Journal of Monetary Economics 32 (1993) 485-512.

North-Holland

The role of macroeconomic factors

in growth

Stanley Fischer*
M.I. T., Cambridge, MA 02139, USA

Received March 1993, final version received September 1993

Using a regression analog of growth accounting, I present cross-sectional and panel regressions
showing that growth is negatively associated with inflation, large budget deficits, and distorted
foreign exchange markets. Supplementary evidence suggests that the causation runs from macroeco-
nomic policy to growth. The framework makes it possible to identify the channels of these effects:
inflation reduces growth by reducing investment and productivity growth; budget deficits also
reduce both capital accumulation and productivity growth. Examination of exceptional cases shows
that while low inflation and small deficits are not necessary for high growth even over long periods,
high inflation is not consistent with sustained growth.

Key words: Growth; Inflation; Budget deficits; Productivity; Investment

JEf. c.Jassijication: EOO; 011; 057

1. Introduction

It is now widely accepted that a stable macroeconomic framework is neces-

sary though not sufficient for sustainable economic growth.’ In this paper
I present international cross-sectional regression evidence that supports the
view that growth is negatively associated with inflation and positively associated
with good fiscal performance and undistorted foreign exchange markets. I also
present evidence suggesting that the causation runs in part from good macro-
economic policy to growth.

Correspondence fo: Stanley Fischer, MIT., E52-373, Cambridge, MA 02139, USA.

* Also affiliated with the NBER. This paper is part of the World Bank’s Growth Project. I am
grateful to Michael Bruno, Jose de Gregorio, Robert King, Ken Rogoff, and Sweder van Wijnbergen
for helpful suggestions and comments, to participants in a Hebrew University seminar, especially
Michael Beenstock and Giora Hanoch, for suggestions, and to Ruth Judson for excellent research
assistance.
‘See World Bank (1989, 1990, 1992).

0304-3932/93/$06.00 0 1993-Elsevier Science Publishers B.V. All rights reserved

486 S. Fischer, Macroeconomic factors in growth

The view that a stable macroeconomic framework is conducive to growth is

also supported by much striking nonregression evidence. In Latin America, the
recovery of economic growth in Chile and Mexico was preceded by the restora-
tion of budget discipline and the reduction of inflation.2 By contrast, the
ongoing growth crisis in Brazil coincides with high inflation punctuated by
stabilization attempts and continued macroeconomic instability. The fast grow-
ing countries of East Asia have generally maintained single- or low double-digit
inflation, have for the most part avoided balance of payments crises, and when
they have had them - as for instance in Korea in 1980 - moved swiftly to deal
with them. The lessons of the case study evidence amassed in the major World
Bank research project, headed by Little, Cooper, Corden, and Rajapatirana
(1992),3 support the conventional view. The notion that macroeconomic stability
is not sufficient for growth is supported by evidence from Africa, where most of the
countries of the franc zone have grown slowly since 1980 despite low inflation.
This paper considerably extends and strengthens results presented in Fischer
(1991). In that paper, I used the conventional approach of adding macroeconomic
variables to the basic growth regression. In this paper, in sections 4 through 6,
I develop an alternative approach due to Elias (1992), a regression analog of
growth accounting. I present both pure cross-sectional regressions as well as panel
regressions, which exploit the time series as well as cross-sectional variation in the
data. I also explore nonlinearities in the relationship between inflation and
growth.4 In section 7 I discuss the issue of the causality between inflation and
economic growth. Then in section 8 I identify and discuss some apparent excep-
tions, countries where high growth took place despite high inflation and/or large
deficits, and conclude that the statement that macroeconomic stability is neces-
sary for sustainable growth is too strong, but that the statement that macroeco-
nomic stability is conducive to sustained growth remains accurate.
The paper opens in section 2 with a discussion of the notion of a stable
macroeconomic framework and of the theoretical considerations linking growth
to macroeconomic policies. In section 3 I briefly review recent evidence on the link
between macroeconomic conditions and growth, most of it based on the standard
mixed regression which includes among its regressors the rate of investment.

2. Definitions and theoretical considerations

In practice the concept of a stable macroeconomic framework is used to
mean a macroeconomic policy environment that is conducive to growth. The

’ However, in both cases it took several years to reduce inflation to the moderate, 15-30 percent,
range.
3Summarized in Corden (1991).
4Nonlinearities in the inflation-growth relationship have also been explored by Levine and
Zervos (1992).
 S. Fischer, Macroeconomic factors in growth 487

macroeconomic framework can be described as stable when inflation is low and

predictable, real interest rates are appropriate, fiscal policy is stable and sustain-
able, the real exchange rate is competitive and predictable, and the balance of
payments situation is perceived as viable. ’ This definition goes beyond the
stability of macroeconomic policy variables to include also the criterion that
policy-related variables are at levels conducive to growth.
Of the five criteria specified in the preceding definition, only low and stable
inflation is readily quantifiable.(j None of the specified variables is directly
controllable by policy, and each should optimally vary in response to shocks.
Given the practical difficulty of defining and measuring the stability of the
macroeconomic framework, or the optimal or appropriate inflation rate, real
interest rate, real exchange rate, and so forth, I instead proceed by specifying
indicators of macroeconomic policy.
The basic indicators of macroeconomic policy are the inflation rate, the
budget surplus or deficit, and the black market exchange premium. I shall use
the inflation rate as the best single indicator of the conduciveness of macroeco-
nomic policies to growth’ and the budget surplus as the second basic indicator.
In essence, the inflation rate serves as an indicator of the overall ability of the
government to manage the economy. Since there are no good arguments for
very high rates of inflation, a government that is producing high inflation is
a government that has lost control. All governments announce that they aim for
low inflation, and the macroeconomic situation in any medium or high inflation
economy can therefore be expected to change. While there are economies in
which inflation remains at moderate levels for prolonged periods [Dornbusch
and Fischer (1993)], economic agents in a high or medium inflation economy
have to expect an attack - typically many attacks - on inflation at some point.
Countries may for a long time succeed in maintaining low and stable inflation
through policies that are not ultimately sustainable. Such countries, for instance
those in the franc zone, may face fiscal or balance of payments crises that could
necessitate sharp changes in macroeconomic policy and that certainly increase
macroeconomic uncertainty. The fiscal deficit is a good, though imperfect,
indicator of such an unsustainable situation. In addition, as discussed below, the
deficit is likely to affect growth through its effects on capital accumulation.

5This definition is based on World Bank (199Oa, p. 4).

6 With regard to quantification of the other four variables: Measures of the fiscal deficit provide
some information about fiscal policy; however it is difficult to characterize fiscal policy by a single
variable [MacKenzie (1989)], and international fiscal data are poor. Estimates of sustainable deficits
could in principle be calculated along the lines of Hamilton and Flavin (1986), but that level of detail
would require a much more extensive study than can be carried out in the current project. The
competitiveness of the real exchange rate could in principle be estimated by its implications for
current and future levels of the current account, while the appropriateness of the real interest rate is
difficult to specify.
‘The potential links between inflation and growth are discussed and developed in Fischer (1983)
and by implication in Fischer and Modigliani (1978), and are taken up below.

J.Mon- E
488 S. Fischer, Macroeconomic factors in growth

I use the black market premium on foreign exchange as an indicator of the

sustainability and appropriateness of the exchange rate. The black market
premium is a good indicator of a distorted or controlled market for foreign
exchange, but is less good as an indicator of the unsustainability of the exchange
rate, since an exchange rate may be overvalued and unsustainable even when
there is no black market premium.
Most developing countries experienced major terms of trade shocks during
the period over which the regressions in this paper are estimated. The terms of
trade are included as a separate exogenous determinant of macroeconomic
performance.’
The usual emphasis on the stability of the macroeconomic framework (rather
than its conduciveness to growth) suggests that the main reason macroeconomic
factors matter for growth is through uncertainty. There are two main channels
through which uncertainty could affect growth. First, policy-induced macroeco-
nomic uncertainty reduces the efficiency of the price mechanism, as in the classic
Lucas (1973) contribution. This uncertainty, associated with high inflation or
instability of the budget or current account, can be expected to reduce the level
of productivity and, in contexts where the reallocation of factors is part of the
growth process, also the rate of increase of productivity. Second, temporary
uncertainty about the macroeconomy tends to reduce the rate of investment, as
potential investors wait for the resolution of the uncertainty before committing
themselves [Pindyck and Solimano (1993)]. This channel suggests that invest-
ment would be lower at times when uncertainty is high, and its presence should
therefore be more noticeable in the time series than cross-sectional data.g
Capital flight, which is likely to increase with domestic instability, provides
another mechanism through which macroeconomic uncertainty reduces invest-
ment in the domestic economy.
The variability of inflation might serve as a more direct indicator of the
uncertainty of the macroeconomic environment. However, the inflation rate and
the variance of the inflation rate are highly correlated in the cross-section, making
it difficult to disentangle the effects on growth of the level of inflation from the
effects of uncertainty about inflation. By adding a time series measure of inflation
variability to the panel regressions, I attempt in this paper to bring further
evidence to bear on the level-uncertainty distinction, but with limited succes~.~~
The 1950s and 1960s growth theory literature on inflation and growth
emphasized the positive impact of inflation on capital accumulation that occurs

x Data sources are described in the appendix.

9 Solimano (1989) presents time series evidence supporting this relationship.
“‘Aizenman and Marion (1991) attempt to quantify policy uncertainty by estimating autoregres-
sive processes for policy variables and using the standard deviations of policy surprises as a measure
of uncertainty. This is a promising approach, which however does not distinguish contemporaneous
variability caused by responses to exogenous shocks from purely random variability.
 S. Fischer, Macroeconomicfacrors in growth 489

as a result of the portfolio shift away from money when the rate of return on
money falls, the Mundell-Tobin effect. Subsequent contributions, noting vari-
ous complementarities between real balances and capital - whether through the
production function or because of a cash-in-advance constraint - predicted that
higher inflation would reduce capital accumulation.” Similarly, all the costs of
inflation detailed in Fischer and Modigliani (1978) - including the impact of
inflation on the taxation of capital - would imply a negative association between
the level of income and inflation and, through the new growth theory mecha-
nisms, between inflation and growth. It is also possible that the relationship
between inflation and growth is nonlinear.
Turning to the other macroeconomic indicators: The budget surplus should
be positively associated with capital accumulation. There are again two reasons.
The first is crowding out. The second is that, like the inflation rate, the
deficit serves as an indicator of a government that is losing control of its
actions.
An increase in the black market exchange premium is an indicator of expecta-
tions of depreciation of the exchange rate and foreign exchange rationing. This
suggests that capital accumulation and the black market premium are likely to
be negatively related. One influence in the opposite direction arises from the fact
that when foreign exchange access is controlled, there is frequently preferential
treatment for the import of investment goods.
Of course, each of these indicators has its shortcomings as a policy measure.
In the short run, neither the inflation rate nor the budget deficit is unaffected by
the growth rate. For instance, a supply shock will both reduce the growth rate
and raise the inflation rate; and given government spending, a reduction in
growth will increase the deficit. Two main types of regressions are reported in
this paper. In the cross-sectional regressions, the period average (usually
1961-88) growth rate or other dependent variable for each country is regressed
on period average values of such right-hand-side variables as inflation and the
budget deficit. In the panel regressions, similar regressions are run using both
the time series variation within each country and the cross-sectional variation.
The problem of reverse causation is more likely to arise in the panel regressions.
In principle, the use of instrumental variables can deal with the endogeneity
problem, but in practice appropriate instruments are difficult to find. The
endogeneity problem is less severe in the cross-sectional regressions, where the
length of period is more than 25 years. Over such long periods, the average rates
of inflation and the deficit are more likely to be determined by the government’s
basic policy stance than by the short-run association between shocks and the
endogenous policy indicators. In addition, I use prior knowledge, the timing of
the 1973 oil shock, to break the period down into one in which demand shocks
predominated (pre-1973) and one in which there were many supply shocks, and

I1 For references to the literature through 1983, see Fischer (1983).

490 S. Fischer, Macroeconomic factors in growth

show that the results based on the pre-1973 data also support the basic
contention of this paper.

3. Existing empirical evidence

Beyond the evidence of the examples presented in the introduction, the simple
statistical evidence supports the basic proposition that macroeconomic stability
is conducive to growth. Inflation in fast-growing Asia is well below the rates of
price increase in slower-growing Africa and Latin America (table 1), and across
the three periods shown in table 1 inflation in each area has moved inversely
with growth. l2 Levine and Renelt (1992) show that high growth countries are
also lower inflation countries, have smaller governments, and lower black
market exchange rate premia - the latter reflecting disequilibria in the official
foreign exchange markets.
The large volume of empirical work inspired by the new growth theory
consists largely of cross-country regressions, typically using the Summers-
Heston (1988) ICP data. i3 Levine and Renelt (1992) list 40 cross-sectional
growth studies published between 1980 and 1990.14 Their paper starts from
a basic regression in which per capita real income growth (GYP, from the World
Bank data base) is regressed on Summers-Heston initial real income (RGDP60),
population growth (GN), the 1960 rate of secondary school enrollment (SEC),
and the share of investment in GDP (INI’). The regression is estimated on
a sample of 101 countries, over the period 1960-89 (t-statistics in parentheses):

GYP = - 0.83 - 0.35 RGDP60 - 0.38 GN

( - 0.98) ( - 2.50) ( - 1.73)

+ 3.17 SEC + 17.5 ZNI’, (1)

(2.46) (6.53)

l?= = 0.46.

“The World Bank SAVEM tables from which table 1 is derived present more regional detail than
does table 1. For both South Asia and East Asia, growth and inflation change in the same direction
between 1965-73 and 1973-80. For the Middle East and North Africa, growth and inflation exhibit
the same general correlation as is seen in table 1, that is, they move in opposite directions from
period to period. [I should also note that a table similar to table 1 is presented in Fischer (1991). The
inflation rate for Asia in that table (for which the first period is 1960-73) is shown as increasing from
period to period, with an average of only 2 percent for 1960-73. Both tables are taken from the same
source, and I am unable to account for the different patterns of Asian inflation, though they may
arise from changes in country coverage and data revisions or possibly a transcription error.]
i3 For examples, see Barro (1991) and the many studies listed in Levine and Renelt (1992).
“‘Their list is necessarily incomplete; in particular, it does not include the comparative cross-
country analysis by Adelman and Morris (1988), which is based on work dating back to the 1960s.
Several other earlier cross-country studies are listed by Chenery [chapter 2 in Chenery, Robinson,
and Syrquin (1986, p. 27)]. Reynolds (1986, p. 101) also presents a cross-sectional growth regression,
despite his general preference for time series studies.
 S. Fischer, Macroeconomic~factors in growth 491

Growth is robustly (in the Learner sense) related to initial income and to
investment, but not to the other variables.
When Levine and Renelt extend the analysis to include a variety of other
variables, they find, first, that several measures of economic policy are related to
long-run growth and, second, that the relationship between growth and almost
every particular macroeconomic indicator other than the investment ratio is
fragile. The strongest results are that investment in physical capital, and either
the level or the rate of change of human capital, increase the rate of growth.
In Fischer (1991) I extended the basic Levine-Renelt growth equation to
include macroeconomic indicators. Per capita growth is negatively associated
with inflation and positively associated with the budget surplus as a share of
GNP. While the coefficients on inflation and the budget surplus are statistically
significant, the negative coefficient on external debt is not, in a sample that
includes all countries for which data were availab1e.15
As discussed in section 2, these macroeconomic indicators cannot be regarded
as truly exogenous. Instruments are difficult to find; for instance, such candi-
dates as measures of political instability not only cause but also are caused by
inflation.r6 Given the difficulties of choosing instruments, I do not pursue
instrumental variables regressions in the remainder of this paper, but address
the issue of endogeneity in section 7.
The negative relationship between inflation and economic growth has been
found also in other papers [for instance, in Fischer (1983) de Gregorio (1993)
and Gylfason (1991)]. To deal with the endogeneity of inflation, Cukierman
et al. (1992) use measures of central bank independence as an instrument for
inflation. They conclude that, even after instrumenting with the better indicators
of central bank independence, there remains a significant negative relationship
between inflation and economic growth. De Long and Summers (1992) likewise
implicitly use the degree of central bank independence as an instrument for
inflation and argue that lower inflation is associated with higher growth.
Levine and Zervos (1992) returning to the questions examined by Levine and
Renelt, show that an inflation variable has a significant coefficient when added
to the basic growth equation, but that the relationship is not robust and can be
traced to several high inflation countries. They also examine possible nonlineari-
ties in the relationship between inflation and growth. Their final innovation is to
create an index of macroeconomic policy, a function of the rate of inflation and
the budget deficit, and to show that growth is positively associated with better

I5 It can be argued that developing countries are sufficiently and systematically different from
industrialized countries that the latter should be excluded from the regressions. While it is easy to
agree with this view at the extremes, it is hard to know where to draw the line, and I therefore
worked mostly with all countries for which there were data. For some regressions (not reported
here), I excluded all countries that in 1970 had an income level above Italy’s; if anything, this gave
stronger results with respect to macroeconomic variables, particularly the debt.
I6 Results obtained using different sets of instruments are presented in Fischer (1991).
492 S. Fischer, Macroeconomic factors in growth

(low inflation, larger budget surplus) macroeconomic policy indicators. Easterly

and Rebel0 (1992) find a consistent negative relationship between growth and
budget deficits.
The simple correlations suggested by table 1, and the more detailed empirical
work that builds on eq. (l), thus support the view that a stable macroeconomic
framework is conducive to growth.

4. Interpreting the evidence

The basic growth regression (1) includes the investment rate as a regressor.
The effects of macroeconomic policy variables are usually studied by adding
them as right-hand-side variables to the basic regression. The resultant regres-
sion therefore presents severe difficulties of interpretation when used to examine
the role of policy variables or other indicators in the growth process. Presum-
ably the interpretation of such equations is that, conditional on the rate of
investment, other variables affect growth. But it is hard to conceive of variables
that would not affect growth through their effect on investment as well as
through other routes, mostly the rate of productivity increase - and this is
especially true of macroeconomic variables.
Recognizing this, Barro (1991) also presents investment equations, as does
Fischer (1991). Nonetheless, since some of the same variables explain both
growth and investment, the policy variable-augmented growth regression has no
straightforward interpretation. Rather there seem to be mongrel regressions,
born out of a legitimate study of convergence and the desire to study the effects
of policy on growth.”
In this section I use a simple alternative to the mixed regression, a production
function-based approach pioneered by Elias (1992). The approach is a regres-
sion analog of growth accounting, which helps identify the channels through
which macroeconomic variables affect economic growth. As a matter of ac-
counting, growth can be attributed to increases in supplies of factors and
to a residual productivity category, reflecting changes in the efficiency with
which factors are used. The approach is to examine the relationships between
growth and macroeconomic variables, and then between the macroeconomic
variables and changes in both the supplies of factors, and the residual, or
productivity.
Consider the production function

I7 Some of the more recent papers, [for instance, Cukierman et al. (1992) and Levine and Zervos
(1992)] do not include investment in the equation that also includes inflation, but do include other
conditioning variables such as initial real income.
 Table 1
Inflation and economic growth (% per annum).a

Africa Asia Latin America

65-13 73-80 8G-90 65-73 73-80 80-90 65-73 73-80 80-90

GDP growth 3.7 3.4 2.1 5.8 5.8 6.9 6.0 5.0 1.1
GDP per cap. growth 1.1 0.4 - 1.0 3.2 3.7 4.9 3.3 2.5 - 0.9
Inflation 5.2 15.8 18.9 14.8 8.9 6.9 22 53 249
-
“Source: World Bank.
494 S. Fischer, Macroeconomic .&actors in growth

where K, L, and H are physical capital, raw labor, and human capital, respec-
tively, and A, is an overall efficiency factor, including not only the level of
technology, but also for example representing the quality of government man-
agement of the economy or institutional factors.
Differentiating (2), we obtain the conventional growth accounting equa-
tion

e/y=MGK) + v&L) + rl3(fufo+ v&m, (3)

where t/i is the elasticity with respect to argument i in eq. (2). The product
q4(k/A) will be referred to as the productivity residual.
Macroeconomic factors can in principle affect economic growth through all
four factors on the right-hand side of eq. (3). The standard procedure of adding
macroeconomic variables to a growth regression that already includes some of
the right-hand-side variables thus implicitly assumes that policy variable does
not affect the other included variables, and affects growth only through its
impact on the right-hand-side variables in (3) not explicitly included in the
regression, typically the productivity residual.

Productivity residuals

Three alternate estimates of productivity residuals were made. Bhalla resid-

uals start from an estimated panel regression equation like (3) with the three
factor inputs included explicitly. The data are those provided by Surjit Bhalla
through the Bank’s 1991 World Development Report (WDR) database. The
Bhalla panel regression implies productivity residuals for each country for each
year; the mean productivity residual for each country, plus the dummy for its
region, is an estimate of the average rate of productivity increase for that
country, on the (maintained) assumption that the production function for each
country is the same up to the productivity variable.‘*

r* The Bhalla production function estimated on the full panel by GLS (t-statistics in parentheses)
is

ZGDP = 0.398 ZKAP + 0.440 ZLAB + 0.012 ZED + RD,, N = 1912. (PI)
(14.25) (3.53) (0.38)

ZGDP is the growth rate of real GDP (in 1980 prices); ZKAP is the growth rate of capital; ZLAB is
the growth rate of the labor force; and ZED is the growth rate of the educational stock in the labor
force (calculated as the product of the average years of education of the adult population and the
labor force). Regional dummies (RDJ are included for the five World Bank regions as of 1991 and the
OECD. Coefficients are: EMENA (Europe, Middle East, and North Africa), 0.011; LACAR (Latin
America and Caribbean), 0.002; AFRIC, - 0.004, EASIA, 0.006, SASlA, 0.001; and OECD, 0.007.
These coefficients are small in absolute value and only those on EMENA and OECD are signifi-
cantly different from zero.
 S. Fischer, Macroeconomic factors in growth 495

Two other sets of residuals were calculated for each country. Solow residuals
are calculated as

RESi, = ZGDPi, - 0.4 ZKAPi, - 0.6 ZLABif, (4)

i= 1,. . . , 68, t = 1961,. . . , 1988.

Mankiw-Romer- Weil residuals are calculated as

REMR Wi, = ZGDPi, - 0.333 ZKAPi, - 0.333 ZLAB,

- 0.333 ZEDit, (5)

i= 1,. . , 68, t = 1961,. . . ,1988.

Calculation of the Solow residuals imposes a common Cobb-Douglas produc-

tion function in which the share of capital is somewhat higher than in the
industrialized countries, as it generally is estimated to be in developing coun-
tries. Mankiw-Romer-Weil residuals are calculated imposing coefficients used
in their 1992 paper.
The productivity residuals constructed by these three methods are very highly
correlated in the time series for each country (with pairwise R2s all exceeding
0.98) and we therefore use the Solow residuals in the remainder of the paper.
Table 2 presents the minima and maxima of the mean rates of Solow
productivity growth calculated for each of the five 1991 World Bank regions and
the OECD. These estimates raise obvious questions about the underlying
Summers and Heston data, or perhaps the input data. When similar calculations
were made using World Bank income data, the productivity residuals looked
more plausible. For instance, Pakistan had the highest rate of productivity
growth in South Asia and Congo had the highest in Africa. However, since the
Summers-Heston income data are widely used, I chose to work with those,
leaving the investigation of the apparent anomalies in table 2 for later research.
The difference between the maximum (Brazil) and minimum (Haiti) rates of
productivity increase is very large, 6.7 percent per annum. Even the range across
regions - 2.19 percent - is large.

5. Results in the growth accounting framework

Cross-sectional regressions for the largest possible number of countries on
single macroeconomic indicator variables are presented in table 3.19 These are

i9 Differences in data coverage raise the issue of whether all regressions should be run on the
maximal possible common set of countries or on as many countries as possible for the particular
regression. Since the intersection of the data sets covers only 32 countires, I have chosen the latter
approach. I have also excluded any data series that includes less than 10 observations.
496 S. Fischer, Macroeconomic factors in growth
 S. Fischer, Macroeconomic factors in growth 497

Table 3
Cross-sectional growth regressions (r-statistics in parentheses).”

No. of
Eq. INFLAT SVRRAT ZTOTI EXCHPREM SMAPI obs.

(6) - 0.037 80
(- 2.13)

(7) 0.133 40
(2.07)

(8) 0.113 80
(0.83)

(9) - 0.022 94
( - 2.95)

(10) - 0.093 80
( - 2.98)

(11) - 0.026 0.277 - 0.040 - 0.041 22

( - 1.34) (3.36) ( - 0.20) ( - 3.32)

a Dependent variable is ZGDP, growth rate of real GDP. Other variable definitions are: INFLAT,
inflation rate; SVRRAT, ratio of budget surplus to GDP; ZTOTI, change in terms of trade;
EXCHPREM, black market exchange premium; SMAPI, mean of the standard deviation of the
inflation rate around its mean for overlapping seven-year periods. (Variable definitions are in the
appendix.)

regressions in which there are no regional dummies, and only a constant in

addition to the variable indicated. However, the coefficients change very little
when regional dummies are added. The inflation rate, budget surplus, black
market exchange premium, and the standard deviation of inflation are each
individually significantly correlated with the growth rate.*’
Regression (11) is included for completeness, though there is only a small
number of countries for which the full set of data is available.21 The coefficients
on the budget surplus and the black market exchange premium are strongly
significant.**

“ln Fischer (1992), in a similar table, only the inflation rate and the budget surplus were
significantly correlated with the growth rate. The change is a result of the increase in sample sizes
since that paper was written. I have also substituted the moving average measure of inflation for the
standard deviation of the inflation rate over the entire period (SINFLAT) in eq. (IO) for comparabil-
ity with the panel regressions. The coefficient on SINFLAT in the analog of eq. (10) is - 0.026, with
a t-statistic of - 2.34.
‘I They are: Ghana, Cote d’lvoire, Kenya, Malawi, Morocco, Zambia, Dominican Republic,
Jamaica, Mexico, Argentina, Chile, Colombia, Ecuador, Paraguay, Venezuela, India, Indonesia,
Korea, Pakistan, Thailand, Greece, Turkey.
“AS noted above, the high correlation between the inflation rate and its standard deviation
preclude the inclusion of both variables in the regressions.
498 S. Fischer, Macroeconomic factors in growth

Table 4
Panel growth regressions (r-statistics in parentheses).”

No. of
Eq. INFLAT SURRAT ZTOTl EXCHPREM SMAPI obs.

(12) - 0.046 1998

( - 7.43)

(13) 0.226 714

(6.30)

(14) 0.057 1732

(5.93)

(15) - 0.026 2088

( - 1.48)
(16) - 0.064 1685
( - 4.54)

(17) - 0.039 0.228 0.043 - 0.017 351

( - 4.65) (4.49) (2.71) ( - 2.76)

“Variables are as defined in table 3. Regressions are run using GLS (seemingly unrelated
regressions).

This first cross-sectional look at correlations between growth and macroeco-

nomic variables is broadly consistent with prior expectations. However, in using
only period averages the cross-sectional regressions discard the information in
the time series for individual countries. The results of similar panel regressions
are presented in table 4.
The simple panel regressions in table 4 [eqs. (12) to (1611 confirm the
relationships between inflation and inflation variabilityz3 and growth, and also
between the budget surplus and growth, seen in table 3. In the time series, the
black market exchange premium correlation with growth is lower than in the
cross-section, while the correlation between changes in terms of trade and
growth is increased; improvements in the terms of trade are associated with
higher growth. The numerical value of the coefficient on inflation in eq. (12) is
a bit higher than that in eq. (6), while the coefficient on the standard deviation of
inflation falls between the cross-section and the time series. The coefficient on
the budget surplus in eq. (13) is almost double that in eq. (7), possibly a result of
reverse causation between growth and the budget within the time series for
individual countries.

23 Values of SMAPI in this sample range from 1.8 (South Africa) to 44.5 (Bolivia). The regression
implies that the high inflation variability in Bolivia would reduce its growth 2.7 percentage points
relative to South Africa.
 S. Fischer, Macroeconomic factors in growth 499

Regression (17) includes all the regressors except inflation uncertainty. All the
coefficients are significantly different from zero. They imply that a country that
has an inflation rate 100 percentage points higher than another (e.g., 110 percent
per annum rather than 10 percent per annum) will have a growth rate that is 3.9
percent lower, and that a country with a budget surplus that is higher by
1 percent of GDP will have a growth rate that is 0.23 percent larger. Countries
with higher black market exchange premia grow more slowly. The unitsz4
imply that the black market premium in the country where it was largest,
Mozambique, would be associated with a reduction in the growth rate of 2.5
percent. Adverse changes in the terms of trade reduce growth, though the
coefficient is small relative to the range of the change in the terms of trade.
Similar regressions that include regional dummies give almost identical coeffi-
cients on the macroeconomic variables.
The regressions reported in table 4 reinforce the evidence in favor of the view
that macroeconomic stability, as measured by the (inverse of the ) inflation rate,
and indicators of macroeconomic policy, like the budget surplus and the black
market exchange premium, are associated with higher growth and are on
average good for growth. We turn now to the mechanisms through which the
macroeconomic variables affect growth.

5.1. Capital accumulation

Pursuing the approach described in section 4, we start with equations in

which the rate of capital accumulation is regressed on the same macroeconomic
variables as in tables 3 and 4. The results presented in table 5 are all for panel
regressions estimated by GLS. (Results for the corresponding cross-section
regressions will be discussed below.) In the simple regressions (18) through (22)
all the coefficients are significantly different from zero, and all have the expected
sign.
In regression (23) the coefficients on the inflation rate and the black market
exchange premium are significantly different from zero, while surprisingly the
budget surplus and the terms of trade coefficients lose their significance. The
coefficient on inflation implies that an increase in the inflation rate by 100
percentage points (e.g., from 10 to 110 percent per annum) reduces the growth
rate of the capital stock by 3.7 percentage points. This is a large effect: if the
investment rate is about 20 percent of GDP and the capital output ratio is 2.5,
then the growth rate of capital is 8 percent. According to the regression, capital
in such a country would stop growing when the inflation rate reaches about 210
percent per annum. The point estimate of the coefficient on the budget surplus
implies that an increase in the budget deficit of 1 percent of GDP would reduce
the growth rate of capital by 0.08 percentage points. Again assuming a capital

24The black market exchange premium enters the equation in the form ln(1 + EXCHPREM).
500 S. Fischer, Macroeconomic factors in growth

Table 5
Panel regressions, capital accumulation (r-statistics in parentheses).”

No. of
Eq. INFLAT SURRAT ZTOTI EXCHPREM SMAPI obs.

(18) - 0.046 1626

( - 11.05)

(19) 0.222 716

(7.11)

0.028 1300
(3.54)

(21) - 0.027 1653

( - 12.01)
(22) - 0.094 1340
( - 9.46)
(23) - 0.037 0.075 0.008 - 0.019 352
( - 4.77) (1.61) (0.62) ( - 3.56)

a Dependent variable is ZKAP, the growth rate of the real capital stock. Variable definitions are as
in table 3. Regressions are estimated by GLS.

output ratio of 2.5, the investment share in GNP would decline by 0.2 percent-
age points. This estimate implies a relatively low level of crowding out on
average. The effect implied in the one-variable regression (19) is above 0.5
percentage points. The coefficient on the black market premium again suggests
that it has quite large effects on investment and capital accumulation.
In single-variable cross-sectional regressions corresponding to those in table
5, the coefficients on all variables except the terms of trade are significantly
different from zero, and all are of the same sign as in table 5. However, the
coefficients are generally larger than in table 5. In the overall cross-sectional
regression, corresponding to eq. (23), the coefficient on the inflation rate is
insignificant, while that on the deficit becomes larger (0.50) and strongly signifi-
cant.
These results suggest that one important route through which inflation affects
growth is by reducing capital accumulation;25 similarly, an increase in the black
market exchange premium, which reflects foreign exchange controls and expec-
tations of devaluation, reduces capital accumulation. An increase in the budget
surplus is associated with more capital accumulation, but the effect is not
significant even at the 10 percent level. The numerical values of the coefficients
are plausible, even though these cannot be thought of as structural regressions.

” De Gregorio (1993) also finds strong effects of inflation on investment.

 S. Fischer, Macroeconomic factors in growth 501

Table 6
Panel regressions, productivity growth (r-statistics in parentheses).

No. of
Eq. INFLAT SURRAT ZTOTI EXCHPREM SMAPI obs.

(24) - 0.016 1598

( - 2.88)

(25) 0.125 714

(4.57)

(26) 0.039 1251

(3.85)

(27) - 0.014 1566

( - 4.46)

(28) - 0.022 1327

( - 1.89)
(29) - 0.018 0.137 0.038 - 0.006 351
( - 2.49) (3.23) (2.60) (- 1.17)

a Dependent variable is RES, the Solow residual, calculated as in eq. (4). Other variable definitions
are as in table 3. Regressions are estimated by GLS.

5.2. Productivity growth

The impacts of the macroeconomic variables on productivity growth estim-

ated by the Solow residual are presented in table 6. The inflation rate is
significantly negatively correlated with the rate of productivity growth, with
a coefficient which implies that an increase in the inflation rate by 100 percent is
associated with a decline in the rate of productivity growth of 1.8 percent per
annum. Increases in the budget surplus and improvements in the terms of trade
are associated with improvements in productivity growth. The effect of inflation
is robust to the inclusion of other variables. The black market exchange rate
premium is significantly negatively correlated with the rate of productivity
growth, but the coefficient on the black market premium loses its significance in
the multiple regression.
Theories in which inflation distorts price signals suggest that uncertainty
about inflation should have an impact on productivity. The negative coefficient
on the standard deviation of inflation (SMAPZ) in eq. (29) is consistent with this
view, but the coefficient is not statistically significant.
In the cross-sectional regressions equivalent to (24) to (28), none of the
coefficients in any of the single-variable regressions were significantly different
from zero. This implies that the significant correlations in table 6 are mainly
a result of the time series variation between the regressors and productivity
growth. In the overall regression equivalent to (29), the coefficients on inflation
502 S. Fischer. Macroeconomic factors in growth

Table 7
Panel regressions, labor force growth (t-statistics in parentheses).”
_______
No. of
Eq. INFLAT SURRAT ZTOTI EXCHPREM SMAPI obs.

(30) - 0.001 2021

(1.75)

(31) - 0.0015 716

( - 0.14)

(32) - 0.0007 1668

( - 0.64)
(33) 0.0009 2020
(2.60)

(34) 0.003 1669

(2.57)

(35) - 0.002 - 0.007 0.0009 0.0003 352

(- 1.14) ( - 0.53) (0.29) (0.22)

a Dependent variable is ZLAB, the growth rate of the labor force. Regressions are estimated by
GLS.

and the budget surplus were similar to those in (29) but again not statistically
significant.

5.3. Labor force growth

For the sake of completeness, table 7 presents estimates of the panel equations
for labor force growth. It would be surprising if the macroeconomic variables
had a major impact on the growth of the labor force. In fact, the regressions in
table 7 show no coefficients to be significantly different from zero in the overall
regression (35), and only the correlations with the exchange premium and
inflation variablity to be significant in the one-independent-variable regressions.

5.4. Summary

The strongest result that comes out of the regressions reported in tables
5 through 8 is the consistent negative correlation between inflation and growth.
Inflation is negatively associated with both capital accumulation and productiv-
ity growth. There is a strong positive correlation between the budget surplus and
growth, with the evidence suggesting some influence of the surplus on capital
accumulation and a stronger effect on the rate of growth of productivity.
Adverse changes in the terms of trade reduce growth, mainly through their effect
on productivity growth. The black market exchange premium is negatively
 S. Fischer, Macroeconomic factors in grow& 503

related to growth, mainly through lower capita1 accumulation. The macroeco-

nomic variables are not significantly associated with labor force growth.

6. Inflation nonlinea~ti~ and other variations

While it is easy to believe that triple-digit inflation has adverse effects on

economic growt through the mechanisms discussed in section 2, and reflected in
the regressions for capital accumulation and productivity growth, it is possible
that there is a range of low inflation rates in which variations in inflation have
very little effect on growth. Thus, in testing for nonlinear effects of infiation,
I expect to find more signi~cant effects of inflation at high than at low inflation
rates.
To allow for possible nonlinearities in the effects of inflation, the basic
regressions for growth, capital accumulation, and productivity were estimated
using a spline function, with breaks at 15 and 40 percent.26 In table 8, the
inflation variables enter as:

INFLL = value of the inflation rate if it is 15 percent or less,

INFLM = value of the inflation rate if it is between 15 and 40 percent,
ZNFLH = value of the inflation rate if it is above 40 percent.

Table 8 shows the variants of panel regressions (17) (23), and (29), with the
inflation rate broken into three categories. The results show that the effects of
inflation are nonlinear, but that, per percentage point of inflation, the associ-
ation between inflation and growth and its determinants on average weakens as
inflation rises.” It is thus not the case, as I had expected, that it is the high
inflation outliers that are responsible for the overall negative correlations
between inflation and growth, capital accumulation, and productivity growth,
seen in tables 5 through 7. Rather the association between inflation and growth
and between inflation and capital accumulation is stronger at the low and
moderate inflation levels than at high inflation. When inflation is decomposed as
in table 8, none of the inflation components in eq. (38) the equation for
productivity growth, is significant, even though inflation enters significantly in
the corresponding linear eq. (23).
Note also that, when the inflation rate is decomposed in this way, the
coefficient on the budget surplus in the capital accumulation equation becomes
statistically significantly different from zero. An increase in the budget deficit is
statistically significantly associated in table 8 with lower growth through both
lower capital accumulation and lower productivity growth.

z6 See Greene (1993,pp. 235-238) for spike regressions.

27 Levine and Zervos (1992) obtain similar results.
504 S. Fischer, Macroeconomic factors in growth

Table 8
Nonlinear effects of inflation (f-statistics in parentheses), N = 351:

Es. 1361 Eq. (37) Eq. (38)

Dependent variable

Variable ZGDP ZKAP RES

INFLL - 0.127 - 0.008 - 0.079

(- 1.99) (- 0.15) (- 1.37)
1NFLM - 0.075 - 0.115 - 0.029
( - 1.84) (- 3.23) (- 0.77)
INFLH - 0,019 - 0.017 - 0.009
( - 1.43) (- 1.46) (- 0.76)
SURRAT 0.230 0.115 0.141
(4.66) (2.50) (3.24)
ZTOTI 0.048 0.009 0.04 1
(2.97) (0.70) (2.80)
EXCHPREM - 0.014 - 0.018 - 0.004
(- 2.19) ( - 3.28) ( - 0.69)

p See variable definitions in the appendix. Regressions are estimated by GLS.

The results in table 8 suggest that the basic nonlinearity in the relationship
between inflation and growth could be captured by a function in which
log(1 + x) appears. When regressions like (17), (23), and (29) are run with
log(1 + n) replacing the inflation rate, the t-statistic on the inflation variable
rises in each case, and the remaining coefficients are little affected.
ln~u~io~ uncer~ui~ty: Grier and Tullock (1989) report a significant negative
association between inflation variability and growth and a relationship between
inflation and growth that varies across regions. Tables 3-6 show the simple
relationship between the moving standard deviation of inflation (SMAPZ) and
the dependent variables. In all cases, the direction of the relationship is the same
as that between inRation and the dependent variable.
Both the inflation rate and SMAPI have been included in several regressions,
to try to separate out the effects of high from uncertain inflation. No consistent
pattern of results emerged. In the panel regressions, both with and without the
other variables in the regression, the coefficient on the inflation rate was almost
always negative, and that on the standard deviation measures was sometimes
negative and more often positive, sometimes significantly so.
standard ~aria~~es~ In table 9 I report the results of adding the standard
cross-country variables to regressions (17), (23), and (29). These all enter as
period averages or initial values. Initial real GNP per capita enters the growth
and capital accumulation equations significantly and negatively; a measure of
tariff protection openness, defined as the product of the volume of trade relative
to GNP and the tariff rate, affects productivity growth negatively; and the
 S. Fischer, Macroeconomic factors in growth 505

Table 9
Addition of standard variables (t-statistics in parentheses), N = 206.”

Eq. (39) Eq. (40) Eq. (41)

Dependent variable

Variable ZGDP ZKAP RES

INFLAT - 0.03 1 - 0.032 - 0.017

( - 2.72) ( - 4.21) ( - 1.59)
SURRAT 0.241 0.038 0.146
(3.00) (0.61) (2.04)
ZTOTI 0.066 0.002 0.063
(3.39) (0.13) (3.41)
EXCHPREM - 0.015 - 0.014 - 0.007
(- 1.94) - 2.72) - 1.08)
In(GNP0) - 0.021 - 0.035 - 0.007
(- 2.18) - 2.55) - 0.82)

OPEN TAR - 0.003 - 0.0002 - 0.003

( - 1.27) ( - 0.06) ( - 2.13)

BHKA VG 0.005 0.013 O.ooOl

(1.44) (2.72) (0.03)

LLY - 0.020 - 0.039 - 0.016

( - 0.36) ( - 0.50) ( - 0.35)

“GNP0 is the Summers-Heston 1960 per capita GNP; OPENTAR is a measure of tariff
protection, equal to ((X + M)/ZGDP) In(1 + tar) where X and M are exports and imports and tar is
the WDR measure of tariffs and other surcharges on imports; BHKAVG is the Barro-Lee measure
of human capital; and LL Y [from Levine and Zervos (1992)] is the average ratio of liquid liabilities
to GDP for the period 1960-89.

human capital measure is estimated to increase capital accumulation. The

measure of financial intermediation does not enter any of the equations signifi-
cantly.
The most important result in table 9 is that the addition of these variables
leaves the basic relationships between the dependent and macroeconomic policy
variables unchanged.

7. Causality

While inflation is negatively associated with growth and with its production
function determinants, it is not clear - especially in the panel regressions - which
way the causation runs. If supply shocks predominate, then possibly adverse
supply shocks cause both inflation and slower growth, and the regressions may
merely be reflecting that association.
506 S. Fischer, Macroeconomic factors in growth

Table 10
Inflation-growth correlations, subperiods 1961-72 and 1973388.

Dependent variable

ZGDP ZKAP RES

61-72 73-88 61-72 73-88 61-72 73-88

Single regression

Bq. (42) (43) (44) (45) (46) (47)

INFLAT - 0.072 - 0.033 - 0.052 - 0.026 - 0.032 - 0.013
( - 3.74) - (4.67) - (3.46) ( - 6.30) ( - 1.47) ( - 2.16)
No. of obs. 773 1225 631 995 640 958

Multiple regressions

Bq. (48) (49) (50) (51) (52) (53)

INFLAT - 0.200 - 0.039 - 0.03 1 - 0.029 - 0.173 - 0.019
( - 3.37) - (4.04) - (0.69) ( - 3.40) ( - 3.09) ( - 2.33)

No. of obs. 44 306 44 306 44 306

a See variable definitions in the appendix. Regressions are estimated by GLS

The inclusion of changes in the terms of trade as a regressor goes a long way
towards dealing with this problem. For most of the developing countries,
changes in the terms of trade are a major source of supply shocks, and these have
been taken into account in the multi-variable regressions in sections 5 and 6.
The use of measures of central bank independence as instruments for inflation in
the cross-sectional regressions, as in Cukierman et al. (1992), provides another
method of dealing with the endogeneity of inflation. Their results suggest that
the causation runs significantly, but not exclusively, from inflation to growth.
Subperiod regressions: In addition, I have split the period up into two parts,
from 1960 to 1972 and from 1973 to 1988.‘* Demand shocks probably predomi-
nated in the first period and supply shocks in the second. If supply shocks are
primarily responsible for the negative association between inflation and growth,
we should expect the negative association to be stronger in the second period
than in the first, where we might even expect to find a positive association.
Table 10 shows the results of this breakdown, presenting only the coefficient
on inflation from the multiple regressions corresponding to (17), (23), and (29). In
the simple regressions, (42) to (47), the coefficient on inflation is always negative,
and absolutely larger in the first period than in the second. The t-statistics are
always lower for the first period. Similarly, in the multiple regressions, the

s* Michael Bruno suggested this approach.

 S. Fischer, Macroeconomic factors in growth 507

absolute value of the coefficients is larger in the first period than in the second,
but there are much fewer degrees of freedom and the r-statistics are smaller.
The breakdown into subperiods thus strengthens the view that the relation-
ship between inflation and growth is not merely a result of supply shocks.

8. Some reservations

The results so far support the conclusions that high inflation, large budget
deficits, and exchange market distortions are associated with lower growth.
Most of the results suggest also that these relationships are to some extent
causal. The positive association between the budget surplus and growth appears
particularly robust, and that between the black market exchange premium and
growth is also strong. Thus, the evidence from the regressions and from case
studies is consistent with the view that the causation is not fully from low growth
to high inflation, and therefore that countries that are able to reduce the inflation
rate in a sustainable way can on average expect higher growth to follow. There is
nothing in the results to contradict the view that inflation is merely a symptom of
a government out of control - but there is nothing in that argument that
contradicts the view that controlling inflation will help restore growth.
While the regressions provide suggestive evidence, it is also useful to look at
the exceptions. Table 11 shows that some countries have experienced rapid

Table 11
High inflation and economic growth (% per annum).’

High growth period Entire spellb

GNP GNP
Country Period Inflation growth Period Inflation growth

Argentina 1977 101.5 6.2 1975-87 112 0.5

1979 95.4 6.8
1986 64.5 5.3
Brazil 1980 60.3 8.7 198&87 90 3.5
1984-86 105.3 7.1
Chile 1977 65.2 9.4 1972277 115 - 1.2
Ghana 1978 54.9 9.4 1977-78 66 5.6

Israel 1979-80 70.3 6.0 1979-85 95 3.8

Peru 1979 51.1 5.6 1979 51 5.6
198687 59.8 7.9 1983-87 73 2.4
Uganda 1981 73.6 8.0 1981 74 8.0
1988 104.3 6.3 1985-88 102 0.1

’ Source: Inflation data from IMF; growth data from World Bank.
b A spell is a period in which the annual inflation rate year exceeds 50 percent each year.
508 S. Fischer, Macroeconomic factors in growth

Table 12
Large deficits, inflation, and growth.”
-
Country Period Deficit/GDP Growth rate Inflation
-
Argentina 1975576 13.4 - 0.3 I34
1981-84 13.9 - I.8 I24
Chile 1973 19.0 - 5.7 153
Cote d’Ivoire 1976 12.4 10.9 II
1979-83 12.3 0.7 II
Ghana 1975 13.2 - 14.3 26
Greece 1981 10.9 0.0 22
1984-88 12.7 2.1 I7
Israel* 197484 19.4 3.6 64
Jamaica 1977-85 17.6 - I.2 21
Mauritius* 1978-82 II.6 2.2 I6
Malawi 1979982 13.4 0.1 II
Mexico 1981-82 13.5 3.9 36
Morocco 1976-79 13.8 6.2 9
1981 13.6 - 1.3 12
1983 11.5 2.3 6
Nicaragua* 1981-86 20.8 - 0.4 70
Turkey 1978 10.6 2.8 31
1980 11.9 - 0.7 74
Zambia 1977-87 16.1 - 1.6 20
Zimbabwe 1981-87 13.3 I.6 18

a Source: Deficit data from Easterly, except for countries indicated by asterisks, where the deficit
data are from the IMF. Other variables are from WDR Database.

growth at high inflation rates. During the period 1961-88, at least 14 countries
in the World Bank database experienced an annual inflation rate greater than 50
percent in at least one year. Growth in some of these countries exceeded
5 percent during a year or more of the 50 percent or more inflation. Table 11 lists
those cases, as well as information about growth and inflation during the entire
period of high inflation of which the high growth period is a part.
Similarly, treating the budget deficit as a macroeconomic indicator, the 15
countries in table 12 have experienced deficits in excess of 10 percent of GDP
during the periods shown. 29 Some of them, including Brazil and Israel, are also

29 For countries for which the Easterly fiscal data are available, the data listed in table 12 are from
that source; for other countries for which IMF deficit data are available (indicated by an asterisk),
that is the source.
 S. Fischer, Macroeconomic factors in growth 509

listed in table 10. Others listed in table 12 include rapid growers such as
Morocco during the period 1976-79.30
The data presented in tables 11 and 12 raise the question of the circumstances
under which countries can continue to grow fast when such standard indicators
of the macroeconomic situation as the deficit and inflation are exceptionally
high. Every country that appears in table 11 ran into severe trouble at some later
stage. Thus, table 11 seems to show only that rapid growth is possible for a time
even with high inflation. In some cases, such as Peru, the period of rapid growth
is associated with a rapidly accelerating inflation and a situation that is heading
rapidly for disaster.
By drawing the line in table 11 at 50 percent inflation, I omit those countries
that have succeeded in growing over sustained periods with inflation that
persisted in the moderate range of 15-30 percent, typically with the assistance of
extensive indexation.31 Such situations are sustainable, provided the govern-
ment takes action to prevent inflation rising above the 30 percent range. The
explosive situations appear to be those in which governments believe the
inflation rate is of no major consequence, and permit it to continue rising even
after it leaves the moderate range.
The data in table 12 provide a much less clear lesson. For most of the
countries in the table, growth rates were low during the periods of high deficits,
but Morocco grew fast during the high deficit period, as did Italy in the 1980s. It
is clearly possible to sustain large deficits for some time, with the assistance of
high saving rates and financial repression. Notice though that inflation rates are
low for almost all the non-Latin American high deficit countries. The lesson
seems to be that a high deficit by itself is not a certain indicator of later trouble.
It may be sustainable for a while, and it may be consistent with low inflation. It
would take supplementary studies of the budgetary situation and debt dynamics
to determine whether a large deficit is sustainable, and therefore consistent with
macroeconomic stability, or unsustainable, and therefore a harbinger of macro-
economic instability.

9. Concluding comments
The broad range of evidence reviewed and presented in this paper supports
the conventional view that a stable macroeconomic environment, meaning
a reasonably low rate of inflation and a small budget deficit, is conducive to
sustained economic growth. The growth accounting framework makes it pos-
sible to identify the main channels through which inflation reduces growth. As
a great deal of prior theory predicts, the results presented here imply that

“Industrialized countries such as Italy are not included in the database from which table 12 is
drawn.
31 See Dornbusch and Fischer (1993).
510 S. Fischer, Macroeconomic factors in growth

inflation reduces growth by reducing investment and by reducing the rate of

productivity growth. Larger budget surpluses are also strongly associated with
more rapid growth, through greater capita1 accumulation and greater productiv-
ity growth. An undistorted foreign exchange market is also conducive to growth.
The cross-sectional regression methodology that is associated with the new
growth theory has been extended in this paper to include pane1 regressions,
whose results typically reinforce those of the simple cross-sections. The en-
dogeneity issue is difficult to deal with formally, but the weight of the evidence
implies that the relationship between inflation and growth is not purely a result
of low growth producing high inflation. The evidence that small deficits are
good for growth is strong, as is the support for the view that distorted foreign
exchange markets, as reflected in a large foreign exchange market premium, are
bad for growth.
The examples presented in tables 11 and 12 show that low inflation and small
deficits are not necessary for high growth, over even quite long periods. They do
imply that very high inflation is not consistent with sustained growth. The
results also suggest that the sustainability of the budget deficit has to be
investigated in more detail than is possible in the aggregative approach that has
been taken in this paper.
To make further progress in defining a stable and sustainable macroeconomic
framework, and in clarifying the channels through which macroeconomic vari-
ables affect growth, it will be necessary to undertake more detailed case studies
of individual countries, based on structural models. A good start on this
approach has aheady been made in some of the contributions in Little et al. and
in many studies of individual countries. The conclusions of those studies agree
with the conclusions in this paper.

Appendix: Data sources

All time series that have less than ten observations have been excluded from
regressions.
ZGDP is the log difference of real GDP, as estimated by Heston and Sum-
mers.
ZKAP is the growth of the capita1 stock, using the World Bank (Nehru) data
set. The data start with an assumed capita1 stock of zero in 1950, which leads to
very rapid rates of growth of the capita1 stock in early years. Further, some
estimates are based on an assumed stock of zero in 1960. All observations for
which the capital stock grows by more than 30 percent per annum have been
excluded.
ZLAB is the log difference of the labor force, from the WDR dataset.
ZED is the log difference of the product of LABOR, the size of the labor force,
and BHK, the Barr+Lee (1993) measure of the average years of educational
 S. Fischer, Macroeconomic factors in growth 511

attainment of the labor force. It is an estimate of the growth rate of human

capital.
INFLA T is the inflation rate, computed from the CPI series in International
Financial Statistics. GDP deflator data from the World Bank were used
to extend inflation series for the Central African Republic, Malawi, and
Chad.
SZNFLAT is the standard deviation of the inflation rate over all the observa-
tions on inflation for a given country.
SMAPZ is a time series estimate of inflation uncertainty, calculated as the
standard deviation of the inflation rate around its mean for overlapping seven-
year periods.
EXCHPREM is the black market exchange rate premium from the WDR
dataset. The variable used in regressions is ln(1 + EXCHPRM).
ZTOTZ is the log difference of the terms of trade from the WDR dataset.
SURRAT is the budget surplus ( + ) or deficit ( - ) provided by William
Easterly.
AFRICA, ASIA, EASIA, LACAR, OECD, and SASIA are regional dummies.

References

Aizenman, Joshua and Nancy Marion, 1991, Macroeconomic uncertainty, persistence and growth,
Mimeo., Dec. (Dartmouth College, Hanover, NH).
Barro, Robert J., 1991, Economic growth in a cross section of countries, Quarterly Journal of
Economics 106,407~444.
Chenery, Hollis B., Sherman Robinson, and Moshe Syrquin, 1986, industrialization and growth
(Oxford University Press, New York, NY).
Corden, Max, 1990, Macroeconomic policy and growth: Some lessons of experience, Proceedings of
the World Bank Annual Conference on Development Economics, 59-84.
Cukierman, Alex, Pantelis Kalaitzidakis, Lawrence H. Summers, and Steven B. Webb, 1992, Central
bank independence, growth, investment, and real rates, Presented at Carnegie-Rochester Con-
ference on Public Policy, Nov.
De Gregorio, Jose, 1993, Inflation taxation, and long-run growth, Journal of Monetary Economics
31, 271-298.
Dellas, Harris, 1990, Stabilization policy and long term growth: Are they related? A Darwinian
perspective, Mimeo. (University of Maryland, College Park, MD).
De Long, J. Bradford and Lawrence H. Summers, 1991, Equipment investment and economic
growth, Quarterly Journal of Economics 106, 369406.
De Long, J. Bradford and Lawrence H. Summers, 1992, Macroeconomic policy and long-run
growth, Federal Reserve Bank of Kansas City Economic Review 77, no. 4, 5-30.
Dornbusch, Rudiger and Stanley Fischer, 1993, Moderate inflation, World Bank Economic
Observer 7, l-44.
Easterly, William and Sergio Rebelo, 1992, Fiscal policy and growth: An empirical investigation,
Mimeo. (World Bank, Washington, DC).
Elias, Victor J., 1992, Sources of growth (ICS Press, San Francisco, CA).
Fischer, Stanley, 1983, Inflation y crecimiento (Inflation and growth), Cuadernos de Economia 20,
267-278 (Sidrauski memorial lecture, in English as NBER working paper no. 1235).
Fischer, Stanley, 1991, Macroeconomics, development, and growth, NBER Macroeconomics
Annual, 329-364.
Fischer, Stanley, 1992, Growth: The role of macroeconomic factors, Mimeo. (M.I.T., Cambridge, MA).
512 S. Fischer, Macroeconomic factors in growth

Fischer, Stanley and Franc0 Modigliani, 1978, Towards an understanding of the real effects and
costs of inflation, Weltwirtschaftliches Archiv, 81(t-832. Reprinted in: S. Fischer, 1986, Indexing,
inflation, and economic growth (M.I.T. Press, Cambridge, MA).
Greene, William H., 1993, Econometric analysis (Macmillan, New York, NY).
Grier, Kevin B. and Gordon Tullock, 1989, An empirical analysis of cross-national economic
growth, 1951-80, Journal of Monetary Economics 24, 259-276.
Gylfason, Thorvaldur, 1991, Inflation, growth, and external debt: A view of the landscape, The
World Economy 14,279-298.
Hamilton, James D. and Marjorie A. Flavin, 1986, On the limitations of government borrowing:
A framework for empirical testing, American Economic Review 76, 808-819.
Kormendi, Roger C. and Philip G. Meguire, 1985, Macroeconomic determinants of growth:
cross-country evidence, Journal of Monetary Economics 16, 141-164.
Levine, Ross and David Renelt, 1992, A sensitivity analysis of cross-country growth regressions,
American Economic Review 82,942-963.
Levine, R. and S. Zervos, 1992, Looking at the facts: What we know about policy and growth from
cross-country analysis, Mimeo., Oct. (World Bank, Washington, DC).
Little, Ian, Richard Cooper, Max Corden, and Sarath Rajapatirana, 1992, Boom, crisis and
adjustment: The macroeconomic experience of developing countries, Mimeo.
Lucas, Robert E., 1973, Some international evidence on output-inflation tradeoffs, American
Economic Review 63, 326-334.
Mackenzie, G. A., 1989, Are all summary indicators of the stance of fiscal policy misleading?, IMF
Staff Papers 36, 743-770.
Mankiw, N. Gregory, David Romer, and David N. Weil, 1992, A contribution to the empirics of
economic growth, Quarterly Journal of Economics 107, 407-438.
Morris, Cynthia T. and Irma Adelman, 1988, Comparative p&terns of economic development
1850-1914 (Johns Hopkins Press, Baltimore, MD).
Pindyck, Robert and Andres Solimano, 1993, Economic instability and aggregate investment,
NBER Macroeconomics Annual, 259-303.
Romer, Paul M., 1989, Human capital and growth: Theory and evidence, NBER working paper no.
3173.
Solimano, Andres, 1989, How private investment reacts to changing macroeconomic conditions:
The case of Chile in the eighties, Working paper no. 212 (World Bank, Washington, DC).
World Bank, 1989, Adjustment lending: An evaluation of ten years of experience, Policy and
research series paper no. 1 (World Bank, Washington, DC).
World Bank, 1990, Adjustment lending policies for sustainable growth, Policy and research series
paper no. 14 (World Bank, Washington, DC).
World Bank, 1991, World development report (World Bank, Washington, DC).
World Bank, 1992, The third report on adjustment lending: Private and public resource for growth,
March (World Bank, Washington, DC).