezes, such as the U.S.
capital was eroded, lending was disrupted, stock market crash of 1987 and the collapse of
and there was often the need for significant U.S. hedge fund Long-Term Capital Manage-
public intervention (see, for example, Caprio ment (LTCM) in 1998. With leverage in bank-
and Klingebiel, 2003). An extension of this ing systems linked to securitization, it would
approach is to augment the narratives about seem important to simultaneously analyze the
banking crises with narratives of currency banking and securities channels of intermedia-
crises, when reserves were depleted and/or tion to determine the degree of interaction
there was a significant change in the exchange between the two.
rate mechanisms (see, for example, Kaminsky To overcome these limitations, this chapter
and Reinhart, 1999; and Reinhart and Rogoff, identifies episodes of financial stress as extreme
2008). Pure securities market stress episodes values of a composite variable—the “Financial
have not been examined as comprehensively, Stress Index” (FSI)—which is built using market-
especially those involving multiple countries, based indicators in real time and of high fre-
quency. The FSI for each country is constructed
as an average of the following indicators:
For a detailed analysis of the interbank funding mar- • three banking-related variables: the “beta” of
ket in the context of the current turmoil, see Chapters 1
and 2 of the October 2008 Global Financial Stability Report
banking sector stocks; the spread between
(IMF, 2008). The supply of liquid funds in the interbank
market dried up because many banks were in need of
such funds, and those with surplus funds refrained from A similar approach is used by Illing and Liu (2006).
lending, owing to concerns about the creditworthiness of The beta of banking stocks is a measure of the correla-
their counterparts. This affected both commercial banks, tion between the total returns to the banking sector stock
which rely largely on retail deposits to fund their lending index and the overall stock market index. A beta greater
activities, and investment banks and broker-dealers, which than 1—indicating that banking stocks move more than
rely more on wholesale funding markets. Universal banks, proportionately against the overall stock market—suggests
which combine features of both commercial and invest- that the banking sector is relatively risky. The FSI com-
ment banks, faced similar constraints. putes the betas as the coefficient on the rolling returns of
132
interbank rates and the yield on treasury First, asset-price-based variables are easy to
bills—the so-called TED spread, which mea- monitor and compute on a comparable basis
sures the premium banks charge each other across a large set of countries. Second, move-
over treasury bill rates; and the slope of the ments in broader financial asset prices can be
yield curve; expected to have a greater impact on the ability
• three securities-market-related variables: of financial firms to supply intermediation
corporate bond spreads, stock market returns, services than on the ability of specific nonfi-
and time-varying stock return volatility; and nancial firms to fund new investment, which is
• one foreign exchange variable: time-varying much more closely tied to developments in their
effective exchange rate volatility. particular sector. Third, it is useful to initially
The advantage of utilizing such an index is consider a broad range of financial stress events
the ability to more precisely date by quarter using asset prices and then use quantity-based
the start, peak, and end of a financial stress variables to identify which of those financial
episode and thereby to calculate its duration. stress episodes are associated with a significant
Moreover, such an index facilitates the identi- economic downturn. The underlying hypothesis
fication of four fundamental characteristics of is that only a subset of the universe of asset-
financial stress events: large shifts in asset prices price-based stress episodes reflects true underly-
(stock and bond market returns); an abrupt ing distress in the balance sheets of financial
increase in risk/uncertainty (stock and foreign intermediaries that have an impact on overall
exchange volatility); abrupt shifts in liquidity economic activity by restricting the supply of
(TED spreads); and the health of the banking credit; others merely reflect normal market
system (the beta of banking sector stocks and corrections.
the yield curve, which affects the profitability of Using the seven variables described above,
intermediating short-term liabilities into long- the FSI is constructed for each of the 17 coun-
term assets). Looking at these subcomponents tries in the sample. Episodes of financial
can help identify which types of financial stress stress are identified as those periods when the
episodes have been associated with larger output index for a country is more than one standard
consequences: banking-related, securities-mar- deviation above its trend (identified using
ket-related, currency-related, or a combination the Hodrick-Prescott filter), which signals
of these. that one or more of the banking, securities,
This chapter uses financial market (asset- and/or foreign exchange market variables
price-based) variables to identify periods when shifted abruptly.
the financial sector is under strain and its Overall, 113 financial stress episodes during
ability to intermediate may be impaired. This the past 30 years were identified in the 17 coun-
strategy has three major advantages over using tries considered in this chapter (Table 4.1). Of
a quantity-based index (an approach that the these, 43 episodes were driven mainly by stress
corporate finance literature might suggest). in the banking sector (that is, the banking vari-
ables accounted for most of the FSI increase),
50 episodes reflected primarily turmoil in the
each country’s banking sector subindex regressed on the
returns of the country’s overall stock market index. The securities market, and 20 episodes reflected
FSI considers the beta only during periods when returns primarily turmoil in the foreign exchange
are negative to focus on adverse shocks to banks. Accord- market. In some cases, stresses in one segment
ingly, in high-stress episodes, this indicator would reflect
an unusually large drop in banking stock prices relative of the financial system eventually migrated into
to overall market prices. the other segments. For example, in 17 of the
All variables are weighted by the inverse of their vari-
70 episodes that reflected stress primarily in
ance in order to reduce the impact on the overall index
of the more volatile variables. See Appendix 4.1 on the
the securities or foreign exchange markets, the
construction of the index. banking variables accounted for at least one-
133
Table 4.1. Descriptive Statistics on Financial Stress

Episodes
Duration of
Number of Episodes1 Episodes
Figure 4.2. Financial Stress Index (average;
(Shaded areas denote financial stress episodes) Total 1980s 1990s 2000s Current quarters)
Financial stress 113 37 42 34 16 2.4
The financial stress index appears to capture the current episode of financial stress
in all countries in the sample. Of which:
Banking 43 12 19 12 4 2.4
Securities markets 50 19 12 19 11 2.4
100 United States United Kingdom 100 Foreign exchange 20 6 11 3 1 2.4
95 Memo:
90 Banking related 60 16 25 19 10 2.6
90 Of which:
80 Above median arm’s
85 length 31 9 13 9 4 2.4
70
Below median arm’s
80 length 27 7 11 9 5 2.9
60 Source: IMF staff calculations.

75 1The countries included are Australia, Austria, Belgium, Canada,
Denmark, Finland, France, Germany, Italy, Japan, Netherlands, Norway,

70 50 Spain, Sweden, Switzerland, United Kingdom, and United States.
1980 85 90 95 2000 05 08: 1980 85 90 95 2000 05 08:
Q3 Q3
100 Japan Germany 80

third of the spike in the FSI. This implies that
90 there are actually 60 episodes in the sample with
70
banking-related financial stress—episodes for
80 which banks were either the most or the second
60
most important factor, contributing at least one-
70
third of the FSI spike.
50 Overall, the FSI appears to accurately capture
60
global financial episodes. The current finan-
50
1980 85 90 95 2000 05 08: 1980 85 90 95 2000 05 08:
40 cial crisis, which began in 2007, has a signifi-
Q3 Q3 cant global dimension, affecting virtually all
countries in the sample (Figures 4.2 and 4.3).
90 France Canada 100
Earlier episodes that simultaneously affected the
majority of countries in the sample include the
90
80 1987 stock market crash, the Nikkei/junk bond
80 collapse in the late 1980s, the Scandinavian
70 banking crises in 1990, the European exchange
70
rate mechanism crisis of 1992, and the collapse
60
60
of LTCM, but the current episode seems to have
the widest impact.
50 50 The FSI also accurately captures the fact
1980 85 90 95 2000 05 08: 1980 85 90 95 2000 05 08:
Q3 Q3 that, although the origins of the current epi-
sode were in the banking sector, by early 2008
Source: IMF staff calculations. the crisis had broadened significantly to affect
Overall, of the 113 episodes of financial stress identi-

fied in the sample, 87 episodes simultaneously affected
two or more countries.
134
the securities and foreign exchange markets as Figure 4.3. Financial Stress and Shocks
well (Figure 4.4). Moreover, the index indi-
cates that all past episodes of banking-related Financial stress episodes tend to be synchronized around systemic events, as shown
by their clustering in a few peaks. Episodes of financial stress generally correspond
stress also had significant securities market to cases of monetary policy shocks, but they don't correlate well with other shocks.
components.
Share of Countries Experiencing Financial Stress 100
The country-specific FSIs for the 17 countries Scandinavian
Nikkei/junk LTCM
considered in this chapter show that the FSI bond collapse banking collapse 80
crises
peaks, which correspond to periods of financial Stock
market 60
stress, generally overlap accurately with well-
crash ERM
known financial stress episodes in these coun- crisis 40
tries over the past three decades, including the 20

current episode (see Figure 4.2). Specifically,
0
the FSI captures over 90 percent of the banking 1980 85 90 95 2000 05 08:
Q1
crises and over 80 percent of the currency crises
Oil Prices1 60
identified in the literature.
Given that spikes in the FSI are associated 40
with well-known events such as stock/bond mar- 20

ket collapses or banking crises, could the index 0
simply mirror changes in other fundamentals -20
that can directly affect the economic cycle? Con-
-40
sidering four types of such shocks—to oil prices,
-60
labor productivity, fiscal policy, and monetary 1980 85 90 95 2000 05 08:
Q1
policy—the evidence indicates that spikes in the
Macroeconomic Policies 4
FSI are not correlated with oil, labor productiv-
Fiscal3 3
ity, or macroeconomic policies (see Figure 4.3, Monetary2
2
bottom three panels).10 There does appear to be
1
a greater correlation with monetary policy, but
0
this can be expected, given that monetary policy
-1
(as measured by the term spread) is a subcom- -2
ponent of the FSI itself, where it is included -3
because financial stress appears to be associated 1980 85 90 95 2000 05 08:
Q1
Labor Productivity4 4
3
2
The FSI captures 100 percent of all episodes identified 1
in the literature if the duration of episodes is interpreted 0
more broadly—that is, if the period around the peak of
-1
financial stress is expanded by a few quarters.
10The definition of these shocks mirrors the definition -2
of financial stress episodes: (1) Shocks to oil prices or -3
labor productivity are defined as instances when changes -4
1980 85 90 95 2000 05 08:
in these indicators are one standard deviation above Q1
trend; (2) fiscal policy shocks, when the ratio of govern-
Sources: Haver Analytics; IMF, Commo dity Price System database; OECD, Analytic
ment net lending/borrowing to GDP is one standard Database; OECD, Economic Outlook (2008); and IMF staff calculations.
deviation above trend; and (3) monetary policy shocks, Note: ERM = exchange rate mechanism; LTCM = Long-Term Capital Management.
1 Oil prices are scaled by U.S. inflation (CPI) and represent deviations from
when the inverse term spread is one standard devia-
tion above trend. In all cases, the deviations from trend Hodrick-Prescott (HP) trend.
2 Monetary policy is measured using the inverse term spread’s deviations from HP trend.
are calculated using Hodrick-Prescott filters. Figure 4.3 3Fiscal policy is measured using government net lending’s deviations from HP trend.
reports the first principal components of these fiscal, 4Labor productivity of the total economy is measured as the ratio of real GDP and total
monetary, and labor productivity indicators across the set employment and represents deviations from HP trend. Data are not available for Austria,
of countries considered in the chapter. Belgium, Denmark, Spain, and Switzerland.
135
with the monetary policy stance as reflected in

the term spread.
Overall, the FSI appears to be a comprehen-
sive indicator that successfully identifies the
main episodes of financial stress for the sample
of countries under consideration and provides a
sound basis for examining the macroeconomic
Figure 4.4. Contribution of Banking, Securities, and consequences of such stress.
Foreign Exchange to Current Financial Stress Episode1
The episode of financial stress that started in 2007 has become more broad-based,
with contributions from banking, securities, and foreign exchange markets. Previous Financial Stress, Economic Slowdown,
episodes also show strong contributions from both banking and securities markets. and Recession
Banking
How many of the financial stress episodes
Securities identified using the FSI were followed by an
Foreign Exchange economic slowdown or an outright reces-
sion? How did episodes that were followed by
Average of Most Recent Quarter
Previous Episodes of Current Episode
economic downturns differ from those that
were not?
Australia • An episode of financial stress is followed by
Austria an economic “slowdown” if the level of real
GDP falls below trend (identified using the
Belgium
Hodrick-Prescott filter) within six quarters of
Canada
the onset of financial stress.
Denmark • An episode of financial stress is followed by
France a “recession” if a peak-to-trough business
cycle, identified using the methodology
Germany
described in Harding and Pagan (2002) and
Italy the April 2003 World Economic Outlook, begins
Japan within six quarters of the onset of financial
Netherlands
stress.11
This chapter seeks to identify the main char-
Norway
acteristics of financial stress episodes that were
Spain eventually followed by economic downturns,
Sweden not to assess whether financial stress “causes”
economic downturns, in recognition of the
Switzerland
significant analytical and empirical challenges in
United Kingdom
establishing causality.12 Nevertheless, the analysis
United States attempts to control, to some extent, for other
0 10 20 30 40 50 0 10 20 30 40 50 shocks—namely, monetary, fiscal, oil price, and
Source: IMF staff calculations.

1The sum of the components equals the level of the Financial Stress Index in the given 11For example, in the United States, the most recent
period. See Appendix 4.1 for details.
recession was in 2001, whereas the most recent slowdown
was when GDP fell below trend during 2007:Q4–2008:Q1.
12For example, many shocks affect both the financial
system and the economy, and although the financial

system may amplify the shocks, it would be hard to disen-
tangle the direct effects of the shocks from the amplifica-
tion effects.
136
Financial Stress, Economic Slowdown, and Recession
Table 4.2. Descriptive Statistics on Financial Stress, Slowdowns, and Recessions

Lag until
Duration (average; quarters) Output Loss
Downturn3
Financial Slowdown or (average; percent of GDP)
(average;
N stress recession1 Cumulative2 Average quarters)
Financial stress 113 2.4
Of which:
Followed by slowdown 29 2.7 7.6 –7.6 –0.7 2.4
Banking-related 18 3.2 8.4 –9.3 –0.8 1.8
Followed by recession 29 3.0 6.8 –13.8 –1.2 2.3
Banking-related 17 4.0 7.6 –19.8 –1.5 2.0
Others 55 2.0
Slowdown not preceded by financial stress4 109 5.1* –4.1* –0.6
Recession not preceded by financial stress4 31 3.1* –5.4* –0.9
1Slowdown duration: number of quarters during which GDP is below trend; recession duration: number of quarters until GDP is at or exceeds
peak level.
2Slowdown output loss: cumulative output loss below trend; recession output loss: cumulative output loss until recovery.
3Number of quarters between start of financial stress and slowdown or recession.
4Asterisks indicate difference from slowdowns preceded by financial stress significant at 10 percent or better.
labor productivity shocks—that may affect the The occurrence of financial stress tends to
relationship between financial stress and eco- change the patterns of downturns (Figure 4.6),
nomic cycles. as apparent by examining the dynamics of
Based on the definitions above, of the 113 selected macroeconomic variables at the begin-
financial stress episodes identified here, 29 were ning of the downturn.13 In particular, when
followed by slowdowns and 29 by recessions. The preceded by financial stress, economic slow-
remaining 55 episodes were not followed by an downs tend to be characterized by a flattening
economic downturn (Table 4.2). in consumption growth, by investment that
The average lag between the onset of follows a boom-and-bust cycle, by appreciable
financial stress and the subsequent downturn turnarounds of current account balances, and
was about seven months. However, this aver- by falling inflation and real interest rates. The
age masks substantial variations: about half the pattern changes are more pronounced for
downturns occurred within a quarter of the slowdowns and less pronounced for recessions,
beginning of financial stress, but it took more perhaps suggesting that the latter may be trig-
than a year for a downturn to materialize dur- gered to a greater extent by the interaction of
ing one-fourth of the episodes (Figure 4.5). financial stress with other shocks.
Most important, the slowdowns and reces-
sions that were preceded by financial stress
episodes were longer in duration and, partly as Why Are Some Financial Stress Episodes
a result, were more severe than those that were Associated with Economic Downturns?
not. Median cumulative output losses (relative Only about half the episodes of financial
to trend or until recovery) were about 3 per- stress identified using the FSI were followed by
cent of GDP for slowdowns following financial economic slowdowns or recessions. What deter-
stress and about 4½ percent of GDP for reces-
sions following financial stress, significantly
larger than for slowdowns and recessions that 13The charts use a window of 12 quarters and show
were not preceded by financial stress (about only “complete” episodes (episodes spanning at least 12
quarters from start to finish). Therefore, they include
1½ percent and 2¼ percent, respectively) (see only those downturns that started between 1983:Q1 and
Figure 4.1). 2005:Q1.
137
mines whether a financial stress episode will be

followed by a downturn? What characterizes the
stress episodes that precede the most severe and
prolonged downturns? To answer these ques-
tions, those financial stress episodes that were
followed by downturns are compared against
each other along two dimensions: (1) the char-
acteristics of the financial stress episode itself
and, in particular, the nature of the financial
shock (whether related to the banking system,
securities market, or foreign exchange market)
and (2) the financial position of financial inter-
Figure 4.5. Lag between Financial Stress and Downturns mediaries, households, and firms at the begin-
(Start of financial stress episode at t = 0; quarters on the x-axis; number of
ning of the episode.
slowdowns or recessions that begin in the given quarter on the y-axis)
In most cases economic downturns materialize soon (within one quarter) after the
start of a financial stress episode, but longer lags have been observed. Is Banking-Related Financial Stress Different?
Banking system stress is associated with
16
larger output consequences than stress epi-
sodes related purely to the securities or foreign
14
Number of recessions exchange markets, in which the banking system
Number of slowdowns remains largely unaffected (Figure 4.7). About
12
60 percent of those financial stress episodes that
10 are followed by downturns are banking-related.
Moreover, slowdowns and recessions preceded
8
by banking-related stress tend to last longer and
6
be associated with larger average GDP losses
than those preceded by other types of finan-
4 cial stress or by no financial stress at all (see
Table 4.2).
2 Bank asset growth slows significantly when
banking-related financial stress episodes are
0
0 1 2 3 4 5 6 followed by recessions or slowdowns, compared
with financial stress episodes that are not fol-
Source: IMF staff calculations. lowed by economic downturns (Figure 4.8,
top panel). In general, downturns tend to be
associated with a fall in the demand for credit,
but during slowdowns or recessions associated
with banking-related financial stress, the cost
of capital is significantly higher (Figure 4.8,
bottom panel).14 While the issue of reverse
14The cost of capital is defined here as a weighted aver-

age of the real cost of equity, the real cost of debt, and
real lending rates, using as weights the relative shares of
equity, bonds, and loans in nonfinancial corporate liabili-
ties. See Appendix 4.1 for details.
138
Financial Stress, Economic Slowdown, and Recession
causality between recessions and financial stress

is difficult to address empirically, suggesting
appropriate caution in interpreting these results, Figure 4.6. Selected Macrovariables around Economic
Downturns with and without Financial Stress1
these findings are consistent with the view that
(Median; start of economic contraction at t = 0; quarters on the x-axis)
a reduction in the supply of credit—the classic
credit crunch or credit squeeze—is a key factor The occurrence of financial stress changes the patterns of economic downturns. In
associating banking-related financial stress epi- particular, the shape of slowdowns seems to be affected.
sodes with economic downturns.
Slowdowns preceded Slowdowns not preceded
by financial stress by financial stress
Recessions preceded Recessions not preceded
Do Initial Conditions Matter? by financial stress by financial stress
Having examined the nature of the shocks, the

110 Consumption Residential Investment 112
analysis now assesses whether the likelihood of (100 at t = 0) (100 at t = 0)
108 110
a downturn depends on initial conditions. The 108
106
evidence suggests that the credit and asset price 106
104
dynaamics and the financial position of financial 102
104
intermediaries, households, and firms preceding 102
100
100
the episode are important in determining the 98 98
economic impact of a financial shock. 96 96
• House prices and the credit-to-GDP ratio 94 94
both tend to rise significantly faster during 92
-12 -8 -4 0 4 8 12
92
-12 -8 -4 0 4 8 12
the upswing of the financial cycle in those
stress episodes that are followed by slowdowns 120 Nonresidential Investment Current Account 2.0
or recessions (Figure 4.9). Statistical tests (100 at t = 0) (percent of GDP)
115 1.5
confirm that financial turmoil is more likely 110
1.0
to be followed by economic slowdown or 105
outright recession when it is preceded by a 0.5
100
more rapid buildup in house prices and credit 0.0
95
(Figure 4.10). -0.5
90
• Firms tend to be more heavily dependent -1.0
85
on external sources of funding—that is, with
80 -1.5
higher net borrowing ratios—in the run-up -12 -8 -4 0 4 8 12 -12 -8 -4 0 4 8 12
to financial stress episodes that are followed
Inflation Change in Real Interest Rates
by economic downturns (see Figures 4.9 and 6
(percent difference in CPI from (percentage points)
2.0
4.10). A higher initial reliance on external one year earlier) 1.5
funding makes firms more vulnerable dur- 5 1.0
ing the downswing of the financial cycle, and 0.5
may force them to adjust their spending plans 4 0.0
more drastically in the aftermath of financial -0.5
stress, setting the stage for a larger impact on 3 -1.0
the real economy. -1.5
• Only financial stress episodes that are fol- 2 -2.0
-12 -8 -4 0 4 8 12 -12 -8 -4 0 4 8 12
lowed by recessions (not by slowdowns) seem
to be characterized by a “more exposed” Sources: Haver Analytics; IMF, International Financial Statistics database; OECD, Analytic
household sector in terms of reliance on Database; and IMF staff calculations.
1 All in real terms. The sample is constant for all quarters.
external financing (see Figure 4.9). Indeed,
the median household net borrowing ratio
139
(in deviation from trend) is significantly

higher at the outset of financial stress epi-
sodes that are followed by recessions than
during those that are followed by slowdowns
Figure 4.7. Banking-Related Financial Stress, or by no decline in economic activity (see
Slowdowns, and Recessions
Figure 4.10).
(Median; quarters on the x-axis)
The analysis suggests that when the finan-
The most severe economic downturns are associated with banking-related financial cial cycle turns, as signaled by the onset of
stress. stress in financial markets, there is a greater
likelihood of a downturn in economic activ-
Financial Stress1
(real GDP percent change from one year earlier) ity the larger the initial financial imbal-
4.0
Non-banking-related ances—when firms and households are more
financial stress All financial stress episodes 3.5 exposed to a decline in credit and asset prices.
3.0 In particular, households’ exposure seems to
2.5 be associated with more severe contractions in
2.0
economic activity.
Banking-related The degree to which initial financial imbal-
financial stress 1.5
ances and firms’ and households’ reliance on
1.0
external funding explain the severity of ensu-
-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12
0.5 ing economic downturns can be investigated
Quarters before and after the start of financial stress more formally. The cumulative loss of output
in the aftermath of financial stress episodes is
Slowdowns and Recessions2 regressed on (1) the run-up in credit and asset
(real GDP percent change from one year earlier)
prices before the onset of the financial stress,
Slowdowns preceded Slowdowns not preceded
by banking-related financial stress by financial stress (2) firms’ and households’ net borrowing ratios
Recessions preceded Recessions not preceded at the start of the episodes, and (3) a proxy for
by banking-related financial stress by financial stress
5 the severity of the financial shock, namely, the
4
duration of the stress episode.
The main results of the regressions are pre-
3
sented in Table 4.3:15
2
• Firms’ net borrowing ratio at the onset of the
1 financial stress episode enters significantly
0 in almost all specifications, confirming the
-1 importance of the link between firms’ initial
reliance on external credit and the severity of
-2
-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 the decline in economic activity.
Quarters before and after the start of economic downturn
• Households’ net borrowing ratio at the onset
of the financial stress episode is statistically
Sources: Haver Analytics; OECD, Analytic Database; OECD, Economic Outlook (2008); and
IMF staff calculations. significant when considered alone but loses
1 The difference between banking-related and non-banking-related episodes is significant
at a minimum of 10 percent for quarters 0, 1, 2, 3, and 4. The sample is constant for all significance when firms’ net borrowing
quarters.
2 The difference between slowdowns preceded by banking-related financial stress and
position is added. It continues to affect the
slowdowns not preceded by financial stress is significant at a minimum of 10 percent for severity of the output losses, however, when
t – 6 to t + 6. The sample is constant for all quarters.
15Slowdown severity is measured using the cumulative
output loss during the period during which output is

below trend; see Table 4.1 for further details. Recession
severity is measured by losses until recovery.
140
Has Financial Innovation Affected the Interplay between Financial Stress and Economic Cycles?
interacted with the duration of the financial

episode, suggesting that households’ position
matters especially when the economy is hit by
a sustained financial shock.
This chapter seeks to identify factors that
determine whether or not episodes of financial
stress are followed by economic downturns.
Therefore, a key challenge is to determine the Figure 4.8. Cost of Capital and Bank Asset Growth
origins of the shocks that hit the economy and around Banking Financial Stress Episodes
set off the complex interactions between the (Average; change from one year before to one year after the start of
financial stress)
financial and real sectors. The fact that finan-
cial sector stress precedes an economic down-
Banking-related financial stress episodes associated with recessions are
turn does not mean that financial stress drives characterized by lower bank asset growth and higher cost of capital in the aftermath.
subsequent real sector developments: because
financial market participants are forward-look- Change in Bank Assets1
(percent) 0.25
ing, financial stress may be merely a mani-
p-value = 0.05
festation that they anticipate a fundamental 0.20
deterioration in the real sector. To address this
problem, the chapter considers four types of 0.15
p-value = 0.05
shock that could be considered fundamental,
0.10
exogenous deteriorations in the real sector
environment: oil-price shock, monetary policy 0.05
shock, fiscal policy shock, and labor productiv-
ity shock. As shown in Figure 4.11, when one of 0.00
Recessions2 Slowdowns 2 Other 3
these shocks is combined with financial stress,
there is a more severe downturn than when Change in Cost of Capital1
(percentage points) 1.2
there is the shock alone. This suggests that
financial stress has a separately identifiable 1.0
impact. 0.8
p-value = 0.00
0.6
Has Financial Innovation Affected the 0.4
Interplay between Financial Stress and 0.2
Economic Cycles? 0.0
Banks continue to be important in explain- -0.2

Recessions2 Slowdowns2 Other 3
ing why certain financial stress episodes are
associated with greater output consequences Sources: OECD, Analytic Database; and IMF staff calculations.
1 p-value reported refers to the two-tailed t-tests on the difference between the two
(see Figure 4.7). Why do banks remain crucial averages.
despite financial innovation and the emer- 2Banking-related financial stress episodes followed by slowdowns or recessions.
3 Financial stress episodes not followed by slowdowns or recessions.
gence of nonbank sources of funding? Finan-
cial innovation would seem able to reduce the
pivotal role of banks by providing alternative
channels for firms and households to access
financing, channels that loosen collateral con-
straints for borrowers and soften the adverse
impact of financial stress on the cost of capital
for banks. However, even though the role of
141
Figure 4.9. Selected Macrovariables around Financial banks has evolved over time, their symbiotic
Stress Episodes1 relationship with securities markets remains
(Median; start of financial stress episode at t = 0; x-axis as stated) an essential feature of many financial systems,
especially those characterized by arm’s-length
Financial stress episodes followed by slowdowns or recessions tend to be preceded
by rapid buildups in asset prices and credit ratios and are associated with higher
financing (World Economic Outlook, Septem-
initial net borrowing from nonfinancial firms and (in cases of recession) from ber 2006).16 As a result, episodes of banking
households.
distress continue to affect nonbank sources of
financing as well.
Followed by slowdowns only
Followed by recessions only To explain the continuing importance of
Others banks, it is revealing to explore the procycli-
cal behavior of bank leverage around financial
Stock Prices House Prices
cycles. In particular, how banks manage their
20 (percent deviation from trend; (percent deviation from trend; 4
quarters on the x-axis) quarters on the x-axis) leverage during upturns and downturns in the
15 3
cycle appears fundamental to explaining why
10 2 banking stress translates into a reduced credit
5 1 supply, a higher cost of capital, and a soften-
0 0 ing of economic activity. The hypothesis is as
-5 -1
follows: When banks overextend their balance
sheets during booms, on the back of higher
-10 -2
asset values and lower perceived risk, there is
-15 -3
-12 -8 -4 0 4 8 12 -12 -8 -4 0 4 8 12 a buildup of financial imbalances and a rapid
expansion in activity, which further boosts asset
Credit/GDP Interest Rate values and reduces perceived risk, thereby fos-
3 (percent deviation from trend; (change from one year 2.5
quarters on the x-axis) earlier; percentage points; tering another round of lending and economic
2.0
quarters on the x-axis) expansion.17 Under such conditions, a financial
2 1.5
1.0
shock that either increases risk or reduces the
1
0.5 return on assets could prompt a cycle of severe
0.0 deleveraging, with banks sharply reducing their
0
-0.5 lending (or their growth in lending) as bank
-1 -1.0 capital falls, prompting an economic slowdown
-1.5 that feeds back into a further reduction in
-2 -2.0 credit supply.
-12 -8 -4 0 4 8 12 -12 -8 -4 0 4 8 12
The procyclicality of leverage is more pro-
Household Net Nonfinancial Corporate Net nounced when banks are more exposed to
1.5 Lending/Borrowing Lending/Borrowing 1.0
(deviation from trend; years (deviation from trend; years fluctuations in the market value of assets—for
1.0 on the x-axis) on the x-axis) example, through their holdings of securities and
0.5
0.5
0.0
0.0
16Banks increasingly depend on market-based funding
-0.5
-0.5 sources to finance their assets (such as through their
-1.0 certificates of deposit and off-balance-sheet commercial
-1.0 paper programs). Conversely, investment banks and
increasingly commercial banks also remain at the center
-1.5 -1.5
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 of the originate-to-distribute model of securitized financ-
ing, providing credit through repurchase facilities to
Sources: IMF, International Financial Statistics database; OECD, Analytic Database; and hedge funds and other leveraged intermediaries to invest
IMF staff calculations. in securities markets.
1All in real terms, except for household and nonfinancial corporate net lending ratios. The 17This is in line with Minsky’s financial instability
sample is constant for all quarters and years.
hypothesis (Minsky, 1992).
142
Has Financial Innovation Affected the Interplay between Financial Stress and Economic Cycles?
their repurchase facilities.18 Because this is typical

for nondepository financial intermediaries, espe-
cially investment banks, there should be evidence
of procyclical leverage among them (Shin, 2008).
On the other hand, commercial banks should be
less prone to procyclically adjusting their balance Figure 4.10. Initial Conditions of Financial Stress
sheets during asset price or liquidity booms and Episodes
(Average)
busts, because they rely less on wholesale fund-
ing and more on retail deposits, and also because Faster buildup in the credit ratio and house prices and higher nonfinancial corporate
they are less subject to mark-to-market changes in borrowing are statistically significant indicators of the likelihood that financial stress
episodes will be followed by slowdowns or recessions. Household net borrowing is
the value of their assets. significantly associated only with subsequent recessions.
Evidence confirms that investment banks’
leverage tends to be procyclical: they expand Household Net Nonfinancial Corporate
Lending/Borrowing Net Lending/Borrowing
their leverage when they are expanding their (percent of gross disposable (percent of GDP; deviation from
assets (Figure 4.12, upper panel).19 The evi- income; deviation from trend trend one year before start of
one year before start of financial stress)
dence is less uniform on the procyclicality 0.5 financial stress) 0.0
of commercial banks, which rely much more p-value = 0.04
on retail deposits and whose main activity is 0.0

-0.2
making long-term, illiquid loans (Figure 4.12,
-0.5
lower two panels). However, the evidence does -0.4
suggest that commercial banks tend to be more -1.0
procyclical when operating in more-arm’s- p-value = 0.05
-0.6
length financial systems, where a greater share -1.5
p-value = 0.02
of intermediation occurs through financial
-2.0 -0.8
markets rather than through traditional rela- Recessions Slowdowns Other1 Recessions Slowdowns Other1
tionship-based (and bank-dominated) activities

Real House Prices Credit
(Figure 4.13). Thus, more-arm’s-length financial (cumulative percent (percent of GDP; cumulative
systems are associated with overall more pro- deviation from trend over six percent deviation from trend
quarters before start of over six quarters before start
cyclical bank behavior, and as a result may be 30
financial stress) of financial stress)
8
more vulnerable to banking stress.20 Of note, 6

20
the proportion of banking-related financial p-value = 0.00 4
stress episodes occurring in more-arm’s-length 10 2
p-value = 0.00
financial systems has remained about equal to
0 0
the proportion occurring in more-relationship-
p-value = 0.05 -2
based systems (see Table 4.1).
-10
Indeed, slowdowns and recessions tend to be p-value = 0.00 -4
deeper in economies with more-arm’s-length -20 -6

Recessions Slowdowns Other1 Recessions Slowdowns Other1
financial systems, although the duration of such
Sources: Haver Analytics; IMF, International Financial Statistics database; OECD, Analytic
18In Database; and IMF staff calculations.
a systemic crisis, it would be difficult for all banks Note: p-value reported refers to the two-tailed t-tests on the difference between the two
to adjust their leverage simultaneously, because there averages.
1Financial stress episodes not followed by slowdowns or recessions.
would be few buyers for these assets among other banks;
the only likely buyers would be cash-rich investors who do
not rely on bank leverage to fund their positions.
19See also Adrian and Shin (2008a and 2008b).
20This is consistent with the findings in Chapter 3 of
IMF (2008), showing that fair value accounting tends to

lead to more procyclical movements in financial interme-
diaries’ balance sheets.
143
Table 4.3. Cross-Section Regressions

Dependent Variable: Cumulative Output Loss1 [1] [2] [3] [4] [5] [6] [7] [8] [9]
Duration of financial stress 1.324 1.591 0.641 1.023 0.963 1.377 1.034 0.888 0.711
(0.10) (0.10) (0.43) (0.33) (0.30) (0.09) (0.20) (0.07) (0.30)
Real interest rate2 0.89 0.808 0.47 0.835 0.877 0.87 0.841 0.887 0.869
(0.00) (0.01) (0.12) (0.01) (0.00) (0.00) (0.01) (0.00) (0.01)
Nonfinancial corporate net borrowing ratio3 2.0 1.753 0.849 1.986 1.439
(0.01) (0.02) (0.35) (0.00) (0.07)
Household net borrowing ratio3 1.05 0.48 0.668 –1.086 –0.89
(0.10) (0.44) (0.33) (0.22) (0.32)
Real house price4 14.304
(0.01)
Real stock price4 –0.785
(0.72)
Credit ratio4 –1.09
(0.90)
Household net borrowing ratio x duration 0.593 0.561
(0.04) (0.04)
Nonfinancial corporate net borrowing ratio x duration 0.229 0.136
(0.43) (0.48)
Constant –2.014 –0.803 2.076 0.809 0.482 –1.877 –0.727 –1.161 –0.519
(0.45) (0.79) (0.39) (0.74) (0.86) (0.50) (0.78) (0.64) (0.85)
N 42 40 52 52 52 40 40 40 40
Adjusted R-squared 0.418 0.287 0.254 0.128 0.126 0.418 0.42 0.493 0.485
Note: Robust p-values in parentheses.
1Dependent variable is cumulative output loss in episodes of financial stress followed by slowdowns or recessions.
2Average of real interest rates during six quarters before the financial stress.
3Net borrowing ratios one year before the financial stress (deviation from trend).
4Cumulative percent deviation from trend over six quarters before the financial stress.
downturns is broadly similar in both types of the availability of the twin engines for financial
systems (Figure 4.14), indicating that deleverag- intermediation (banks and markets).
ing matters and that its impact depends on the Arm’s-length financial systems do offer sev-
degree of procyclicality in the banking system.21 eral advantages over relationship-based systems
This implies that more-arm’s-length systems are in terms of reallocating resources in response
vulnerable to sharper contractions in activity fol- to changing economic opportunities (see World
lowing banking stress. Consistent with this chan- Economic Outlook, October 2006). However, as
nel, the leverage of banks in more-arm’s-length the current crisis underscores, the trend toward
systems also tends to fall more sharply than that greater securitization in more-arm’s-length sys-
of banks in other types of financial systems, tems, while permitting portfolio diversification
albeit from a lower starting level. This casts to offset the costs of monitoring the idiosyn-
doubt on the presumption that arm’s-length sys- cratic risks that are inherent in traditional rela-
tems can better soften the blow from financial- tionship-based systems, does not eliminate the
stress-driven economic downturns because of need for banks and markets to independently
assess the risk of their exposures. Indeed, a
21 Diverging experiences with economic cycles by lack of information about the value and risk of
economies characterized by arm’s-length versus relation- many securitized products, and about the losses
ship-based financial systems may also reflect contrasts in subsequently associated with these products,
other areas, notably in the degree of flexibility in labor
and product markets and the types of social welfare sys- appears to have played a significant role in
tems (see World Economic Outlook, October 2006). amplifying the current crisis.
144
The Current Financial Crisis in Historical Context
The Current Financial Crisis in Historical

Context
This section compares the current episode
of financial stress to six well-known episodes of
banking-related financial stress that occurred
Figure 4.11. Financial Stress and Economic Downturns:
in advanced economies during the 1990s. These
Controlling for Four Main Shocks
episodes affected Finland, Norway, Sweden, the (Median of GDP percent change from one year earlier; start of economic
United Kingdom, and the United States in the contraction at t = 0; quarters on the x-axis)
early 1990s and Japan throughout the 1990s.
Given the importance of banking distress in the Economic downturns tend to be more severe when preceded by financial stress
episodes.
current financial turmoil, these episodes can
serve as a useful benchmark for analyzing the
Preceded by financial Preceded by shock only
current conjuncture and for gauging its poten- stress and shock
tial macroeconomic impact.22
Examining the initial conditions before the 5 Oil-Price Shock1 Monetary Policy Shock 2 5
onset of these six episodes confirms the main
4
4
findings of the event analysis, namely, that
financial stress episodes are more likely to be 3
3
followed by severe economic downturns when 2
they occur in the context of a rapid buildup in 2
1
credit and house prices and a heavier reliance
1
on credit by firms and households. Broadly 0
speaking, the episodes characterized by the 0 -1

-12 -8 -4 0 4 8 12 -12 -8 -4 0 4 8 12
largest asset price and credit booms were
in the Nordic countries in the early 1990s,
5 Fiscal Policy Shock 3 Labor Productivity 4 4
where increases in credit ratios, assets prices,
and bank assets were abnormally high (see 4
3
Table 4.4, top panel). At the same time, borrow- 3
ing by households and firms was initially much 2
2
larger in Finland and Norway than in the other
1
countries. Japanese firms relied extensively on 1
external financing, but this was balanced at 0
0
least partially by the saving patterns of house-
holds. In contrast, before the crises in the -1
-12 -8 -4 0 4 8 12 -12 -8 -4 0 4 8 12
-1
Anglophone countries, asset price imbalances

were moderate and balance sheets were not Sources: Haver Analytics; IMF, Commo dity Price System database; OECD, Analytic
Database; OECD, Economic Outlook (2008); and IMF staff calculations.
under tremendous strain. 1 Oil prices are scaled by U.S. inflation (CPI). Oil-price shock is identified if oil price is
The countries with larger financial imbalances one standard deviation above Hodrick-Prescott (HP) trend.
2 Monetary policy is measured using the inverse term spread. Monetary policy shock is
and balance sheet vulnerabilities at the onset identified if the inverse term spread is one standard deviation above HP trend.
3 Fiscal policy is measured using government net lending. Fiscal shock is identified if
of an episode experienced more severe output government net lending is one standard deviation above HP trend.
4 Labor productivity of the total economy is measured as the ratio of real GDP and total
contractions (see Table 4.4, lower panel). The
employment. Productivity shock is identified if labor productivity is one standard deviation
most dramatic collapses in asset prices, bank below HP trend. Data are not available for Austria, Belgium, Denmark, Spain, and
Switzerland.
asset growth, and credit occurred in the coun-
22Thisapproach is in line with other studies, includ-

ing Reinhart and Rogoff (2008). See also J.P. Morgan
Research (2008).
145
tries with the largest buildups of financial imbal-

Figure 4.12. The Procyclicality of Leverage in ances. These countries also suffered the deepest
Investment and Commercial Banks and longest recessions.
(Annual change, percent) Moreover, households and firms in these
countries also generally experienced a stronger
Investment banks’ leverage (assets divided by equity) is procyclical; for commercial
banks, the evidence is mixed across countries.
deleveraging process. Deleveraging to reduce
the stock of debt through higher savings and
Linear regression consequently lower consumption and invest-
Linear regression, excluding outliers ment had direct implications for the dynamics
Median regression
of growth. The degree of corporate deleverag-
ing, in particular, corresponds remarkably well
Top 50 Investment Banks1 1.5
B = 0.47 with the length and depth of the recessions—
R 2 = 0.26 1.0 underscored by the contrast between the Nordic
0.5
and Anglophone country groups.
Asset growth
These historical experiences underline the
0.0
key role of policy responses to financial stress.
-0.5 Policies appropriate to restore sound financial
intermediation are discussed in Box 4.1, based
-1.0
around four main principles. First, there must
-1.5 be a sound framework in place for ensuring
-4 -3 -2 -1 0 1 2 3
Leverage growth financial stability, which encompasses a frame-
work for intervention and appropriate legal,
Commercial Banks: United States (top 50 banks) 1 0.6
B = 0.32 institutional, and procedural mechanisms to
R 2 = 0.12 deal with distress. Second, policy responses must
0.4
be rapid and involve the early recognition of
Asset growth
0.2
losses, a quick assessment of the scale of the
0.0 problem, and timely measures to ensure that
financial institutions are adequately capitalized.
-0.2
Third, the adverse impact of financial stress on
-0.4 the real economy may need to be contained
directly, in order to preserve or restore the
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5
Leverage growth health of households’ and firms’ balance sheets.
Finally, adequate safeguards must be in place
Commercial Banks: Germany (top 10 banks)1 0.4 to limit the fiscal cost of support and prevent
B = 0.18
R 2 = 0.04 the creation of inappropriate incentives for the
0.2 longer term that could lead to excessive reliance
Asset growth
on publicly financed bailouts.

0.0
Implications for the Current Crisis in the United

-0.2
States and Euro Area
-0.4 Figure 4.15 compares data for the current
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5
crisis in the United States and euro area against
Leverage growth
the medians of selected macroeconomic vari-
Sources: Bankscope; and IMF staff calculations. ables around the beginning of the six major
1B and R 2 refer to the linear regression, excluding outliers, of asset growth over
leverage growth. financial stress episodes examined above and
against the averages for these variables across all
146
The Current Financial Crisis in Historical Context
financial stress episodes that were followed by

recessions. The current imbalances and adjust-
ments appear generally much smaller than those
for the six episodes examined here, except for
U.S. residential real estate investment and the
U.S. current account.23 The patterns of credit
and asset prices in the United States prior to
the current crisis are very similar to those for
the typical financial-stress-driven recession.
The deleveraging process by households in the
United States is proceeding faster than in the Figure 4.13. Procyclical Leverage and Arm’s-Length
typical recession, although deleveraging by firms Financial Systems
seems to be proceeding somewhat more slowly
There is more evidence of procyclical leverage by commercial banks in arm’s-length
and from a stronger initial position. Finally, financial systems.
although bank assets remained robust during
the second half of 2007, partly reflecting the 0.4
United Kingdom
reintermediation of off-balance-sheet commit-
ments, the ratio of credit to GDP declined sig-
nificantly in the first quarter of 2008, suggesting Belgium
Netherlands 0.3
Denmark
that the pace of deleveraging may have picked United States
Degree of procyclical leverage

Canada
(from median regressions)

up (see also Chapter 1). Austria
0.2
The current crisis is different for the United
France
States in important ways from previous epi- Australia
Germany
Norway
sodes. Corporate balance sheets and firms’ Spain
Finland 0.1
reliance on external financing were on a more Sweden
solid footing entering the current crisis, which
Italy
should provide some resilience. However, the 0.0
sheer size of the U.S. mortgage market, which Japan
is at the heart of the crisis, and the role of resi-
dential investment suggest that household sav- -0.1
0.2 0.3 0.4 0.5 0.6 0.7 0.8
ing and consumption behavior may play a much Arm’s-length index
larger role in the current downturn than in the
past. On a positive note, the policy stance in the Source: IMF staff calculations.
United States has been proactive, as exemplified

by the aggressive cuts in policy rates and the
measures taken to shore up liquidity in both
commercial banks and investment banks. More-
over, banks have raised substantial amounts of
capital, although continuing declines in the
mark-to-market value of assets suggest that
substantially more capital will be needed before
the financial system can resume significant
discretionary lending.
23However, a strong drop in residential real estate

investment is a distinguishing feature of almost all U.S.
recessions (see Leamer, 2007).
147
In the euro area, the adjustment of house

prices and credit has thus far been milder than
in the United States, but there is evidence that
the adjustment is gathering momentum. The
net borrowing pattern of firms in the euro area
is similar to that in the United States: starting
from a stronger base than typical for a finan-
Figure 4.14. Arm’s-Length Financial Systems, GDP cial-stress-driven recession but weakening. How-
Growth, and Bank Leverage ever, euro area households are in a considerably
(Median; start of financial stress episode at t = 0; years on the x-axis)1
stronger position, and this is a distinguishing
Evidence of more procyclical leverage in arm’s-length financial systems may explain
feature of financial stress episodes that are not
differences in spillovers from financial crises to the real economy. followed by recessions. The euro area’s vulner-
ability may also be somewhat reduced because
Countries with above-median arm’s-length financial systems the financial systems in many countries tend to
Countries with below-median arm’s-length financial systems
be less arm’s-length than in the United States.
Financial Stress Followed by Recessions: Output Within the euro area, there are important
5
(real GDP percent change from one year earlier) intercountry differences. Credit growth is on a
4 more pronounced downward path in Ireland
Difference between medians
statistically significant at time t = 1 and Spain than in other countries. A number
3
of countries have experienced unusually large
2
run-ups in house prices and residential invest-
1 ment (see Box 1.2 and Chapter 2), and activ-
0
ity in this sector is slowing markedly. Finally,
although Germany is experiencing large exter-
-1
-3 -2 -1 0 1 2 3 nal surpluses, there are large current account
deficits in a number of countries, including
Financial Stress Followed by Recessions: Bank Leverage 2 28 Greece, Portugal, Spain, and, to a lesser extent,
(percent)
26 Ireland (see Box 2.1).
24 Overall, these results suggest that the eco-
22 nomic impact of financial stress may be greater
20 in the United States than in the euro area.
18
The U.S. economic downturn may well become
16
more severe and could evolve into a recession.
14
The evidence for the euro area is more con-
12
sistent with the pattern for a slowdown than a
-3 -2 -1 0 1 2 3 recession, and the dynamics also appear to be
evolving with some lag.
Sources: Haver Analytics; IMF, International Financial Statistics database; OECD,
Analytic Database; OECD, Economic Outlook (2008); and IMF staff calculations.
1 The sample is constant for all years.
2 Difference in medians is statistically significant for all values of t. Bank leverage is
defined as assets divided by equity.
Conclusions
This chapter uses the Financial Stress
Index to analyze episodes of stress in bank-
ing, securities, and foreign exchange markets
in 17 advanced economies during the past
30 years. The focus is on attempting to iden-
tify factors that determine the extent to which
financial stress affects economic activity.
148
Conclusions
Table 4.4. Six Major Periods of Financial Stress and Economic Contractions
Initial Condition
Asset price buildup1 Net lending ratio3
Equity prices House prices Credit/GDP Bank assets2 Households Firms
Finland, early 1990s 80.0 36.1 16.6 21.0 –6.5 –5.1
Sweden, early 1990s 68.5 17.5 19.1 27.2 ... ...
Norway, early 1990s 73.9 26.5 18.8 27.6 –6.9 –3.6
Japan, 1990s 54.4 12.2 7.4 22.4 5.3 –5.1
United Kingdom, early 1990s 19.9 22.9 2.5 16.1 –2.3 –3.4
United States, early 1990s 14.5 4.9 3.1 9.5 –0.4 –0.3
Average 51.9 20.0 11.3 20.6 –2.1 –3.5
Current episode
United States 27.7 5.9 3.0 9.9 1.4 –0.7
Euro area 44.0 2.9 4.5 9.7 –0.5 –0.4
Japan4 25.1 5.4 6.4 –0.8 0.1 –2.9
United Kingdom 29.4 3.2 5.1 11.2 –0.8 –0.8
Outcomes
Asset price decline
Equity House Credit/ Bank Macroeconomic deleveraging7 Quarters to
prices prices GDP assets6 Households Firms Output loss8 recovery
Finland, early 1990s –85.9 –39.8 –16.8 –5.1 16.2 17.0 –13.6 27
Sweden, early 1990s –69.5 –20.1 –21.3 –4.9 ... ... –5.8 19
Norway, early 1990s –76.9 –24.6 –2.7 –12.5 16.5 8.5 –3.9 12
Japan, 1990s –58.5 –11.1 –6.8 –8.5 0.5 15.4 –5.1 19
United Kingdom, early 1990s –21.4 –23.3 –5.6 –6.5 9.6 4.4 –2.6 13
United States, early 1990s –21.0 –4.8 –3.8 –5.4 0.8 0.6 –1.3 5
Average –55.5 –20.6 –9.5 –7.2 8.7 9.2 –5.4 15.8
1Trough-to-peak changes before the start of the crisis in the detrended Hodrick-Prescott (HP) filter level of the variables.
2Maximum percent deviation from detrended (HP filter) levels of bank assets before the start of the crisis.
3Deviation from HP trend the year before the crisis.
4Data in net lending ratios are for 2006.
5Peak-to-trough changes after the start of the crisis in the detrended (HP filter) level of the variables.
6Minimum percent deviation from the detrended (HP filter) level of bank assets after the start of the crisis.
7Trough-to-peak changes in detrended (HP filter) net lending ratios.
8Output loss is measured as the loss from peak to trough in percent of peak level of output.
The analysis finds that financial stress is as opposed to traditional relationship-based

often, but not always, a precursor to economic intermediation, tend to exhibit higher procy-
slowdown or recession. A rapid expansion of clical leverage, indicating that the amplifying
credit, a run-up in house prices, and heavy role of financial systems in propagating shocks
borrowing by households and firms all increase is more pronounced. As a result, when shocks
the likelihood that stress in the financial affect core financial institutions, the subsequent
system will lead to a more severe economic downturns tend to be deeper in more-arm’s-
downturn. Banking stress, in particular, tends length financial systems. Even so, arm’s-length
to lead to greater effects on activity, despite systems are not generally more prone to such
financial innovation that has increased the shocks, and such systems are better able to
role of securities markets in many countries. reallocate resources across various sectors of
This can be explained by the procyclicality of the economy in response to changing economic
leverage, especially for investment banks but opportunities.
also for commercial banks in many countries. Based on a comparison of the current epi-
Indeed, economies with financial systems sode of financial stress with previous episodes,
dominated by more-arm’s-length transactions, there remains a substantial likelihood of a
149
Figure 4.15. The Current Financial Stress Episode in the United States and Euro Area in Historical Context 1
(Start of financial stress episode at t = 0; x-axis as stated)
Credit ratios and asset prices are mirroring previous episodes, but bank asset growth remains resilient in both the United States and the euro area. Initial corporate
financial positions have been stronger than in past episodes, but are deteriorating. A correction in household financial positions is under way in the United States. The
slowdown of consumption and investment growth mirrors earlier episodes in the United States, although it has only recently started in the euro area.
Median across financial stress episodes for six case studies

Median across financial stress episodes followed by recessions
Recent Episodes: United States Euro area
House Prices Household Net Output Consumption

(percent deviation from trend; Lending/Borrowing Ratio (GDP percent change from one (percent change from one year
10 quarters on the x-axis) (deviation from trend; years on 4 year earlier; quarters on the 5 earlier; quarters on the x-axis) 5
the x-axis) x-axis)
3 4 4
5 2
3
3
1
2
0 0 2
1
-1
1
0
-5 -2
-1 0
-3
-10 -4 -2 -1
-12 -8 -4 0 4 8 12 -3 -2 -1 0 1 2 3 -12 -8 -4 0 4 8 12 -12 -8 -4 0 4 8 12
Credit Nonfinancial Corporate Residential Investment Nonresidential Investment

(percent of GDP; percent Net Lending/Borrowing Ratio (percent change from one year (percent change from one year
8 deviation from trend; quarters (deviation from trend; years on 3 earlier; quarters on the x-axis) 10 earlier; quarters on the x-axis) 10
on the x-axis) the x-axis)
6 2
5
0
4 1
0
2 0 -10
-5
0 -1
-20
-2 -10
-2
-4 -3 -30 -15
-12 -8 -4 0 4 8 12 -3 -2 -1 0 1 2 3 -12 -8 -4 0 4 8 12 -12 -8 -4 0 4 8 12
Stock Prices Bank Assets Current Account Real Interest Rate

20 (percent deviation from trend; (percent deviation from trend; 20 (percent of GDP; quarters on 2 (percent; quarters on the x-axis) 12
quarters on the x-axis) years on the x-axis) the x-axis)
15
15 10
0
10
10 8
5 -2
0 5 6
-5 -4
0 4
-10
-6
-5 2
-15
-20 -10 -8 0
-12 -8 -4 0 4 8 12 -3 -2 -1 0 1 2 3 -12 -8 -4 0 4 8 12 -12 -8 -4 0 4 8 12
Sources: European Central Bank; Haver Analytics; IMF, International Financial Statistics database; OECD, Analytic Database; OECD, Economic Outlook (2008); and IMF staff calculations.
1All in real terms, except fo r household and nonfinancial corporate net lending ratios and bank assets.
150
Conclusions
Box 4.1. Policies to Resolve Financial System Stress and Restore Sound Financial
Intermediation
How can policymakers respond to financial clear later that delaying decisive intervention
stress, including the current global financial increased the stress on the financial system
turmoil, in a way that ensures that the financial and the economy. To avoid this, policymakers
system is restored to health, while limiting the should force the early recognition of losses
fallout on the economy and avoiding long- and take steps to ensure that financial insti-
term moral hazard? Well-timed interventions tutions are adequately capitalized.
aimed at financial institutions and borrowers • The adverse impact of financial system dis-
can help restore balance sheets and incentives, tress on the real economy may need to allevi-
mitigate the negative shock to the economy ated through measures that directly support
of a financial system under stress, and help to firms and households—for example, through
restart productive investment. But in undertak- targeted debt relief programs to distressed
ing these interventions, governments face the borrowers and corporate restructuring
key challenge of restoring financial intermedia- programs.
tion while keeping the costs to taxpayers down, • Steps should be taken to limit the costs
avoiding misallocations of capital, and main- and moral hazard implications of these
taining proper incentives. policy responses. Shareholders must first
absorb losses by a write-down of their equity
General Principles of Intervention capital. In the case of large losses, creditors
The experience from past episodes of finan- also need to contribute by reducing and
cial system distress suggests that the effective- restructuring their claims. Borrowers must
ness and cost of policy responses depend on absorb some of the costs, especially if they
four key elements: have been imprudent. Mechanisms that link
• Having a sound framework for ensuring government support (such as preferred stock
financial sector stability helps prevent and purchases) to privately raised capital can also
contain financial stress. Key elements of this help identify those banks that are truly worth
framework include (1) pre-crisis sanctions on saving and limit future distortions arising
undercapitalized financial institutions that from moral hazard.
pose systemic risks; (2) legal and institutional
mechanisms to deal swiftly with weak finan- Specific Policy Responses
cial institutions, such as bank-specific bank- Policymakers can employ a wide range of
ruptcy regimes; (3) well-defined tools and specific emergency measures (aimed at contain-
processes for closing and rapidly reopening ing the crisis) and restructuring tools (aimed at
banks; and (4) an effective deposit insurance restoring the normal functioning of the credit
system. system and rebuilding banks’ and borrowers’
• Speed is of the essence to minimize the balance sheets).
impact on the real economy. Too often, Emergency measures include (1) regulatory
regulatory forbearance and liquidity support capital forbearance, (2) emergency liquid-
have been used to help insolvent financial ity support, (3) government deposit guaran-
institutions recover—only to have it become tees, and (4) suspension of convertibility of
deposits. Each of these actions can have very
The author of this box is Luc Laeven. This box

draws heavily on Calomiris, Klingebiel, and Laeven Examining a sample of 40 banking crisis epi-
(2005). sodes, Laeven and Valencia (forthcoming) show that

For an overview of existing literature on crisis emergency measures have often included emergency
resolution policies, see Hoelscher and Quintyn (2003) liquidity support and government deposit guarantees.
and Honohan and Laeven (2005). Regulatory capital forbearance—suspending pruden-
151
Box 4.1 (concluded)
different consequences on the supply of credit is small relative to the negative net worth of
and thus on economic activity. The appropriate recipient institutions. Looking at the recapi-
policy measure depends on whether the trigger talization schemes adopted in the United
for the crisis is a loss of depositor confidence, States (starting in 1933) and Japan (1998 and
the (regulatory) recognition of existing insol- 1999) helps illustrate some key issues. In the
vency, or the knock-on effects of asset price United States, the program mainly involved the
volatility, including exchange rate or house purchase of preferred stock to enhance bank
price pressures. Even during the emergency capital and included appropriate screening and
phase, however, longer-term implications must incentives for participants so that only banks
be taken into account—the risk being that worth saving and those that managed their risk
restoring stability in the heat of the crisis may and capital structure more prudently received
lead central banks to extend loans to some taxpayer funds. Moreover, banks receiving
financial institutions that are almost certain to assistance were monitored to ensure that they
prove insolvent. made proper use of public aid. In Japan, the
Specific resolution policies include (1) recapi- first program (launched in 1998) involved only
talizating financial institutions, (2) using asset small amounts, was mostly targeted to purchases
management companies (AMCs) to resolve of subordinated debt and loans, and was broadly
distressed loans, (3) offering debt forgiveness, spread across the banking system. A more suc-
and (4) providing incentives for loan loss write- cessful recapitalization program was launched
offs. Countries typically apply a combination of in 1999, which involved much larger purchases
resolution strategies—with some directed more of preferred stocks, included more rigorous
toward financial institutions and others geared benchmarks, and participation was more nar-
more toward borrowers—and in the process the rowly focused.
government often incurs substantial fiscal costs. Asset management companies (AMCs): The main
Here are some experiences with these types of objective of government-owned AMCs is to
resolution policies. accelerate financial restructuring by taking
Recapitalization: Measures aimed at quickly over nonperforming assets from banks. Two
improving the capital bases of financial institu- examples of successful AMCs are Securum and
tions do not directly improve debtor capacity, Retrieva in Sweden, created in 1992 to manage
but they make it easier for banks to recognize the problem loans of two major Swedish banks,
losses and thereby facilitate corporate restruc- Nordbanken and Gota Bank. Both compa-
turing. Government-assisted recapitalizations nies managed to recover much of their initial
can, however, create moral hazard for share- investment by selling off their assets. Factors
holders, especially if government intervention that contributed to their success include an
efficient judicial system, which allowed them
tial regulations and allowing technically insolvent

banks to continue operating—is also a rather common The two Japanese programs together involved
response. By contrast, measures aimed at avoiding public purchase of ¥10 trillion (2 percent of GDP) of
bank runs through deposit freezes and bank holidays bank capital.
are rarely used. The specific form of bank recapitalization often
Laeven and Valencia (forthcoming) show that depends on the country’s insolvency regime for
bank recapitalization occurred in three-quarters of the financial institutions. In many countries today such
crises they considered, with an average fiscal cost of regimes do not allow for a speedy resolution of
6 percent of GDP. AMCs were set up in slightly more crises but rather prolong them. Another lesson for
than half of the episodes in their database. successful bank recapitalization is that bank capital
The average fiscal cost of government intervention regulations must be enforced rigorously, which can
in the cases studied by Laeven and Valencia (forth- involve imposing limitations on the distribution of
coming) is about 16 percent of GDP. dividends.
152
Conclusions
to force insolvent debtors into bankruptcy; distress. Because of the risks of moral hazard,
the real-estate-related nature of their assets, however, debt forgiveness should be considered
which made it easier to restructure; and the only as a last resort.
strong governance mechanisms and skilled Loan loss write-off programs: Loan loss write-off
management teams in place at the companies. programs are directed at supporting borrowers.
However, other countries have found it harder Although they can be implemented quickly,
to realize these advantages, in part owing to loan loss write-offs may worsen incentives for
weak legal, regulatory, and political institu- prudent behavior as they do not impose losses
tions—banks’ assets often are transferred to on banks or their borrowers.
the AMC at prices abovemarket value, resulting Overall, the mix of policy responses will
in backdoor bank recapitalization and creating ultimately be crisis-specific and must reflect
moral hazard. a variety of factors, including the nature and
Debt forgiveness: Key advantages of this depth of the financial crisis and the specific
measure are its simplicity and speed—debt country circumstances. The four principles for
forgiveness recognizes loan losses up front intervention outlined here have proven to have
and thus provides immediate relief to borrow- general applicability and should be followed in
ers. At the same time, however, debt forgive- every crisis, including the current one.
ness poses incentive problems because it does
not impose losses on borrowers and bank The U.S. experience in the 1930s, when gold pay-
shareholders. It can also undermine trust in ment clauses in debt contracts were abrogated, shows
monetary institutions and the rule of law, as that debt forgiveness can help solve coordination
it can violate monetary standards and inter- problems in renegotiating debt. While few individual
fere in private contracting. Whether it works creditors were willing to voluntarily remove these
clauses, when they were forced to do so collectively,
ultimately depends on the frequency of its use the improvement in aggregate economic circum-
and the specific circumstances of financial stances left both creditors and debtors better off.
sharp downturn in the United States, given One factor that helps predict whether a
the similarities between the current dynam- financial stress episode will lead to a downturn
ics of asset prices, credit ratios, and household is the buildup in balance sheet vulnerabilities
financial positions and previous episodes that associated with rising asset prices and credit.
were followed by recession. Mitigating factors Policymakers therefore need to be alert to these
are the rapid monetary policy response and indicators during the upswing of the financial
a relatively low real interest rate. In the euro cycle. Prudential measures and monetary policy
area, by contrast, the relatively strong position should be used to address buildups that may
of households offers some protection against a leave the economy vulnerable to greater output
sharp downturn, despite the appreciable run-up losses in the event of a severe shock.
in asset prices and the credit ratio ahead of Should significant financial stress affect the
the current financial turmoil. The euro area’s core of the banking system, the early recogni-
vulnerability to a deeper downturn may also be tion of losses and measures to support the
somewhat reduced because many of its financial speedy restoration of capital can help reduce
systems are less arm’s-length, as evidenced most the output consequences. At the same time,
notably by the much smaller role for the origi- policymakers must seek to avoid longer-term
nate-to-distribute mortgage banking business moral hazard implications of any strategy to
model. restore financial stability.
153
Appendix 4.1. Data and Methodology The Financial Stress Index

The main authors of this appendix are Angela Espiritu and This section of the appendix describes the
Gavin Asdorian.
components and the methodology used to con-
struct the Financial Stress Index (FSI). The FSI
Table 4.5. Data is an equal-variance weighted average of seven
Variable Source Frequency variables, grouped into three categories.
GDP Haver Analytics, Quarterly
Organization for
Economic Cooperation Banking Sector
and Development
(OECD) Analytical • Banking sector β: rolling 12-month covari-
Database
ance of the year-over-year percent change of a
Average petroleum spot World Economic Quarterly country’s banking sector equity index and its
price Outlook database
overall stock market index, divided by the roll-
CPI inflation Haver Analytics, OECD Quarterly
Analytical Database ing 12-month variance of the year-over-year
Real private consumption OECD Analytical Quarterly percent change of the overall stock market
Database index. Sources: Thomson Datastream, Haver
Money supply OECD Analytical Quarterly Analytics, and OECD.
Database
• TED spread: three-month LIBOR or commer-
Interest rate Haver Analytics, Quarterly
International Financial cial paper rate minus the government short-
Statistics (IFS) term rate. Source: Haver Analytics.
database
• Inverted term spread: government short-
Banking sector equity Thomson Datastream, Monthly
index Haver Analytics term rate minus government long-term rate.
Stock market index OECD Monthly
Sources: Thomson Datastream and Haver
3-month London Haver Analytics Monthly
Analytics.
interbank offered rate
(LIBOR) or commercial
paper rate Securities Market
Government short-term Haver Analytics Monthly
rate • Corporate spread: corporate bond yield minus
Government bond yields Haver Analytics, Monthly long-term government bond yield. Sources:
Thomson Datastream Thomson Datastream and Haver Analytics.
Corporate bond yields Thomson Datastream, Monthly • Stock decline: stock index at t – 1 minus stock
Haver Analytics
index at t, then divided by stock index at t – 1.
Real effective exchange International Monetary Monthly
rate Fund Source: OECD.
Residential investment OECD Analytical Quarterly • Time-varying stock volatility: GARCH(1,1) vol-
Database atility of overall stock market index monthly
Nonresidential investment OECD Analytical Quarterly return. Source: OECD.
Database
Current account OECD Analytical Quarterly
Database
Foreign Exchange
Real house prices OECD Quarterly
• Time-varying real effective exchange rate
Credit IFS database Quarterly
volatility: GARCH(1,1) volatility of real effec-
Household net lending OECD Annual
tive exchange rate monthly percent change.
Government net lending OECD Annual
Source: IMF.
Nonfinancial corporate net OECD Annual
lending All components are originally in monthly
Bank assets OECD Annual frequency. The index is constructed by taking
Bank equity OECD Annual the average of the components after adjusting
154
Appendix 4.1. Data and Methodology
for the sample mean and standardizing by the • Real cost of debt: corporate bond yield minus
sample standard deviation. The index is then one-year-forward Consensus inflation forecast.
rebased so that it ranges from 0 to 100. Finally, Sources: Thomson Datastream, Haver Analyt-
it is converted into quarterly frequency by tak- ics, and Consensus Economics.
ing the average of the monthly data. The FSI is • Real cost of equity: derived using a model
available for 17 advanced economies starting in specified in Box 2 on p. 76 of the ECB’s
1980.24 November 2004 Monthly Bulletin. Using avail-
Episodes of financial stress are identified able data for the other variables, the real cost
when the index is one standard deviation above of equity, ht, can be calculated using the fol-
its trend. Episodes that are only two quarters lowing equation:
apart are considered a single episode. To clas-
sify the cause of an episode of financial stress— Dt[(1 + g) + 8(gtIBES – g)]
Pt = ———————————,
either banking-related, securities-related, or ht – g
foreign-exchange-related—we look at the
where
change between the FSI from the quarter prior
Pt = real stock price,
to the start of the episode and the maximum
Dt = the current level of real dividends,
value of the FSI within the episode. If most of
gtIBES = I/B/E/S long-term earnings-per-share
the increase stems from banking sector com-
growth forecast minus Consensus long-
ponents, the FSI is classified as “banking.” The
term inflation forecast,
same rule applies if the change results mainly
g = long-term growth rate of real corpo-
from the securities markets components or
rate earnings, assumed constant at
the foreign exchange component. Moreover, if
2.5 percent.
banking contributes at least one-third of the
The overall cost of capital is calculated as a
change in the FSI, the episode is also classified
weighted average of these three components
as “banking-related.”
with the weights defined, respectively, as loans,
debt, and equity as shares of nonfinancial corpo-
The Cost of Capital rate liabilities as reported in the OECD national
accounts data.
“Cost of capital” is defined in this chapter
as a weighted average of the real cost of bank
loans, the real cost of debt, and the real cost Bankscope Data
of equity, using as weights the relative shares
Two data sets were constructed using bank-
of equity, bonds, and loans in nonfinancial
level data obtained from the Bankscope
corporate liabilities. The cost of capital is based
database.25 The first data set included only
on the calculation outlined in Box 4 on p. 37
investment banks as classified by the Bankscope
of the European Central Bank’s (ECB’s) March
database (“Investment Bank/Securities House”).
2005 Monthly Bulletin. The real cost of bank
The second data set, referenced in the chapter
loans, real cost of debt, and real cost of equity
as “commercial banks,” included banks with the
are derived as follows:
following Bankscope classifications: Commercial
• Real cost of bank loans: bank lending rates
Bank, Savings Bank, Cooperative Bank, Real
minus one-year-forward Consensus inflation
Estate/Mortgage Bank, and Medium & Long
forecast. Sources: IFS, ECB, and Consensus
Term Credit Bank.
Economics.
24Data on long-term corporate bond yields for Greece,
Ireland, New Zealand, and Portugal were not available 25Bankscope database published by Bureau van Dijk
and therefore were excluded from the sample. Electronic Publishing (BvDEP): www.bvdep.com.
155
Sample of Banks Table 4.6. Average Yearly Share of Total Bank

Assets of Banks in Sample
Investment banks
Number of Top Banks Percent of Country’s
The sample of banks contained banks that Country a Year in Sample Total Bank Assets
were among the top 50 investment banks glob- Australia 10 78
ally in terms of total assets in one or more years Austria 10 71
Belgium 10 94
from 1988 to 2007. Canada 10 88
Denmark 10 92
Commercial banks Finland 10 79
France 10 73
The sample of banks consisted of banks that Germany 10 65
were among the top 10 banks26 in terms of Italy 10 40
total assets for each country in one or more Japan 30 74
Netherlands 10 90
years from 1988 to 2007. Also included were Norway 10 78
any banks that were acquired by or that merged Spain 10 78
Sweden 10 94
with a top-10 bank. (See below for an explana- Switzerland 10 64
tion of accounting for mergers and acquisition.) United Kingdom 30 67
The number of commercial banks in each United States 50 60
country used in the sample was chosen to pro-

vide a representative sample of banking activ-
addition, the data were cleaned (by country,
ity within each country. Table 4.6 summarizes
in the case of commercial banks) by excluding
the average yearly share of total bank assets
observations in which the growth rate of total
(as reported by the OECD) represented by the
assets was above the 95th percentile or below
banks in the sample.
the 5th percentile.
Consolidated Versus Unconsolidated

Balance Sheets Mergers and Acquisitions
Investment banks For consistency, banks that were acquired by
or merged with banks included in the original
Data from consolidated statements were used
sample set were also included in the data set.
for investment banks. If consolidated data were
For years prior to a merger or acquisition, the
unavailable, data from unconsolidated state-
banks involved were treated as separate banks;
ments were used.
for years subsequent to a merger or acquisition,
Commercial banks the bank resulting from the merger or acquisi-
tion was naturally a single bank in the database.
In order to isolate as much as possible the
In order to calculate level changes or growth
domestic activities of commercial banks, uncon-
rates of a bank in the year of a merger or acquisi-
solidated bank data were used for commercial
tion, a data point was constructed for the year
banks in the sample. If unconsolidated state-
prior to the merger or acquisition by summing
ments were unavailable, data from consolidated
the data values of the banks involved in the
data were used.
merger or acquisition.
Data from multiple statements for the same
bank were combined to form a single set of
bank-level data if the statement types (consoli- References
dated or unconsolidated) were the same. In
Adrian, Tobias, and Hyun Song Shin, 2008a, “Finan-
cial Intermediary Leverage and Value at Risk,” Staff
26Top 30 banks for United Kingdom and Japan; top Report No. 338 (New York: Federal Reserve Bank
50 banks for United States. of New York).
156
References
——— , 2008b, “Liquidity and Leverage,” Staff Report Hoelscher, David S., and Marc Quintyn, 2003, Manag-
No. 328 (New York: Federal Reserve Bank of New ing Systemic Banking Crises (Washington: Interna-
York). tional Monetary Fund).
Altunbas, Yener, Leonardo Gambacorta, and David Honohan, Patrick, and Luc Laeven, eds., 2005, Finan-
Marqués, 2007, “Securitisation and the Bank Lend- cial Crises: Containment and Resolution (Cambridge,
ing Channel,” ECB Working Paper No. 838 (Frank- Massachusetts: Cambridge University Press).
furt am Main: European Central Bank). Illing, Mark, and Ying Liu, 2006, “Measuring Financial
Bayoumi, Tamim, and Ola Melander, 2008, “Credit Stress in a Developed Country: An Application to
Matters: Empirical Evidence on U.S. Macro-Finan- Canada,” Journal of Financial Stability, Vol. 2 (Octo-
cial Linkages,” IMF Working Paper 08/169 (Wash- ber), pp. 243–65.
ington: International Monetary Fund). International Monetary Fund (IMF), 2008, Global
Bernanke, Ben, and Mark Gertler, 1995, “Inside the Financial Stability Report: Financial Stress and Dele-
Black Box: The Credit Channel of Monetary Policy veraging—Macro-Financial Implications and Policy
Transmission,” Journal of Economic Perspectives, Vol. 9 (Washington, October).
(Autumn), pp. 27–48. J.P. Morgan Research, 2008, “Credit and Growth: The
———, and Simon Gilchrist, 1999, “The Financial Case of the Euro Area” (June 19).
Accelerator in a Quantitative Business Cycle Frame- Kaminsky, Graciela, and Carmen Reinhart, 1999, “The
work,” Handbook of Macroeconomics, Vol. 1C (Amster- Twin Crises: The Causes of Banking and Balance-of-
dam: North-Holland), pp. 1341–93. Payments Problems,” American Economic Review, Vol.
Bernanke, Ben, Cara Lown, and Benjamin Friedman, 89 (June), pp. 473–500.
1991, “The Credit Crunch,” Brookings Papers on Eco- Kashyap, Anil, Raghuram Rajan, and Jeremy Stein, 2008,
nomic Activity: 2, Brookings Institution, pp. 205–47. “Rethinking Capital Regulation,” paper prepared for
Borio, Claudio, 2007, “Change and Constancy in the the Federal Reserve Bank of Kansas City symposium,
Financial System: Implications for Financial Distress “Maintaining Stability in a Changing Financial Sys-
and Policy,” BIS Working Paper No. 237 (Basel: tem,” Jackson Hole, Wyoming, August 21–23.
Bank for International Settlements, October). Kashyap, Anil, and Jeremy Stein, 1995, “The Impact of
Calomiris, Charles W., Daniela Klingebiel, and Luc Monetary Policy on Bank Balance Sheets,” Carnegie-
Laeven, 2005, “Financial Crisis Policies and Resolu- Rochester Conference Series on Public Policy, Vol. 42
tion Mechanisms: A Taxonomy from Cross-Country (June), pp. 151–95.
Experience,” in Systemic Financial Crises: Containment Kindleberger, Charles, and Robert Aliber, 2005,
and Resolution, ed. by Patrick Honohan and Luc Manias, Panics, and Crashes: A History of Financial
Laeven (Cambridge, Massachusetts: Cambridge Crises (Hoboken, New Jersey: John Wiley & Sons,
University Press), pp. 25–75. 5th ed.).
Caprio, Gerard, and Daniela Klingebiel, 2003, Kiyotaki, Nobuhiro, and John Moore, 1997, “Credit
“Episodes of Systematic and Borderline Financial Cycles,” Journal of Political Economy, Vol. 105 (April),
Crises,” Dataset 1 (Washington: World Bank). pp. 211–48.
Cihak, Martin, and Petya Koeva Brooks, forthcoming, Laeven, Luc, and Fabian Valencia, forthcoming, “Sys-
“From Subprime Loans to Subprime Growth? Evi- temic Banking Crises: A New Database” Working
dence for the Euro Area,” Working Paper (Wash- Paper (Washington: International Monetary Fund).
ington: International Monetary Fund). Leamer, Edward, 2007, “Housing Is the Business
Claessens, Stijn, M. Ayhan Kose, and Marco E. Ter- Cycle,” presented at the Jackson Hole Symposium
rones, 2008, “What Happens During Recessions, on Housing, Housing Finance, and Monetary
Crunches and Busts?” (unpublished; Washington: Policy, August 30–September 1, 2007.
International Monetary Fund, August 5). Minsky, Hyman, 1992, “The Financial Instability
Goodhart, Charles, 1996, “An Incentive Structure Hypothesis,” Jerome Levy Economics Institute
for Financial Regulation,” LSE Financial Markets Working Paper No. 74 (Annandale-on-Hudson,
Group Special Paper No. 88 (London: London New York: Bard College, May).
School of Economics and Political Science, July). Peek, Joe, and Eric Rosengren, 1995, “The Capital
Harding, Don, and Adrian Pagan, 2002, “Dissecting Crunch: Neither a Borrower Nor a Lender Be,”
the Cycle: A Methodological Investigation,” Journal Journal of Money, Credit and Banking, Vol. 27, No. 3,
of Monetary Economics, Vol. 49 (March), pp. 365–81. pp. 625–38.
157
Rajan, Raghuram, and Luigi Zingales, 2003, Comparison,” NBER Working Paper No. 13761
“Banks and Markets: The Changing Character (Cambridge, Massachusetts: National Bureau of
of European Finance,” CEPR Discussion Paper Economic Research).
No. 3865 (London: Centre for Economic Policy Shiller, Robert J., 1999, “Measuring Bubble Expec-
Research). tations and Investor Confidence,” NBER Work-
Reinhart, Carmen, and Kenneth Rogoff, 2004, “The ing Paper No. 7008 (Cambridge, Massachusetts:
Modern History of Exchange Rate Arrangements: National Bureau of Economic Research).
A Reinterpretation,” Quarterly Journal of Economics, Shin, Hyun Song, 2008, “Reflections on Modern Bank
Vol. 119, No. 1 (February), pp. 1–48. Runs: A Case Study of Northern Rock,” Working
———, 2008, “Is the 2007 U.S. Sub-Prime Financial Paper (unpublished; Princeton, New Jersey: Prince
Crisis So Different? An International Historical ton University).
158

c4pdf PDF

Uploaded by

Copyright:

Available Formats

c4pdf PDF

Uploaded by

Document Information

Original Title

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

c4pdf PDF

Uploaded by

Copyright:

Available Formats

4

Financial Stress and Economic Downturns

Against this background, this chapter

on the empirical relationship between asset prices—such

Table 4.1. Descriptive Statistics on Financial Stress

60 Source: IMF staff calculations.

Denmark, Finland, France, Germany, Italy, Japan, Netherlands, Norway,

100 Japan Germany 80

Overall, of the 113 episodes of financial stress identi-

tries over the past three decades, including the 20

with well-known events such as stock/bond mar- 20

with the monetary policy stance as reflected in

Source: IMF staff calculations.

system and the economy, and although the financial

Table 4.2. Descriptive Statistics on Financial Stress, Slowdowns, and Recessions

mines whether a financial stress episode will be

14The cost of capital is defined here as a weighted aver-

causality between recessions and financial stress

Having examined the nature of the shocks, the

(in deviation from trend) is significantly

15Slowdown severity is measured using the cumulative

output loss during the period during which output is

interacted with the duration of the financial

Has Financial Innovation Affected the 0.4

Interplay between Financial Stress and 0.2

Economic Cycles? 0.0

Banks continue to be important in explain- -0.2

their repurchase facilities.18 Because this is typical

on retail deposits and whose main activity is 0.0

tionship-based (and bank-dominated) activities

more vulnerable to banking stress.20 Of note, 6

deeper in economies with more-arm’s-length -20 -6

IMF (2008), showing that fair value accounting tends to

Table 4.3. Cross-Section Regressions

The Current Financial Crisis in Historical

speaking, the episodes characterized by the 0 -1

Anglophone countries, asset price imbalances

22Thisapproach is in line with other studies, includ-

tries with the largest buildups of financial imbal-

on publicly financed bailouts.

Implications for the Current Crisis in the United

financial stress episodes that were followed by

Degree of procyclical leverage

(from median regressions)

United States has been proactive, as exemplified

23However, a strong drop in residential real estate

In the euro area, the adjustment of house

The analysis finds that financial stress is as opposed to traditional relationship-based

Median across financial stress episodes for six case studies

House Prices Household Net Output Consumption

Credit Nonfinancial Corporate Residential Investment Nonresidential Investment

Stock Prices Bank Assets Current Account Real Interest Rate

The author of this box is Luc Laeven. This box

(2005). sodes, Laeven and Valencia (forthcoming) show that

Box 4.1 (concluded)

tial regulations and allowing technically insolvent

Appendix 4.1. Data and Methodology The Financial Stress Index

24Data on long-term corporate bond yields for Greece,

Sample of Banks Table 4.6. Average Yearly Share of Total Bank