Chapter

CHAPTER 5 Economic Growth: What Has Been Achieved and How?

Author(s):
Susan Schadler, and Hugh Bredenkamp
Published Date:
June 1999
Share
  • ShareShare
Show Summary Details
Author(s)
Kalpana Kochhar and Sharmini Coorey 

The promotion of sustainable economic growth is one of the principal objectives of ESAF-supported programs. This chapter examines the pattern of growth in ESAF countries over the past 10–15 years, drawing comparisons with other developing countries, and seeking to explain differences in growth outcomes among countries and over time. The chapter begins with a brief outline of the stylized facts. It then presents a general overview of the methodology used to examine the empirical behavior of long-term growth and surveys the extensive literature on this subject to identify the various policy-related and other factors that have been found to influence growth. An equation relating growth to these factors is estimated on data for 84 low- and middle-income countries, and it is used to compare the growth performance of ESAF countries with that of other developing countries (information on the data and results is reported in an appendix). A more detailed analysis focuses on the impact of structural policies on growth in ESAF countries. The chapter concludes with a summary of its findings.

The Stylized Facts

During the early 1980s, developing countries—on average—experienced virtual stagnation in per capita real GDP. For those nontransition countries that subsequently availed themselves of financial support under the SAF and the ESAF, the first half of the 1980s was a particularly bleak period during which real per capita incomes dropped at an average annual rate of 1.4 percent. In the ten years thereafter, growth in these countries picked up, and the gap between their growth rates and the average of other developing countries was eliminated by 1995 (Table 5.1).

Table 5.1Growth in Real Per Capita GDP in ESAF Countries and Other Developing Countries(Annual average, in percent)
1981–851986–901991–951995
ESAF countries-1.10.4-2.0
ESAF (excluding transition)-1.40.40.31.5
Africa-1.80.4-0.31.2
CFA-2.0-0.6-0.23.5
Non-CFA-1.71.0-0.4-0.1
Asia (excluding transition)2.32.32.73.4
Western Hemisphere-3.2-1.91.51.5
Non-ESAF developing countries10.31.01.01.4
Sources: World Economic Outlook (Washington: IMF, various issues); and IMF staff estimates.

Eighty-four low- and middle-income nontransition developing countries comprising 90 non-ESAF developing countries as defined in the World Economic Outlook less 6 countries classified as high-income by the World Bank (World Development Indicators database).

Sources: World Economic Outlook (Washington: IMF, various issues); and IMF staff estimates.

Eighty-four low- and middle-income nontransition developing countries comprising 90 non-ESAF developing countries as defined in the World Economic Outlook less 6 countries classified as high-income by the World Bank (World Development Indicators database).

However, this improvement in growth among the world’s poorest developing countries was not evenly based across different regions. Growth in Western Hemisphere ESAF countries, which was sharply negative during the 1980s, turned around dramatically in the early 1990s, while that of Asian ESAF countries continued to be strong. In contrast, there appeared to be a setback among African ESAF countries, whose growth again turned negative in the early 1990s so that, on average, per capita GDP growth rates in Africa during the past decade remained close to zero. However, a closer look at the individual African countries reveals that the growth picture varied widely, with some countries showing sizable gains throughout this period (Figure 5.1). More encouraging, there appears to have been a marked upturn in growth across the continent since 1994, in part reflecting the devaluation of the CFA franc.

Figure 5.1Real Per Capita GDP Growth in ESAF Countries1

(Period average; in percent)

Source: IMF staff estimates.

1Excluding transition economies.

The improvement in growth in ESAF countries was, in general, mirrored in increased investment and saving rates, although both remain lower in these countries, on average, than in other developing countries—the gap is especially large in the case of saving rates (Table 5.2). There was also substantial regional variation in the behavior of saving and investment. In contrast to Asia and the Western Hemisphere, saving showed little improvement in Africa. Although investment picked up in non-CFA Africa and the Western Hemisphere, it stagnated in relation to GDP in CFA Africa and Asia.

Table 5.2Trends in Saving and Investment Rates(Annual average, in percent of GDP)
National SavingInvestment
1981–851986–901991–951981–851986–901991–95
ESAF (excluding transition)18.08.99.917.018.520.3
Africa8.39.79.516.818.720.3
CFA8.610.49.816.717.017.2
Non-CFA8.29.39.416.819.722.0
Asia9.811.013.319.218.719.7
Western Hemisphere13.30.57.816.016.722.1
Non-ESAF developing countries218.618.617.424.622.623.3
Source: World Economic Outlook (Washington: IMF, various issues); and IMF staff estimates.

Excludes Guyana.

Eighty-four low- and middle-income nontransition developing countries comprising 90 non-ESAF developing countries as defined in the World Economic Outlook less 6 countries classified as high-income by the World Bank (World Development Indicators database).

Source: World Economic Outlook (Washington: IMF, various issues); and IMF staff estimates.

Excludes Guyana.

Eighty-four low- and middle-income nontransition developing countries comprising 90 non-ESAF developing countries as defined in the World Economic Outlook less 6 countries classified as high-income by the World Bank (World Development Indicators database).

A comparison of developments in growth, saving, and investment before and since countries embarked on their first SAF/ESAF arrangement reveals a brighter picture. With respect to real growth, the data shown in Figure 5.2 depict a marked increase in per capita GDP growth, on average, for all ESAF countries in the three years after the start of SAF/ESAF-supported adjustment compared with the three preceding years. Only in countries in the CFA franc zone did per capita GDP continue to decline during the first SAF/ESAF arrangements. The turnaround is most dramatic in non-CFA African countries and countries in the Western Hemisphere. Transition economies also recorded a significant improvement in growth during ESAF-supported programs.

Figure 5.2Real Per Capita GDP Growth in Adjustment Time

(In percent)

Source: IMF staff estimates.

1Excluding transition economies.

Saving and investment rates before and after the first SAF/ESAF arrangement are shown in Figure 5.3. Here, too, there were marked improvements after SAF/ESAF-supported adjustment began. Investment picked up sharply in non-CFA Africa, the Western Hemisphere, and transition economies. Saving rates also rose markedly, albeit from a low base, in the latter two regions, as well as in CFA Africa.

Figure 5.3Saving and Investment Ratios

(In percent of GDP)

Sources: IMF, World Economic Outlook; and IMF staff estimates.

1Excluding transition economies.

2Excluding Guyana.

Overview of the Methodology

The forces underlying economic growth have been the subject of renewed interest in the economic profession since the mid–1980s, particularly following the development of a theory of economic growth in which distortions and policy interventions can be shown to affect not only the levels of income, but also its steady-state growth rate—the so-called theory of endogenous growth.1 Since then, a vast and growing theoretical and empirical literature has emerged on the factors that influence long-term growth and that can help distinguish between high and low growth rates across countries and over time.2 There are several approaches that have been taken in the empirical literature, all based on cross-country data—either cross-sectional, or pooled cross-sectional and time-series—each with its own uses and limitations. One such approach is to estimate, using cross-sectional data and regression analysis, the parameters of an underlying common production function. Although this approach permits the examination of the contribution of accumulation of various factors of production to growth, it leaves unexplained the determinants of factor accumulation and productivity and, in particular, the role of policies in promoting or retarding growth.

Partly in response to this limitation and partly inspired by the development of the theory of endogenous growth, a second, more eclectic approach was developed in which growth is viewed as a function of factor accumulation and other conditioning variables, the latter typically proxying the effects of economic policies. One problem with this approach, as noted by Fischer (1993), is the difficulty of interpreting the results from a regression that includes the investment rate and macroeconomic and structural policy variables on the right-hand side, since most or all of the policy variables would also influence the pace of capital accumulation. Furthermore, including the investment rate in the growth regression gives rise to a potential bias in the estimated coefficients, owing to the simultaneity between the two variables. These limitations gave rise to another empirical approach, which involves decomposing growth rates into the part attributable to factor accumulation and that attributable to changes in factor productivity, imposing constant (over time) and common (across countries) factor shares. Separate regressions are then estimated for growth, capital accumulation, and growth in total factor productivity, with a number of policy and other relevant variables included as regressors (see Fischer, 1993; and Bosworth, Collins, and Chen, 1995). This approach permits the examination of the impact of policies on growth and then, separately, on the principal mechanisms through which the effects of policies are transmitted to growth.

Since the major focus of this chapter is on identifying policy variables and other influences that help to distinguish the growth outcomes of ESAF countries from those in other developing countries, the methodology adopted here focuses on the first of the three regressions in the last approach described above—that is, one in which pooled cross-sectional and time-series analysis is used to examine directly the influence of policies and other factors on growth. Under the maintained hypothesis that growth, capital accumulation, and factor productivity are influenced by policies (as well as by exogenous shocks and initial conditions), the growth equation estimated here can be interpreted as a reduced-form equation.3 This regression analysis is intended to provide a broad overview of the determinants of growth and adds to and complements the conclusions emerging from a recent study of the response of investment and growth to adjustment policies, which was based on the detailed analysis of eight country cases (Goldsbrough and others, 1996; see Box 5.1).

Determinants of Growth

Empirical growth studies have examined a large number of variables as determinants of growth. Given the inherent difficulty of constructing simple measures of policy variables, it has generally proven necessary to use proxies that can also be affected by non-policy-related factors. In view of the complex nature of the interaction between policies and growth, it has often been difficult to identify statistically robust relationships between growth and most of the observable proxies for policies. Nevertheless, the following factors have been identified with some regularity in the growth literature.

Convergence

The first and most widely examined issue is the one of “convergence” between richer and poorer countries. An important implication of the neoclassical growth model is that incomes in poorer countries will, over time, approach real incomes in rich countries, in part because of the closing of the technology gap. Most evidence on growth rates contradicts this prediction—poor countries typically also have the lowest growth rates. Ample evidence, however, supports the notion of “conditional convergence”—that is, once account is taken of other growth determinants, including policies, there is evidence that poor countries do grow more rapidly than richer ones. This hypothesis is tested in empirical growth models by including, as an explanatory viable, initial-period real GDP levels or the gap between a country’s initial-period GDP and that of a benchmark country, typically the United States.

Human Capital Accumulation

Another widely examined implication of the recent advances in growth theory is that human capital accumulation is a key determinant of growth. The important feature of the new growth theory that gives rise to the implication that policies can affect growth rates is that of nondiminishing returns to capital accumulation. The question of explaining growth is thus one of determining whether diminishing returns are really a constraint, and, if not, how they can be overcome. One possibility is that “knowledge” or human capital accumulation eliminates diminishing returns to capital, implying that labor inputs should be proxied not simply by population or labor force growth but rather by a combination of that and “knowledge.” Empirical studies have included various proxies to capture the accumulation of knowledge, including school enrollment ratios and other more sophisticated measures of schooling (Barro and Lee, 1993; Nehru and Dhareshwar, 1993). Typically, the empirical relationship between growth and these proxies for human capital accumulation has been found to be robust, although the significance of the specific measure of human capital accumulation varies across different samples of countries and time periods.

Macroeconomic Policies

Typically, the effect of macroeconomic policies or stability on growth is proxied by the rate of inflation and some measures of its variability. Fischer (1991, 1993) and others have argued that the inflation rate is the “single best indicator” of the extent to which macroeconomic policies are conducive to growth. Thus, although inflation may affect growth directly, the inflation variable may also capture the effects of policies tied to inflation. Notwithstanding the theoretical appeal of the arguments that high inflation is damaging to growth, the empirical support for this relationship has been mixed. However, more recent studies, particularly those allowing a nonlinear relationship and using panel (rather than purely cross-sectional) data, have tended to support a negative relationship, at least for inflation above the single-digit range. (The issue of the robustness of the inflation-growth association is addressed extensively in Chapter 6, as are questions of its causal interpretation.)

The effects of macroeconomic policies are also examined more directly by including fiscal policy indicators in the growth equation. Budget balances are often included as a proxy for stable and sustainable macroeconomic policy and are typically measured as the ratio of central or general government balances to GDP (see Fischer, 1991, 1993; Levine and Zervos, 1993; and various others).4 The theoretical rationale for including this measure is that, other things being equal, a higher government deficit implies (1) a larger claim by the government on the economy’s financial resources, (2) crowding out of the private sector by lowering access to credit and by raising real interest rates, and (3) a more appreciated real exchange rate, all of which can have an adverse effect on growth. Most empirical studies have tended to find that larger overall budget deficits are associated with lower growth.

Box 5.1Main Findings of Reinvigorating Growth in Developing Countries: Lessons from Adjustment Policies in Eight Economies (Goldsbrough and others, 1996)

This study examined why the medium-term response of growth and investment to reform has frequently been slow, even in countries that have undertaken significant adjustment. It drew upon evidence from eight countries (Bangladesh, Chile, Ghana, India, Mexico, Morocco, Senegal, and Thailand), three of which were ESAF countries. The main findings were as follows:

  • Adjustment episodes were typically triggered by large financial imbalances, extensive structural problems, or both. Most countries had suffered adverse external shocks that typically interacted with an inadequate policy response and heavy external borrowing to precipitate a crisis and a severe external financing constraint.
  • Countries that experienced especially severe episodes of macroeconomic instability, including very high inflation, had growth rates well below the world average during these episodes. By contrast, the confidence generated by a track record of stable, market-oriented economic policies appears to be important in generating growth rates consistently above the world average.
  • The abrupt contraction of domestic absorption resulting from a crisis typically felt heavily on investment—public and private—and gave little time to sustain output by switching resources to tradables. Several countries experienced a “pause” in private investment that lasted two to four years after the adoption of improved macroeconomic policies and structural reforms, reflecting the negative influence on investment of uncertainty and instability. The implementation of mutually consistent policies—an integral component of which is a sustainable medium-term fiscal position—was essential to promote more rapid resource switching and to reduce lags in the response of investment.
  • The largest and most lasting increases in saving were achieved in countries with the best growth records, suggesting a mutually reinforcing effect over the medium term. As also suggested by a number of other studies, the most important policy influence on overall saving appeared to be changes in the level of public saving, although a history of macro-economic stability and public sector reforms (of the tax, public enterprise, and pension systems) also appeared to contribute to increases in saving.
  • Robust empirical links could not be identified between external financing and growth, although the country examples tended to support the view that timely availability of additional financing is likely to enhance growth prospects, provided that supporting policies are in place. Over the medium term, countries with favorable structural policies and improved public saving were the most successful in attracting capital inflows and channeling them toward increased investment, especially in tradables.
  • Countries that began with relatively small structural distortions or made significant progress toward eliminating major distortions tended to experience the most rapid productivity gains, and vice versa. Although there is evidence of complementarities between certain key reforms, no single blueprint emerges for the appropriate sequencing of structural reforms.
  • Countries with a weak growth record tended to have considerable initial labor market rigidities that impeded real wage flexibility; these features remained largely unchanged during adjustment. Patchy evidence suggests that real wage flexibility had important effects on the responsiveness of employment to output and on limiting the negative effects of disinflation on output and employment.

Other macroeconomic policy-related indicators that have been widely examined are the level of external indebtedness and the degree of misalignment of real exchange rates. There has been a great deal of interest, both theoretical and empirical, in the effects of a heavy burden of external indebtedness on economic growth (Borensztein, De Gregorio, and Lee, 1994; Eaton, 1987; Cohen, 1996; and Elbadawi, Ndulu, and Ndung’u, 1996). From a theoretical standpoint, several channels of influence have been identified: higher debt burdens may undermine macroeconomic stability by raising budget deficits; heavy debt-service burdens could raise uncertainty about future increases in potentially distortionary taxation needed to bring about the necessary resource transfer (and possibly also foreign exchange shortages), thus choking off access to private capital markets and damaging private incentives to invest; and, finally, using domestic resources to service debt may crowd out domestic investment, potentially giving rise to a vicious cycle of stagnant growth and greater debt. However, the empirical evidence is mixed regarding the negative impact of external debt or debt service on growth. Fischer (1991), Cohen (1996), and Elbadawi, Ndulu, and Ndung’u (1996) are examples of studies that found that high external indebtedness is negatively associated with growth, while Hadjimichael and others (1995) did not find a significant association between external indebtedness and growth.

Some studies (Elbadawi, Ndulu, and Ndung’u, 1996; Bosworth, Collins, and Chen, 1995; and Ghura and Hadjimichael, 1995) have included the change in the real effective exchange rate as an indicator of the appropriateness of exchange rate policy on the grounds that most (non-oil-exporting) developing countries experienced a major deterioration in their terms of trade during much of the period under consideration, which in turn would imply, other things being equal, a depreciation of the equilibrium real value of the domestic currency. Thus, it is argued that an observed real effective depreciation can be interpreted as a movement toward equilibrium (and sustainable) exchange rates. These studies do find a significant negative correlation between real exchange rate appreciation and growth.

Openness

The degree of openness of an economy can affect growth through many channels. International trade not only allows countries to access larger markets and thereby take advantage of economies of scale, but it also improves resource allocation—hence generating efficiency gains—by lowering distortions in internal relative prices. Trade orientation is also postulated to affect growth through the beneficial effects of increased exposure to competition (Grossman and Helpman, 1991) and through positive externalities that result from the transfer of knowledge and technology (Romer, 1986 and 1990). These predictions have been tested empirically using various measures of openness to trade, including ratios of total trade to GDP, trade taxes, and more sophisticated ones such as those constructed by Leamer (1988) and used by Edwards (1992). Leamer’s measures of trade policies were derived by estimating a Heckscher-Ohlin model of trade, with the deviation between the actual and predicted pattern of trade taken as a measure of trade intervention and that between the actual and predicted level of trade taken as a measure of openness.5 Several studies, including Levine and Renelt (1992) and Sachs and Warner (1995), did find a robust correlation between growth and openness to international trade.

Structural Distortions

One widely used proxy for structural distortions is the black market exchange rate premium, which is typically interpreted as a crude indicator of overall distortions, a measure of the appropriateness of the exchange rate, and a more general indicator of policy uncertainties.6 This variable has not been found to be robust in studies of growth determinants. Fischer (1991 and 1993) found, for example, that it is strongly negatively correlated with investment but does not appear to affect the rate of growth, whereas Levine and Zervos (1993) found that it does have a robust negative correlation with growth.

Several researchers have also examined the relationship between the size of the government, proxied by the ratio of government consumption to GDP, and growth. Although this indicator does not necessarily capture the productivity of expenditures, it is sometimes interpreted as a measure of the distortions associated with unproductive government spending, which not only lowers current per capita growth but also, by creating expectations of an increase in future tax liabilities, lowers future growth (Barro, 1991). On the other hand, government consumption includes expenditures that could be beneficial to growth, particularly in the areas of health and education. There is little consensus in the theoretical literature about the extent of government involvement in the economy that is conducive to growth. The empirical support is similarly mixed, with most studies finding that the correlation between government consumption and growth is negative, but not always strongly significant.

The effect of financial sector development on growth is another subject that has recently witnessed a resurgence in interest. Early proponents of financial sector liberalization as a factor influencing growth were McKinnon (1973) and Shaw (1973). More recently, the mechanisms through which the financial system can affect growth have been examined in Greenwood and Jovanovic (1990) and King and Levine (1993). According to these studies, financial sector development is positively associated with growth because financial intermediation enables savers to pool resources and share risks. Saving channeled through financial intermediaries is allocated more efficiently because such intermediation tends to reduce liquidity risks and partly overcomes the problem of adverse selection in credit markets; by expanding the availability of credit, financial intermediaries facilitate innovative activity and the process of “creative destruction.” Roubini and Sala-i-Martin (1991) used various measures of financial repression and found them to be significantly negatively correlated with growth. Similarly, De Gregorio and Guidotti (1992) and King and Levine (1993) found that proxies for the degree of financial sector development such as the ratio of broad money to GDP and the ratio of private sector credit to GDP are significantly positively correlated with growth.

Other Factors

In addition to policy-related variables, proxies for human capital accumulation, and initial conditions, most growth studies include measures of a variety of shocks, such as changes in the terms of trade, disruptive political events (such as coups, assassinations, and revolutions), and weather-related disturbances. Other studies have examined the correlation among growth and natural resource endowments (Sachs and Warner, 1995); the quality of governance, including law-and-order traditions, the quality of political institutions, corruption, and economic freedom (Sachs and Warner, 1995; Mauro, 1996); ethnic diversity (Easterly and Levine, 1996); and a host of other noneconomic features of countries.

Many empirical studies also include regional dummy variables to examine whether there is a significant unexplained growth differential between countries of a particular region and the average for the whole sample of countries under review. They typically find that, even after taking into account identifiable factors influencing growth, there is a significantly negative unexplained growth differential between African countries and the rest of the sample of countries—the negative “Africa dummy.” A few studies, however—notably Easterly and Levine (1996) and Sachs and Warner (1995)—have been able to “explain away” the Africa dummy by including measures of ethnic diversity and regional spillovers proxied by growth in neighboring countries (Easterly and Levine) or by taking into account the lack of physical access to markets and high reliance on natural resource exports (Sachs and Warner).

Specification of the Growth Equation and Estimation Results

In light of the large body of evidence that has already been extracted from the data, the intention here is not to expand or elaborate on the conventional array of growth determinants already identified. Rather, the principal aim is to examine the strength of the identified relationships for the ESAF developing countries. In an important paper, Levine and Renelt (1992) demonstrated that many of the relationships discussed above are not robust to changes in specification and sample characteristics.7 One question that is considered, therefore, is whether growth in ESAF countries is influenced by the same factors as in other developing countries. From there, the analysis examines what an estimated growth equation can reveal about the factors that account for the differences among growth performance in ESAF users and other developing countries, and about the evolution of ESAF countries’ growth over time. To address these questions, a large sample of low- and middle-income developing countries was used with annual data covering the period 1991–95, which broadly spans the pre-SAF/ESAF period and subsequent phases of structural adjustment for most countries under review.

Most of the variables discussed above are considered in the analysis that follows. Macroeconomic policies are proxied by the rate of inflation, budget balances (BUDBAL), measures of external indebtedness, and changes in real exchange rates (DREER). To examine the effect of inflation on growth, however, this analysis follows a suggestion made by Fischer (1993) and used by Sarel (1996) and others to include a logarithmic transformation of the inflation rate in the regressions (LINFL). The rationale for using this specification is twofold. First, it enables us to account for the intuitively appealing notion that moving from inflation rates of 10 to 20 percent is not the same as moving from rates of 110 to 120 percent or 1010 to 1020 percent. Second, the distribution of inflation rates is highly asymmetric, with a concentration of observations in the low-inflation range and a few observations in the very high inflation range. Thus, using inflation rates themselves would tend to place very heavy weight on a small number of high inflation observations. In contrast, the logarithm of inflation has a more balanced distribution.8 In addition, a dummy variable for observations with a negative inflation rate (DEFL) is included to reflect the possibility that deflation may be harmful to growth.

A number of authors have suggested that the negative inflation-growth association breaks down somewhere below 10 percent. Therefore, following Sarel (1996), the possibility of a further nonlinearity in the effect of inflation on growth is examined by including as a regressor a variable, EXTRAΠ*, defined as follows:9

EXTRAΠ* = dumΠ*[In (Π) - In(Π*)],

where Π* is the rate of inflation at which the non-linearity occurs, and dumΠ is equal to 0 if Π > Π* and equal to 1 otherwise. The coefficient on EXTRAΠ* measures the difference in the coefficients on inflation on each side of the “kink.” The bivariate relationship between annual inflation rates and growth suggests that such a break or kink occurs at a rate near 5 percent (see Chapter 6). Sarel (1996), using a different data set and equation specification, found that the break occurs at a rate of about 8 percent. Hence the search for the break point in the sample was conducted over the range from 2 percent to 10 percent. The rate (Π*) at which the kink in the otherwise negative relationship between growth and inflation occurs was found to be 4.5 percent.10

Three indicators of the burden of external indebtedness were used in this analysis—the (logarithm of the) ratio of total external debt to GNP, the debt-service ratio,11 and a dummy variable that takes the value of 1 for the first year of a debt-rescheduling agreement and the value of 0 otherwise.

As regards structural policies, in addition to the logarithm of the black market exchange rate premium, proxies were included for the size of government, openness to trade, and financial sector development. The size of government was proxied by the ratio of government consumption to GDP (GCONS). Openness to trade (OPENIND) was measured by an index constructed from the residual of a regression of the ratio of total trade to GDP on the size of the population—the latter used as a proxy for the size of the economy. The more outward-oriented the economy, the larger should be the regression residual. The degree of financial sector development was proxied by the ratio of average broad money to GDP (DEPTH) and the ratio of private sector credit to GDP.

The equation also attempts to link growth to key features of the economic and political environment affecting private investment decisions. It includes a variable, ECONSEC (based on data taken from Knack and Keefer, 1995), that measures factors affecting economic security such as expropriation risk, repudiation of contracts by the government, corruption, the quality of the bureaucracy, and the law-and-order tradition.12

In addition to the rate of population growth (POPG), several measures of human capital development were considered—gross primary and secondary school enrollment ratios, and the logarithm of life expectancy at birth (LLIFE). Initial conditions (convergence effects) were controlled for by including the logarithm of GDP in 1980 measured in U.S. dollars adjusted for purchasing power parity (PPP) (LGDP80).

The regressions also included three types of exogenous supply shocks: the change in the terms of trade lagged one year (DTOT1); a dummy variable that takes on the value of 1 for wars, episodes of civil unrest or political instability, and the value of 0 otherwise (WAR); and a dummy variable (WEATHER) that is interpreted as capturing abnormal weather conditions such as droughts and floods, and takes the value of 1 when per capita food production falls by 5 or more percentage points and the value of 0 otherwise.13

Two issues remain open to judgment. The first is whether the rate of capital accumulation (proxied by the rate of investment) should be included as an explanatory variable. Omitting the investment rate from the equation raises the possibility of an omitted-variables bias in the estimated coefficients. On the other hand, including it involves the potential for a simultaneity bias in the estimate, given that growth itself can have an important effect on investment. Since the purpose of this chapter is to examine the influence of policy and exogenous variables on growth, and the estimated equation is intended to be interpreted as a reduced form (see the earlier discussion on methodology), it was decided to exclude the rate of investment from the specification. A Hausman specification test suggests that—after controlling for other influences on growth—investment is not exogenous in the growth equation (Table 5.3).14 As regards the potential omitted-variables problem, insofar as many of the right-hand-side variables that are included also affect investment, this bias is not likely to be a major problem.

Table 5.3Determinants of Growth(Pooled annual data for 84 low- and middle-income developing countries, 1981–95, as available; dependent variable: real per capita GDP growth)
VariableCoefficientt-Statistic (absolute values, based on heteroscedastic consistent standard errors)
CONSTANT-5.6581.18
POPG-0.8254.38**
LLIFE5.0233.42**
LGDP80-1.3924.10**
LINFL (> 5 percent)-0.7525.96**
LINFL (≤ 5 percent)0.4561.62
EXTRA5 (kink)1.2083.63**
DEFL-1.9532.64**
BUDBAL0.1014.00**
OPENIND0.0193.42**
GCONS-0.0712.40**
ECONSEC0.2221.74*
WEATHER-2.0687.36**
WAR-0.7462.01**
DTOT10.0332.733
Source: IMF staff estimates.Note:

Number of observations= 994Hypothesis tests
Adjusted R2= 0.22Hausman Test for exogeneity of investment Test statistic: 7.29**
Jarque-Bera Normality test= 254.7 (**)Test statistic: 7.29**
F-statistic (zero slopes)= 22.1 (**)F-test for joint equivalence of coefficients across ESAF and non-ESAF subsamples Test statistic: 1.43**

POPG= annual population growth (in percent)
LLIFE= log of life expectancy at birth (in years), lagged five years
LGDP80= log of real per capita GDP in 1980 (in PPP-adjusted U.S. dollars)
LINFL= log of the absolute value of annual average CPI inflation
EXTRA5= DUM5*(LINFL-log(5.0), where DUM5 takes the value 0 when inflation exceeds 5 percent and 1 otherwise)
DEFL= dummy variable equal to 1 when inflation is negative and 0 otherwise
BUDBAL= general government balance as a ratio to GDP
OPENIND= adjusted openness variable defined as the residual from a regression of the ratio of total trade to GDP on population and converted to an index
GCONS= government consumption as a ratio to GDP (in current prices)
ECONSEC= index measuring “economic security” as proxied by five indicators measuring expropriation risk, repudiation of contracts by the government, corruption, the quality of bureaucracy, and the law and order tradition. A higher number represents greater security.
WEATHER= dummy variable equal to 1 when annual per capita food production declines by 5 percent or more and 0 otherwise
WAR= dummy variable equal to 1 when there is a war or episode of civil unrest and instability and 0 otherwise
DTOT1= annual average percent change in the terms of trade, lagged one period

Indicates significance at the 5 percent level; * indicates significance at the 10 percent level.

Source: IMF staff estimates.Note:

Number of observations= 994Hypothesis tests
Adjusted R2= 0.22Hausman Test for exogeneity of investment Test statistic: 7.29**
Jarque-Bera Normality test= 254.7 (**)Test statistic: 7.29**
F-statistic (zero slopes)= 22.1 (**)F-test for joint equivalence of coefficients across ESAF and non-ESAF subsamples Test statistic: 1.43**

POPG= annual population growth (in percent)
LLIFE= log of life expectancy at birth (in years), lagged five years
LGDP80= log of real per capita GDP in 1980 (in PPP-adjusted U.S. dollars)
LINFL= log of the absolute value of annual average CPI inflation
EXTRA5= DUM5*(LINFL-log(5.0), where DUM5 takes the value 0 when inflation exceeds 5 percent and 1 otherwise)
DEFL= dummy variable equal to 1 when inflation is negative and 0 otherwise
BUDBAL= general government balance as a ratio to GDP
OPENIND= adjusted openness variable defined as the residual from a regression of the ratio of total trade to GDP on population and converted to an index
GCONS= government consumption as a ratio to GDP (in current prices)
ECONSEC= index measuring “economic security” as proxied by five indicators measuring expropriation risk, repudiation of contracts by the government, corruption, the quality of bureaucracy, and the law and order tradition. A higher number represents greater security.
WEATHER= dummy variable equal to 1 when annual per capita food production declines by 5 percent or more and 0 otherwise
WAR= dummy variable equal to 1 when there is a war or episode of civil unrest and instability and 0 otherwise
DTOT1= annual average percent change in the terms of trade, lagged one period

Indicates significance at the 5 percent level; * indicates significance at the 10 percent level.

A second, and related, issue concerns the direction of causality between the dependent and some of the right-hand-side variables. This problem is most relevant for inflation and the budget balance. It is possible that causation runs from growth to inflation or the budget balance, or that growth, inflation, and the budget balance are all “caused” by some other variable.15 For example, it is possible that an adverse supply shock causes inflation, slower growth, and higher budget deficits. The inclusion of supply shocks of the type discussed above would, to a large extent, mitigate this problem and would strengthen the case for interpreting the relationship between these macroeconomic indicators and growth as causal.

The regression analysis was conducted using annual data for 84 low- and middle-income developing countries for 1981–95. (See the appendix, Tables 5.13 and 5.14 for a complete list of countries included in the regression and the sources of data.)16 Since an important aim of the exercise is to compare factors contributing to growth in ESAF countries relative to a broadly similar group of other developing countries, high-income industrial and developing countries and transition economies were excluded from the sample.17 The sample includes all 30 nontransition ESAF users under review.18 Sample means for the complete set of variables are shown in the appendix (Table 5.6) along with simple pairwise correlations (Table 5.7). The equation was estimated by ordinary least squares, with a correction for heteroscedasticity performed on the standard errors of the estimated coefficients (see Table 5.3).19 The explanatory power (estimated coefficient of determination, R2) is relatively low, a common result in the empirical analysis of growth using annual observations for samples of developing countries.20 However, the value and significance level of the F-test, indicating that all the parameters are jointly significant, provides support for the specification, as does the finding that almost all of the estimated coefficients in the final specification remained relatively robust to changes in the specification.21 The key results from this analysis are as follows:

  • In keeping with the empirical literature, the coefficient on the logarithm of initial-period real GDP, used to test the hypothesis of conditional convergence of real incomes between poorer and richer countries, is found to be negative and statistically significant. The estimate suggests a sizable growth advantage—controlling for policies and exogenous shocks—for countries with low-income levels because of the closing of the technology gap.The measure of human capital development used here—the logarithm of life expectancy, lagged five years—is significantly positively correlated with growth. The finding that neither lagged primary nor secondary school enrollment ratios are significant in the equation is not typical to most empirical analyses of the determinants of growth.22 For this sample, however, that school enrollment ratios are highly positively correlated with the life expectancy variable (correlations of 0.78 and 0.73; see the appendix, Table 5.7) as well as the initial GDP variable may explain why it is not possible to identify significant independent effects from all three variables. The life expectancy variable can be seen as a more encompassing proxy, because it captures both the health and education dimensions of human capital development. The estimated effect is strong, suggesting, for instance, that increasing life expectancy from 50 years to 60 years—approximately the difference in the sample means between ESAF and non-ESAF countries (see the appendix, Table 5.6)—would increase per capita growth by almost 1 percentage point.23
  • The population growth rate is found to exert a strong negative influence on per capita growth, notwithstanding the positive contribution to growth associated with the faster growth of the labor force. Coupled with the finding of significant negative correlation (see the appendix, Table 5.7) between population growth and measures of human capital development, this suggests that policies intended to increase human capital development can have both a direct and an indirect positive impact on growth.
  • With respect to macroeconomic and structural policies, budget balances are found to be significantly positively associated with growth, whereas the size of government, measured by the ratio of government consumption to GDP, is significantly negatively correlated with growth. Likewise, the analysis suggests support in this sample for the idea that more open economies tend to enjoy higher growth.
  • As for the relationship between inflation and growth, the estimated coefficients of LINFL and EXTRA5 suggest that, at rates above 5 percent, inflation tends to have a significantly negative effect on growth, whereas at rates below that threshold the effect of inflation on growth is positive. The t-statistic on EXTRA5 indicates that the kink in the relationship between inflation and growth is statistically significant. In addition, the significant negative coefficient on the deflation dummy DEFL suggests that deflation is detrimental to growth.24 These findings should not be interpreted as pinpointing the exact rate at which a break occurs in the relationship between inflation and growth (as noted earlier, Sarel, 1996, found the kink at 8 percent). Rather, they provide evidence that such nonlinearities do exist, so that the estimated effect of intermediate and higher rates of inflation on growth may be underestimated in studies that have not taken nonlinearities into account, and they suggest a range of inflation rates that it would be prudent to target as part of a growth-enhancing policy package, other things being equal (see Chapter 6).
  • Economic security measured by the variable ECONSEC is found to have a positive and significant effect on growth, suggesting that stronger bureaucracies and law-and-order traditions and better governance, in general, are associated with more rapid growth. With respect to shocks, lagged improvements in the terms of trade are found to be positively associated with growth, whereas wars taking place on national territory and other disruptive episodes of civil unrest or instability and weather-related shocks are strongly negatively associated with growth.25
  • External indebtedness was proxied in different specifications by the (logarithm of the) ratio of total external debt to GDP, the debt-service ratio, or a dummy variable that takes the value of 1 for the first year of every rescheduling arrangement that a country made. The first two variables were insignificant in all specifications; this is not surprising in the case of the debt ratio because the numerator is not in net present value terms (owing to data unavailability) and hence does not reflect differences in the concessionality of external debt among countries. The sign in the rescheduling dummy could, in principle, be negative or positive depending on whether the variable captures more strongly cross-country differences in debt-servicing problems or the timing of rescheduling agreements within individual countries. In the former case, the variable would tend to capture mainly the negative effect on growth of countries that encounter frequent debt-servicing difficulties, whereas in the latter case it would mainly reflect the positive effect on growth of resolving uncertainty and, perhaps, reducing the debt burden. Although the coefficient on the dummy tended to be negative, it was not statistically significant and was dropped in the final step of the specification search. Thus, in this sample it is difficult to find strong evidence of debt-servicing difficulties affecting growth, after controlling for the effects of policies and exogenous shocks. Likewise, a strong or robust association is not found between growth and the black market premium, the change in the real exchange rate, or financial depth (measured by the ratio of broad money to GDP). These issues are discussed further below, under “Structural Reforms and Growth.”

Comparative Growth Performance

The next step in the analysis is to use the estimated coefficients reported in Table 5.3 to compare the different influences on growth from policies, shocks, and other factors between ESAF and non-ESAF countries. First, parameter stability across the subsamples of ESAF and non-ESAF countries was tested using a standard F-test (or Chow test). The results suggest that the null hypothesis that the parameter values for the ESAF subsample are (jointly) the same as those for the non-ESAF subsample cannot be rejected at the 5 (or 10) percent level (Table 5.3). This allows the full-sample equation to be used to draw inferences about the contributions of the various determinants of growth in the ESAF countries.26

The growth decompositions were conducted for three five-year periods (1981–85, 1986–90, and 1991–95), with the first period corresponding to the period just prior to the introduction of the SAF. Using the sample means and the estimated coefficients in Table 5.3, it is possible to examine the explained and unexplained sources of the growth differential between ESAF and non-ESAF countries over the period under review (Table 5.4, page 82).27 Was the convergence in the growth differential achieved by ESAF countries during the late 1980s and early 1990s because of good policies or good luck, or both?

Table 5.4Differences in Growth Between ESAF and Non-ESAF Countries1
Period I (1981–85)Period II (1986–90)Period III (1991–95)II over IIII over I
Actual growth differential-1.22-0.160.021.061.23
Estimated growth differential-0.91-0.46-0.300.450.61
Differential contributions
Macroeconomic policies-0.45-0.180.080.270.53
Inflation2-0.130.070.240.200.37
Budget balance-0.32-0.24-0.160.080.16
Structural policies-0.05-0.20-0.28-0.14-0.23
Openness0.03-0.06-0.07-0.10-0.10
Size of government0.01-0.08-0.04-0.10-0.06
Economic security-0.10-0.05-0.170.05-0.07
Population growth and human capital accumulation-1.38-1.31-1.510.07-0.14
Technological convergence1.291.291.290.000.00
Shocks3-0.32-0.070.130.250.45
Unexplained factors-0.300.310.320.610.62
Source: IMF staff estimates.

For each of the three periods, differential growth contributions are calculated by taking the difference (ESAF countries over non-ESAF countries) of the product of the estimated regression coefficient (reported in Table 5.3) and the respective sample mean for each variable. Estimated growth differentials are differences in the fitted growth rates of ESAF countries over non-ESAF countries. See appendix, Table 5.10 for the underlying growth decompositions.

Includes all three inflation terms, including the dummy variable for deflation.

Includes dummies for weather, war, and terms of trade shocks.

Source: IMF staff estimates.

For each of the three periods, differential growth contributions are calculated by taking the difference (ESAF countries over non-ESAF countries) of the product of the estimated regression coefficient (reported in Table 5.3) and the respective sample mean for each variable. Estimated growth differentials are differences in the fitted growth rates of ESAF countries over non-ESAF countries. See appendix, Table 5.10 for the underlying growth decompositions.

Includes all three inflation terms, including the dummy variable for deflation.

Includes dummies for weather, war, and terms of trade shocks.

The results suggest that over two-fifths of the narrowing in the actual growth differential over the past decade was attributable to improvements in macroeconomic policies, while about a third was due to a more favorable set of exogenous conditions (last column of Table 5.4). These positive influences were partially offset by the negative contribution of structural policies and human resource factors. Despite some trade liberalization, ESAF countries did not make as much progress in opening their economies to external trade as did non-ESAF countries, nor succeed in reducing the size of government, notwithstanding progress in some countries.

Looking at the level of the growth differential in the early 1990s, reductions in inflation and positive exogenous shocks helped to pull up growth in ESAF countries relative to non-ESAF countries, whereas lower economic security and higher budget deficits had a negative effect (column for period III in Table 5.4). An interesting result is that the gap between ESAF and non-ESAF countries in factors affecting economic security accounts for the major part of the shortfall in structural policies in the early 1990s, and, generally, for the 1985–95 period as a whole. Most striking, however, the advantage for ESAF countries—in terms of the potential for more rapid growth—of starting from a lower level of initial per capita income (“technological convergence”) was more than offset by the sizable negative impact of higher population growth rates and lower levels of human capital development.28

As regards exogenous shocks, ESAF countries, after suffering a sharp decline in their external terms of trade in the 1980s, experienced an improvement in the early 1990s, whereas the terms of trade of non-ESAF countries continued to deteriorate. The contrast between ESAF and non-ESAF countries’ terms of trade movements is especially sharp when comparing the early 1990s with the early 1980s. In addition, ESAF countries experienced fewer weather-related shocks in the late 1980s and early 1990s compared with the early 1980s, whereas non-ESAF countries faced more adverse weather conditions during this period.

However, about half of the turnaround in the growth gap appears to be attributable to factors not included in the regressions—possibly the effects of structural reforms that were not fully captured by the measures of policies included in the regression.

Structural Reforms and Growth

As discussed in Chapter 4, a major objective underlying the construction of the score indices of structural policies is to attempt to identify more systematically their links with economic growth.29 Because the data requirements for the construction of such indices for all low- and middle-income developing countries were prohibitive, this exercise focuses solely on the sample of ESAF users. The indices of structural policies in the four areas of interest—exchange and trade systems (EXT), financial sector development (FINSEC), public enterprise policies (ENTPR), and pricing and marketing policies (PRICE)—were discussed in greater detail in Chapter 4. Two types of scores are considered in each area: a period-specific score for each of three five-year subperiods (1981–85, 1986–90, and 1991–95) and a cumulative score obtained by summing the scores for each five-year subperiod.

The first step was to examine, in a multivariate regression framework that controls for other factors, the links between growth and the period-specific and cumulative scores, respectively, in the four structural policy areas. However, with the exception of the pricing and marketing policies variable (PRICE), the indices were statistically insignificant, sometimes of the wrong sign, and generally not robust to changes in specification and sample.30 One possible explanation for this finding is that there is a high degree of multi-collinearity among the individual policy scores and between the policy scores and the other explanatory variables.31

Moving, therefore, to a framework with a smaller set of variables, simple bivariate regressions of period-average per capita growth and the individual policy indicators in each five-year period were estimated using data pooled over time and across ESAF countries. The results, at five-year and annual frequencies, are shown in Table 5.5. On the basis of the uncumulated period scores, price liberalization is strongly positively associated with growth, but financial sector reforms have a somewhat weaker, positive relationship with growth. External and public enterprise reforms, however, do not appear to be significantly correlated with growth. This result may, to some extent, reflect the offsetting, short-run effects on output of trade and public enterprise reform.32 When the scores are cumulated over time, a much stronger positive association emerges between each individual index and growth, suggesting a payoff both to having a relatively undistorted structure to begin with and to making significant early progress with reforms.

Table 5.5Growth and Structural Reforms1
VariableExternal Sector ReformsPrice ReformsFinancial Sector ReformsPublic Enterprise ReformsUnweighted Average2First Principal Component2
Bivariate regressions of real per capita GDP growth, using uncumulated five-year period scores
Five-year data
Estimated coefficient0.180.60*0.44-0.110.540.41
t-Statistic (HCSE)0.681.861.26-0.351.281.36
Annual data3
Estimated coefficient0.180.59**0.44*-0.100.53*0.41*
t-Statistic (HCSE)0.952.411.78-0.471.701.80
Bivariate regressions of real per capita GDP growth, using cumulative scores
Five-year data
Estimated coefficient0.120.22**0.24**0.110.19**0.67**
t-Statistic (HCSE)1.622.392.601.332.052.10
Annual data3
Estimated coefficient0.12**0.22**0.24**0.11*0.19**0.66**
t-Statistic (HCSE)2.163.263.511.732.752.80
Source: IMF staff estimates.

t-Statistics are based on heteroscedasticity-consistent standard errors; ** denotes significance at the 5 percent level; * denotes significance at the 10 percent level.

Combining all four structural policy indices.

With the structural policy score for each five-year period repeated for each of the five years in the period.

Source: IMF staff estimates.

t-Statistics are based on heteroscedasticity-consistent standard errors; ** denotes significance at the 5 percent level; * denotes significance at the 10 percent level.

Combining all four structural policy indices.

With the structural policy score for each five-year period repeated for each of the five years in the period.

The impact of combinations of structural reforms on growth was also considered. Two alternative measures of the structural policy environment covering different combinations of the four individual indicators were regressed on the rate of growth:33 a simple unweighted average (as in Chapter 4) and the largest principal component.34 The results again suggest that, although the uncumulated effects are generally not significant, the cumulative effect of combinations of reforms—particularly in the areas of pricing and marketing, exchange and trade, and the financial sector—have a positive effect on growth (see the appendix, Table 5.11).

An attempt was also made to examine interactions between growth and the various indicators of structural reform in order to capture possible cross-effects or “complementarities” among structural reforms. The different combinations of the four structural reform indices and their respective cross-product terms were thus regressed on growth.35 The results do not show positive cross-effects or complementarities in the effects of different structural reforms on growth (see the appendix, Table 5.12). The coefficient on the cross-product terms is significantly negative (although much smaller than the coefficients on the individual indices) for the combination of price liberalization and external sector reform as well as price liberalization and financial sector reform for the uncumulated five-year scores. The cross-product coefficients are generally not statistically significant for the cumulated scores, with the exception of the combination of financial and external sector reforms. These results suggest that, although structural reforms in certain areas have a significant effect on growth, the quantitative effect is greater when reforms are undertaken singly than when they are combined with certain other reforms. This is consistent with the notion that the marginal effect of individual structural reforms on growth may diminish somewhat as other reforms are undertaken. The evidence does not, however, imply that the joint undertaking of structural reforms in these areas is harmful to growth, nor that implementing reforms sequentially is necessarily superior to undertaking them jointly.36

To sum up, the results suggest that structural reforms have favorable effects on growth, although the results are not very strong, particularly in a multivariate context. Cumulative effects are generally stronger than effects within an annual or five-year period, suggesting a payoff from not delaying implementation of reforms. There is no empirical support for the notion of complementarities among structural reforms, which would have suggested larger effects on growth of undertaking structural reforms simultaneously than individually. Hence, there is little or no evidence of a need for a “critical mass” of structural reforms: that is, a growth dividend from implementing reforms on a broad front. At the same time, the evidence does not indicate any negative overall effects on growth of implementing structural reforms in several areas simultaneously. Hence, by implication, there may be a case for focusing reform efforts selectively in a few areas where distortions are most severe. These findings must be interpreted with caution because of the severe measurement problems involved in calibrating structural distortions.

Concluding Remarks

The design of ESAF-supported programs has been guided by what have become widely accepted “best practices,” at least in countries starting from severely distorted economic structures. One of the questions this study set out to examine is whether the various determinants of growth—particularly those amenable to the influence of policies—identified in the extensive literature are also relevant for ESAF countries. The results confirm many of the findings of the empirical growth literature for the sample of low- and middle-income developing countries considered here. In particular, the analysis shows that in ESAF countries—just as in other developing countries—policies can play a significant role in improving growth. Stable macroeconomic policies, comprising single-digit inflation rates and low budget deficits, together with outward-oriented trade policies and a streamlining of the role of the government are all good for economic growth. In addition, policies aimed at increasing human capital, lowering population growth, and improving governance can all contribute in important ways to raising growth.

An examination of the contributions to growth of the various identified influences suggests that part of the marked narrowing of the growth gap between ESAF and non-ESAF countries can be attributed to stronger macroeconomic policies (inflation and budget deficits) and to more benign exogenous influences (in particular, the terms of trade and weather). However, policies have fallen short in several areas—notably, in reducing population growth, accelerating the accumulation of human capital, opening up to international trade, reducing the size of the government, and improving the quality of governance. The gap between ESAF and other developing countries has widened in these areas.

A more detailed examination of structural policies in the ESAF countries, with the aid of score indices constructed for the purpose, does not produce findings that are sufficiently robust to support firm policy conclusions. This may well reflect the enormous difficulties in measuring differences in structural policies across countries and over time. Bivariate correlations suggest that reductions in structural distortions are associated with more rapid growth over time. But such effects are barely discernible when full account is taken of macroeconomic policies, human capital accumulation, initial conditions, and exogenous shocks.

The results do not allow one to identify any particular aspect of structural reform as more important for growth; nor, however, do they provide evidence of “complementarities” or support for the hypothesis that a package of reforms together (a “critical mass”) has a larger impact than the sum of reforms taken individually.37 Hence, by implication, the findings would be consistent with an argument that reform efforts should focus on a small number of structural areas where problems are most severe.

In sum, we find some empirical evidence across a large group of countries in support of the basic orthodoxy of ESAF-supported adjustment programs. Perhaps reflecting the severe data limitations and the crude statistical tools at our disposal, the factors usually identified as determinants of growth explain only a small proportion of the variation in growth across countries and across time. Nevertheless, the available evidence does indicate that economic growth responds to the achievement and maintenance of macroeconomic stability, reinforced by structural reforms in key areas.38 The analysis also suggests some areas for improvement. More attention needs to be paid to human resource development and to improving institutions—such as civil service bureaucracies, legal systems, and the protection of property—that influence economic security. Deeper inroads also need to be made in opening up economies to external trade and in reducing fiscal deficits and government consumption.

To generate high and sustained growth in ESAF countries, an important challenge for ESAF-supported programs is to reorient government away from market-distorting activity toward increasing investment in human capital, improving the quality of institutions, and promoting open and well-functioning markets, while maintaining financial discipline.

Appendix 5.1. Details of the Analysis

Tables 5.65.14 provide supporting details regarding the data set and empirical results described in this chapter.

Table 5.6Growth Exercise: Sample Means(In percent, unless indicated otherwise)
VariableWhole SampleESAF Countries (Excluding Transition Economies)Non-ESAF CountriesAfrica ESAFCFA ESAFNon-CFA ESAFAsia ESAFWestern Hemisphere ESAF
Per capita growth0.1-0.20.2-0.5-0.9-0.42.5-1.2
Inflation
Mean104.0109.8100.820.87.728.39.8699.0
Median11.911.812.212.14.018.29.521.6
Budget balance (percent of GDP)-5.3-6.8-4.4-5.8-5.6-5.9-7.3-11.8
Change in real effective exchange rate-1.5-2.5-0.9-3.1-3.6-2.9-2.10.1
Change in terms of trade-0.5-1.0-0.2-0.70.4-1.3-1.8-1.8
Exchange rate premium
Mean134.7238.475.3106.52.7165.846.51,156.0
Median10.111.49.27.82.027.225.135.4
Broad money (percent of GDP)33.126.436.723.723.823.630.936.1
Private sector credit (percent of GDP)23.518.226.316.622.412.719.923.3
Openness (index)1100.098.9100.699.298.599.585.4110.7
Weather dummy0.50.50.50.50.50.50.50.5
War dummy0.20.20.20.10.00.20.20.3
External debt (percent of GNP)90.0127.669.1109.0105.1111.249.6308.0
External debt service ratio (percent of exports of nonfactor goods and services)23.624.723.024.218.927.317.634.4
Rescheduling dummy0.20.30.20.30.40.20.00.4
GDP in 1980 (logarithm; in PPP-adjusted prices)2,381.91,025.33,135.5852.3861.9846.81,180.51,822.0
Primary school enrollment ratios in 197073.352.884.646.541.449.454.885.3
Secondary school enrollment ratios in 197020.411.625.36.96.86.921.327.8
Life expectancy (in years)58.151.461.848.447.149.158.161.6
Population growth2.52.92.33.13.23.02.22.3
Government consumption (in percent of GDP)14.915.414.715.716.515.210.518.9
Investment (percent of GDP in constant PPP-adjusted prices)13.610.215.69.610.79.111.711.6
Investment (percent of GDP in constant prices)19.818.420.618.716.619.818.716.9
Economic security (index)24.84.55.04.64.44.74.23.9
Deflation dummy0.10.10.00.10.30.00.00.0
Source: IMF staff estimates.

A higher number indicates more openness.

A higher number indicates more security.

Source: IMF staff estimates.

A higher number indicates more openness.

A higher number indicates more security.

Table 5.7Matrix of Simple Correlations: Whole Sample1
GRTHLINFLBUDBALDREERDTOTILXPREMDEPTHPRIVYOPENINDWEATHERWARLTOTDBTTOTSRVRESCHEDLGDP80PRI 70SEC 70LLIFEPOPGGCONSINVGDPINVPPECONSECDEFL
GRTH1.00
LINFL-0.131.00
BUDBAL0.21-0.101.00
DREER0.010.12-0.011.00
DTOT10.120.000.080.031.00
LXPREM-0.150.28-0.180.14-0.011.00
DEPTH0.11-0.27-0.10-0.030.040.101.00
PRIVY0.14-0.270.13-0.050.01-0.100.601.00
OPENIND0.17-0.17-0.02-0.080.11-0.150.460.301.00
WEATHER-0.24-0.01-0.060.03-0.030.05-0.05-0.09-0.021.00
WAR-0.110.11-0.100.14-0.020.15-0.12-0.19-0.180.091.00
LTOTDBT-0.160.17-0.31-0.09-0.04-0.01-0.12-0.080.020.04-0.101.00
TOTSRV-0.070.31-0.08-0.10-0.060.04-0.19-0.01-0.28-0.01-0.020.311.00
RESCHED-0.100.09-0.04-0.11-0.04-0.04-0.10-0.05-0.070.02-0.080.300.231.00
LGDP800.080.150.200.040.040.050.350.440.15-0.07-0.13-0.200.090.001.00
PRI700.080.130.060.010.03-0.040.260.330.21-0.03-0.04-0.080.110.020.701.00
SEC700.130.060.000.030.03-0.010.530.480.21-0.09-0.05-0.020.050.010.700.711.00
LLIFE0.180.090.180.000.020.020.430.480.15-0.08-0.02-0.180.09-0.050.790.780.731.00
POPG-0.16-0.04-0.01-0.01-0.020.09-0.17-0.27-0.090.010.000.130.010.00-0.39-0.37-0.47-0.421.00
GCONS-0.10-0.10-0.280.030.020.130.290.150.400.03-0.130.17-0.13-0.04-0.010.030.08-0.060.151.00
INVGDP0.24-0.080.080.040.11-0.090.290.320.46-0.04-0.16-0.11-0.11-0.130.190.260.170.26-0.070.171.00
INVPPP0.25-0.060.210.020.100.000.330.440.33-0.07-0.09-0.24-0.06-0.140.400.390.370.51-0.220.080.691.00
ECONSEC0.18-0.070.290.010.00-0.230.180.390.25-0.08-0.26-0.170.04-0.020.390.270.200.39-0.120.010.270.341.00
DEFL-0.10-0.390.01-0.070.04-0.13-0.06-0.08-0.080.03-0.020.00-0.130.09-0.22-0.23-0.19-0.250.050.09-0.13-0.15-0.041.00
Source: IMF staff estimates.

For a description of the variables, see Table 5.14.

Source: IMF staff estimates.

For a description of the variables, see Table 5.14.

Table 5.8Determinants of Growth: Specification Search(Pooled annual data for 84 low- and middle-income developing countries, 1981–95, as available)
VariableEquation 1Equation 2Equation 3Equation 4Equation 5Equation 6Equation 7
Constant-5.51-5.49-5.46-5.77-5.66-5.06-3.50
(-1.11)(-1.14)(-1.13)(-1.20)(-1.18)(-1.00)(-0.58)
Population growth-0.83**-0.83**-0.83**-0.83**-0.83**-0.82**-0.92**
(-4.30)(-4.36)(-4.36)(-4.35)(-4.38)(-4.36)(-5.16)
Life expectancy (logarithm)5.02**5.02**5.01**5.10**5.02**4.97**4.55**
(3.37)(3.41)(3.40)(3.45)(3.42)(3.35)(2.76)
Initial period GDP (logarithm)-1.39**-1.39**-1.40**-1.40**-1.39**-1.44**-1.28**
(-4.08)(-4.09)(-4.11)(-4.13)(-4.10)(-4.08)(-3.55)
Inflation (logarithm; > 5 percent)0.73**0.73**-0.74**-0.77**-0.75**-0.75**-0.53**
(5.11)(-5.52)(-5.66)(-6.04)(-5.96)(-5.99)(-3.85)
Extra5 (kink)1.18**1.18**1.18**1.23**1.21**1.20**1.03**
(3.49)(3.49)(3.50)(3.68)(3.63)(3.59)(3.15)
Dummy for deflation-1.85**-1.84**-1.83**-1.89**-1.95**-1.96**-1.75**
(-2.51)(-2.51)(-2.49)(-2.55)(-2.64)(-2.64)(-2.39)
Budget balances0.10**0.10**0.10**0.10**0.10**0.10**0.12**
(3.73)(3.90)(3.94)(3.97)(4.00)(3.96)(4.70)
Adjusted openness0.02**0.02**0.02**0.02**0.02**0.02**0.02**
(3.21)(3.49)(3.53)(3.56)(3.42)(3.44)(2.93)
Government consumption in percent of GDP-0.07**-0.07**-0.08**-0.07**-0.07**-0.07**-0.05**
(-2.44)(-2.45)(-2.55)(-2.50)(-2.40)(-2.37)(-1.70)
Weather dummy-2.07**-2.07**-2.07**-2.08**-2.07**-2.06**-1.98**
(-7.35)(-7.38)(-7.40)(-7.40)(-7.36)(-7.33)(-7.08)
War dummy-0.83**-0.86**-0.83**-0.81**-0.75**-0.76**-0.95**
(2.21)(-2.21)(-2.23)(-2.16)(-2.01)(-2.06)(-2.56)
Lagged change in terms of trade0.03**0.03**0.03**0.03**0.03**0.03**0.03**
(2.64)(2.63)(2.64)(2.65)(2.73)(2.72)(2.63)
Economic security0.21*0.21*0.22*0.21**0.22**0.22**0.19
(1.59)(1.61)(1.67)(1.66)(1.74)(1.76)(1.46)
Change in real effective exchange rate0.010.010.010.01
(1.19)(1.19)(1.17)(1.30)
Dummy for debt rescheduling-0.41-0.41-0.41
(-1.16)(-1.17)(-1.15)
Exchange rate premium (logarithm)-0.07-0.07
(-0.24)(-0.26)
Financial depth (M3/GDP)0.0001
(0.01)
Dummy for ESAF countries-0.17
(-0.48)
Dummy for African countries-1.30**
(-1.78)
Dummy for Asian countries0.31
(-0.47)
Dummy for Latin American countries-1.98**
(-2.97)
Adjusted R20.220.220.220.220.220.220.23
F-Statistic (zero slopes)17.07**18.15**19.38**20.65**22.06**20.48**20.01**
Number of observations994994994994994994994
Jarque-Bera normality test246.7**246.7**246.5**253.1**254.7**253.0**284.3**
Source: IMF staff estimates.Note: t-Statistics based on heteroscedastic-consistent standard errors are shown in parenthesis; ** indicates significance at the 5 percent level; * indicates significance at the 10 percent level.
Source: IMF staff estimates.Note: t-Statistics based on heteroscedastic-consistent standard errors are shown in parenthesis; ** indicates significance at the 5 percent level; * indicates significance at the 10 percent level.
Table 5.9Growth Exercise: Five-Year Averages(In percent, unless indicated otherwise)
Whole SampleESAF CountriesNon-ESAF Countries
Variable1981–851986–901991–951981–851986–901991–951981–851986–901991–95
Growth-0.60.50.3-1.40.40.3-0.20.50.3
Population growth2.62.52.43.02.82.92.42.32.2
Life expectancy (logarithm)14.04.04.13.93.94.04.14.14.1
1980 real GDP (logarithm)27.47.47.46.86.86.87.87.87.8
Inflation (logarithm)2.62.52.62.72.42.42.52.52.7
Extra 5-0.1-0.3-0.2-0.1-0.3-0.2-0.1-0.3-0.2
Deflation dummy30.00.10.10.00.10.10.00.10.0
Budget balance4-6.2-5.8-3.9-8.2-7.4-4.9-5.0-4.9-3.3
Openness index597.898.2104.098.996.0101.897.299.4105.2
Government consumption415.515.114.315.315.914.715.514.714.1
Weather dummy60.50.50.50.60.50.40.50.50.5
War dummy70.20.20.20.20.10.20.20.20.2
Change in terms of trade0.7-1.6-0.6-0.5-2.70.31.4-1.0-1.2
Economic security index84.34.65.44.14.44.94.54.75.7
African ESAFAsian ESAFWestern Hemisphere ESAF
Variable1981–851986–901991–951981–851986–901991–951981–851986–901991–95
Growth-1.80.4-0.32.32.32.7-3.0-1.91.5
Population growth3.23.03.12.42.22.22.32.02.5
Life expectancy (logarithm)13.83.93.94.04.04.14.14.14.2
1980 real GDP (logarithm)26.76.76.77.07.07.07.57.57.5
Inflation (logarithm)2.62.22.32.22.21.93.74.23.0
Extra 5-0.1-0.4-0.3-0.1-0.0-0.3-0.0-0.00.0
Deflation dummy30.10.20.10.00.00.10.00.00.0
Budget balance4-6.7-6.3-4.3-8.2-7.4-6.4-16.1-13.2-6.4
Openness index5101.196.899.782.283.191.0103.7104.3124.1
Government consumption415.615.815.48.811.611.120.720.514.1
Weather dummy60.60.50.50.70.50.40.70.50.4
War dummy70.10.10.20.10.30.40.50.30.0
Change in terms of trade0.0-2.91.0-3.0-2.4-0.1-1.0-1.4-3.2
Economic security84.24.74.94.03.94.83.13.45.1
Source: IMF staff estimates.

In years.

PPP-adjusted real GDP in U.S. dollars.

In each country, takes the value 0 when inflation is above 5 percent; otherwise, the difference between inflation and 5 percent (in logarithms).

Ratio to GDP.

A higher value indicates more openness.

In each country, takes a value of 1 for abnormal weather and 0 otherwise.

In each country, takes a value of 1 for war/civil unrest and 0 otherwise.

A higher value indicates more security.

Source: IMF staff estimates.

In years.

PPP-adjusted real GDP in U.S. dollars.

In each country, takes the value 0 when inflation is above 5 percent; otherwise, the difference between inflation and 5 percent (in logarithms).

Ratio to GDP.

A higher value indicates more openness.

In each country, takes a value of 1 for abnormal weather and 0 otherwise.

In each country, takes a value of 1 for war/civil unrest and 0 otherwise.

A higher value indicates more security.

Table 5.10Growth Decompositions: Good Policies or Good Luck?1(In percentage points of real per capita GDP growth)
1981–851986–901991–95
Whole Sample
Macroeconomic policies-2.57-2.46-2.33
Inflation2-1.95-1.88-1.93
Budget balance-0.62-0.58-0.39
Structural policies1.691.772.13
Openness1.831.841.95
Size of government-1.10-1.08-1.02
Economic security0.961.021.20
Shocks3-1.18-1.19-1.16
Other identified factors42.022.302.49
Estimated growth-0.030.421.13
Unexplained factors-0.570.05-0.83
Actual growth-0.610.470.30
ESAF Countries
Macroeconomic policies-2.86-2.57-2.27
Inflation2-2.03-1.83-1.78
Budget balance-0.82-0.74-0.49
Structural policies1.661.651.94
Openness1.851.801.90
Size of government-1.09-1.13-1.05
Economic security0.900.991.09
Shocks3-1.38-1.24-1.07
Other identified factors41.972.282.34
Estimated growth-0.620.120.94
Unexplained factors-0.790.25-0.63
Actual growth-1.410.360.31
Non-ESAF Countries
Macroeconomic policies-2.41-2.40-2.35
Inflation2-1.90-1.90-2.02
Budget balance-0.50-0.50-0.33
Structural policies1.711.842.23
Openness1.821.861.97
Size of government-1.11-0.05-1.00
Economic security1.001.031.26
Shocks3-1.06-1.17-1.20
Other identified factors42.052.302.57
Estimated growth0.290.581.24
Unexplained factors-0.49-0.06-0.94
Actual growth-0.190.520.29
Source: IMF staff estimates.

For each of the three periods, growth decompositions are calculated as the product of the estimated regression coefficient (reported in Table 5.3) and the sample mean of the corresponding series for the whole sample, ESAF countries, and non-ESAF countries, respectively (see Table 5.9).

Includes all three inflation terms, including the dummy variable for deflation.

Includes dummies for weather, war, and terms of trade shocks.

Includes population growth, human capital accumulation, technological convergence, and the constant term.

Source: IMF staff estimates.

For each of the three periods, growth decompositions are calculated as the product of the estimated regression coefficient (reported in Table 5.3) and the sample mean of the corresponding series for the whole sample, ESAF countries, and non-ESAF countries, respectively (see Table 5.9).

Includes all three inflation terms, including the dummy variable for deflation.

Includes dummies for weather, war, and terms of trade shocks.

Includes population growth, human capital accumulation, technological convergence, and the constant term.

Table 5.11Combined Effects of Structural Reforms on Growth(Estimated coefficients, based on five-year averages)
FINSEC and ENTERPRFINSEC and PRICEFINSEC and EXTPRICE and ENTERPRENTERPR and EXTEXT and PRICEFINSEC, PRICE, and ENTERPRFINSEC, PRICE, and EXTFINSEC, EXT, and ENTERPRPRICE, EXT, and ENTERPR
Using uncumulated scores
10.18
(0.67)
20.59*
(1.96)
0.38
(1.34)
40.32
(1.05)
50.07
(0.22)
60.46
(1.45)
70.43
(1.48)
80.52*
(1.71)
90.21
(0.73)
100.34
(1.08)
Using cumulated scores
10.61**
(2.05)
20.80**
(2.58)
0.69**
(2.21)
40.62*
(1.92)
50.49
(1.53)
60.68**
(2.06)
70.69**
(2.21)
80.73**
(2.31)
90.61*
(1.95)
100.60*
(1.86)
Source: IMF staff estimates.Note: From a bivariate regression of real per capita GDP growth on the first principal component of different combinations of structural reforms: t-statistics based on heteroscedastic-consistent standard errors in parentheses; ** denotes significance at the 5 percent level; * denotes significance at the 10 percent level. Variables are as follows:

FINSEC = financial sector reformsENTERPR = public enterprise reforms
PRICE = pricing and marketing reformsEXT = external sector reforms

Source: IMF staff estimates.Note: From a bivariate regression of real per capita GDP growth on the first principal component of different combinations of structural reforms: t-statistics based on heteroscedastic-consistent standard errors in parentheses; ** denotes significance at the 5 percent level; * denotes significance at the 10 percent level. Variables are as follows:

FINSEC = financial sector reformsENTERPR = public enterprise reforms
PRICE = pricing and marketing reformsEXT = external sector reforms

Table 5.12Complementarities in the Effects of Structural Reforms on Growth1(Estimated coefficients based on five-year averages)
VariableUsing Uncumulated ScoresUsing Cumulated Scores
FINSEC-1.170.70**
(-0.84)(3.12)
ENTERPR-1.730.16
(-1.44)(0.69)
FINSEC * ENTERPR0.45-0.04
(1.29)(-1.66)
FINSEC4.09**0.47
(2.73)(1.53)
PRICE3.90**0.43
(3.18)(1.55)
FINSEC * PRICE-1.03**-0.05
(-2.69)(-1.45)
FINSEC1.790.69**
(1.39)(2.94)
EXT1.150.28
(1.11)(1.22)
FINSEC * EXT-0.32-0.05*
(-1.04)(-1.84)*
ENTERPR1.390.02
(1.09)(0.06)
PRICE2.52**0.70**
(2.03)(3.34)
ENTERPR * PRICE-0.49-0.04
(-1.52)(-1.37)
ENTERPR0.480.24
(0.41)(1.00)
EXT0.830.36*
(0.78)(1.94)
ENTERPR * EXT-0.16-0.03
(-0.64)(-1.40)
EXT2.20**-0.04
(2.19)(-0.15)
PRICE3.80**0.61**
(3.39)(2.51)
EXT* PRICE-0.69**-0.02
(-2.65)(-0.85)
Source: IMF staff estimates.

Results of regressions of the form: real per capita growth = constant + ß1*S1 + ß2S2 + ß3*(S1*S2); where S1 and S2 are different structural reform indices. t-Statistics based on heteroscedastic-consistent standard errors in parentheses; ** indicates significance at the 5 percent level; * indicates significance at the 10 percent level.

Source: IMF staff estimates.

Results of regressions of the form: real per capita growth = constant + ß1*S1 + ß2S2 + ß3*(S1*S2); where S1 and S2 are different structural reform indices. t-Statistics based on heteroscedastic-consistent standard errors in parentheses; ** indicates significance at the 5 percent level; * indicates significance at the 10 percent level.

Table 5.13List of Countries Included in the Panel Regressions
AlgeriaGuineaPeru
ArgentinaGuinea-BissauPhilippines
BangladeshGuyanaRwanda
BarbadosHaitiSaudi Arabia
BeninHondurasSenegal
BoliviaIndiaSierra Leone
BotswanaIndonesiaSomalia
BrazilIran, Islamic Republic ofSouth Africa
Burkina FasoJamaicaSri Lanka
BurundiJordanSudan
CameroonKenyaSuriname
Central African RepublicKoreaSwaziland
ChadLesothoSyrian Arab Republic
ChileMadagascarTanzania
ColombiaMalawiThailand
Congo, Dem. Rep. ofMalaysiaTogo
Congo, Rep. ofMaliTrinidad and Tobago
Costa RicaMaltaTunisia
Côte d’IvoireMauritaniaTurkey
Dominican RepublicMauritiusUganda
EcuadorMexicoUruguay
EgyptMoroccoVenezuela
El SalvadorMozambiqueZambia
Equatorial GuineaNepalZimbabwe
EthiopiaNicaragua
FijiNiger
GabonNigeria
Gambia, ThePakistan
GhanaPanama
GuatemalaParaguay
Table 5.14Data Description and Sources1
VariableDescriptionSource
GRTHPer capita real GDP growth measured in national currency termsWorld Bank, World Tables database
INVGDPRatio of investment to GDP in constant pricesWorld Bank, World Tables database
POPG, LFGPopulation and labor force growthWorld Bank, World Tables database
PRI70, SEC70Primary and secondary school enrollment ratios in 1970World Bank, World Tables database
LIFELife expectancy at birth (lagged 5 years)World Bank, Social Indicators of Development database
INFLAnnual average CPI inflationIMF, International Financial Statistics (Washington, various issues)
BUDBALRatio of consolidated central government budget deficits to GDPIMF, World Economic Outlook (Washington, various issues)
DREERChange in the real effective exchange rateIMF, Information Notice System database
DTOTChange in the terms of tradeWorld Bank, World Tables database
WEATHERDummy variable that equals 1 when per capita food production falls by 5 or more percentage pointsWorld Bank, World Tables database; Food and Agriculture Organization of the United Nations (Rome), production tables; and World Metereological Organization (Geneva), statistics on drought
GCONSGovernment consumption as a share of GDP in current pricesWorld Bank, World Tables database
DEPTH, PRIVYBroad money to GDP, private credit to GDPIMF, International Financial Statistics (Washington, various issues)
XPREMBlack market exchange rate premiumGlobal Currency Report (various issues); and IMF, International Financial Statistics (Washington, various issues)
TOTDBTRatio of total external debt to GNPWorld Bank, World Debt Tables database
TOTSRVRatio of total debt service to exports of goods and nonfactor servicesWorld Bank, World Debt Tables database
RESCHEDDummy variable equal to 1 for first year of rescheduling arrangement, 0 otherwiseIMF, Official Financing for Developing Countries (Washington, 1996)
WARDummy variable equal to 1 for wars/civil conflict taking place on national territory, 0 otherwiseRuth L. Sivard, World Military and Social Expenditures, 16th ed. (Leesburg, Virginia: WMSE Publications, 1996)
ECONSECIndex ranging from 1 to 10 is an average of 5 subindices on the law and order tradition, quality of the bureaucracy, corruption in government, risk of expropriation and of government repudiation of contractsIndices constructed by the Center for Institutional Reform and the Informal Sector (IRIS), with data collected by Political Risk Services and published in the International Country Risk Guide (College Park, Maryland: IRIS, University of Maryland)
OPENINDThe ratio of total trade to GDP adjusted for the size of the populationWorld Bank, World Tables database; and IMF, International Financial Statistics (Washington, various issues)
GDP80GDP in 1980 in PPP-adjusted prices in U.S. dollar termsRobert Summers and Alan Heston, “The Penn World Tables (Mark 5): An Expanded Set of International Comparisons, 1950–88,” Quarterly Journal of Economics, Vol. 106 (May 1991), pp. 327–68

Some variables are taken directly from the data source and others have been computed using original data from the source cited.

Some variables are taken directly from the data source and others have been computed using original data from the source cited.

References

    BarroRobert J.1991“Economic Growth in a Cross-Section of Countries,”Quarterly Journal of EconomicsVol. 106No. 2 (May) pp. 40743.

    • Search Google Scholar
    • Export Citation

    BarroRobert J. and Jong-WhaLee1993“International Comparisons of Educational Attainment,”Journal of Monetary EconomicsVol. 32 (December) pp. 36394.

    • Search Google Scholar
    • Export Citation

    BarroRobert J. and XavierSala-i-Martin1995Economic Growth (New York: McGraw Hill).

    BorenszteinEduardo J.JoséDe Gregorio and Jong-WhaLee1994“How Does Foreign Direct Investment Affect Economic Growth?”Working Paper 94/110 (Washington: IMF).

    • Search Google Scholar
    • Export Citation

    BosworthBarry P.S.Collins and Y.Chen1995“Accounting for Differences in Economic Growth” (unpublished; Washington: Brookings).

    • Search Google Scholar
    • Export Citation

    BrunoMichael and WilliamEasterly1995“Inflation Crises and Long-Run Growth,”NBER Working Paper 5209 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

    ClagueChristopherPhilipKeeferStephenKnack and M.Olson1995“Contract-Intensive Money: Contract Enforcement, Property Rights, and Economic Performance,”IRIS Working Paper 151 (College Park, Maryland: Center for Institutional Reform and the Informal Sector, University of Maryland).

    • Search Google Scholar
    • Export Citation

    CohenDaniel1996“The Sustainability of African Debt,”Policy Research Working Paper 1621 (Washington: World Bank).

    De GregorioJosé and PabloGuidotti1992“Financial Development and Economic Growth,”Working Paper 92/101 (Washington: IMF).

    EasterlyWilliam and RossLevine1996“Africa’s Growth Tragedy: Policies and Ethnic Divisions,” (unpublished;Washington: World Bank); see also1995“Africa’s Growth Tragedy: A Retrospecitve, 1960–89,”by same authorsPolicy Research Working Paper 1503 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    EasterlyWilliam and SergioRebelo1993“Fiscal Policy and Economic Growth,”NBER Working Paper 4499 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

    EatonJonathan1987“Public Debt Guarantees and Private Capital Flight,”World Bank Economic ReviewVol. 1 (May) pp. 37795.

    EdwardsSebastian1992“Trade Orientation, Distortions, and Growth in Developing Countries,”Journal of Development EconomicsVol. 39 (July) pp. 3157.

    • Search Google Scholar
    • Export Citation

    ElbadawiIbrahim A.B.Ndulu and N.Ndung’u1996“Debt Overhang and Economic Growth in Sub-Saharan Africa” (unpublished; Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    FischerStanley1991“Growth, Macroeconomics, and Development,”NBER Working Paper 3702 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

    FischerStanley1993“The Role of Macroeconomic Factors in Growth,”Journal of Monetary EconomicsVol. 32 (December) pp. 485512.

    • Search Google Scholar
    • Export Citation

    GhuraDaneshwar1995“Macro Policies, External Forces, and Economic Growth in Sub-Saharan Africa,”Economic Development and Cultural ChangeVol. 43 (July) pp. 75978.

    • Search Google Scholar
    • Export Citation

    GhuraDaneshwar and Michael T.Hadjimichael1995“Public Policies and Private Savings and Investment in Sub-Saharan Africa: An Empirical Investigation,”Working Paper 95/19 (Washington: IMF).

    • Search Google Scholar
    • Export Citation

    GoldsbroughDavidSharminiCooreyLouisDicks-MireauxBalazsHorvathKalpanaKochharMauroMecagniErikOfferdal and JianpingZhou1996Reinvigorating Growth in Developing Countries: Lessons from Adjustment Policies in Eight Economies Occasional Paper 139 (Washington: IMF).

    • Search Google Scholar
    • Export Citation

    GreenwoodJeremy and BoyanJovanovic1990“Financial Development, Growth and the Distribution of Income,”Journal of Political EconomyVol. 98 (October) pp. 10761107.

    • Search Google Scholar
    • Export Citation

    GrossmanGene M. and ElhanonHelpman1991“Trade, Knowledge Spillovers, and Growth,”European Economic ReviewVol. 35 pp. 51726.

    • Search Google Scholar
    • Export Citation

    HadjimichaelMichael T.DaneshwarGhuraMartinMühleisenRogerNord and E. MuratUçer1995Sub-Saharan Africa: Growth Savings and Investment 1986–93 Occasional Paper 118 (Washington: IMF).

    • Search Google Scholar
    • Export Citation

    HavrylyshynOleh1990“Trade Policy and Productivity Gains in Developing Countries: A Survey of the Literature,”World Bank Research ObserverVol. 5No. 1 (January) pp. 124.

    • Search Google Scholar
    • Export Citation

    JaegerWilliam K.1992“The Effects of Economic Policies on African Agriculture,”World Bank Discussion Paper 147 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    KingRobert G. and RossLevine1993“Finance and Growth: Schumpeter May be Right,”Quarterly Journal of EconomicsVol. 108 (August) pp. 71737.

    • Search Google Scholar
    • Export Citation

    KnackStephen and PhilipKeefer1995“Institutions and Economic Performance: Empirical Tests Using Alternative Measures of Institutions,”IRIS Working Paper 109 (College Park, Maryland: Center for Institutional Reform and the Informal Sector, University of Maryland).

    • Search Google Scholar
    • Export Citation

    KormendiRoger C. and Philip G.Meguire1985“Macroeconomic Determinants of Growth: Cross-country Evidence,”Journal of Monetary EconomicsVol. 16 (September) pp. 14163.

    • Search Google Scholar
    • Export Citation

    LeamerE.1988“Measures of Openness,” in Robert E.Baldwined.Trade Policy and Empirical Analysis (Chicago: University of Chicago Press).

    • Search Google Scholar
    • Export Citation

    LevineRoss and DavidRenelt1992“A Sensitivity Analysis of Cross-Country Growth Regressions,”American Economic ReviewVol. 82No. 4 (September) pp. 94263.

    • Search Google Scholar
    • Export Citation

    LevineRoss and SaraZervos1993“Looking at the Facts: What We Know About Policy and Growth from Cross-Country Analysis,”Policy Research Working Paper 1115 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    MaddalaG. S.1992Introduction to Econometrics2nd ed. (New York: Macmillan).

    MankiwN. GregoryDavidRomer and DavidWeil1992“A Contribution to the Empirics of Economic Growth,”Quarterly Journal of EconomicsVol. 107 (May) pp. 40737.

    • Search Google Scholar
    • Export Citation

    MauroPaulo1996“The Effects of Corruption on Growth, Investment, and Government Expenditure,”Working Paper 96/98 (Washington: IMF).

    • Search Google Scholar
    • Export Citation

    McKinnonRonald I.1973Money and Capital in Economic Development (Washington: Brookings).

    NehruVikram and AshokDhareshwar1993“A New Database on Physical Capital Stock: Sources, Methodology, and Results,”Revista de Analisis EconomicoVol. 8No. 1 (June) pp. 3759.

    • Search Google Scholar
    • Export Citation

    RomerPaul M.1986“Increasing Returns to Scale and Long-Run Growth,”Journal of Political EconomyVol. 94 (October) pp. 100237.

    • Search Google Scholar
    • Export Citation

    RomerPaul M.1990“Endogenous Technological Change,”Journal of Political EconomyVol. 98 (October) pp. S71S102.

    RoubiniNouriel and XavierSala-i-Martin1991“Financial Development, the Trade Regime, and Economic Growth,”NBER Working Paper 3876 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

    SachsJeffrey D. and Andrew M.Warner1995“Economic Reform and the Process of Global Integration,”Brookings Papers on Economic Activity: I (Washington: Brookings).

    • Search Google Scholar
    • Export Citation

    Sala-i-MartinXavier1997“I Just Ran Two Million Regressions,”American Economic Review Papers and ProceedingsVol. 87 (May) pp. 17883.

    • Search Google Scholar
    • Export Citation

    SarelMichael1996“Nonlinear Effects of Inflation on Economic Growth,”Staff PapersInternational Monetary FundVol. 43 (March) pp. 199215.

    • Search Google Scholar
    • Export Citation

    ShawEdward S.1973Financial Deepening and Economic Development (New York: Oxford University Press).

    TyboutJames R.1992“Linking Trade and Productivity: New Research Directions,”World Bank Economic ReviewVol. 6 (May) pp. 189211.

    • Search Google Scholar
    • Export Citation
1Many theoretical expositions of this theory as well as survey papers now exist on this subject. See Barro and Sala-i-Martin (1995) for a comprehensive review of growth theory.
2Influential early papers on the subject are Kormendi and Meguire (1985); Barro (1991); Mankiw, Romer, and Weil (1992); and Fischer (1991 and 1993).
3The issue of exogeneity of variables and the direction of causality is discussed further below, under “Specification of the Growth Equation and Estimation Results.”
4Easterly and Rebelo (1993) calculated measures of the total public sector balances as well as public sector investment spending to examine in more detail the links between fiscal policy and growth.
5For a survey of the theoretical and empirical literature examining the links between openness to international trade, global integration, and economic growth, see Havrylyshyn (1990) and Tybout (1992).
6It is, however, possible to have major structural distortions as well as an overvalued exchange rate, even when there is no black market premium (as is typically the case in the CFA franc zone countries and in others with similar exchange rate arrangements).
7The Levine-Renelt standard for robustness has been criticized as being “too stringent,” and by less rigorous criteria many of the hypothesized relationships would indeed be robust (Barro and Sala-i-Martin, 1995; Ghura, 1995; Bosworth, Collins, and Chen, 1995; and Sala-i-Martin, 1997). Moreover, the Levine- Renelt tests were based on cross-sectional regressions with data averaged over long periods. The use of panel data may support the finding of a larger number of robust relationships.
8A technical problem arises in using this transformation when the inflation rate is negative. Although such deflation is limited mainly to certain years in countries in the CFA franc zone, omitting these observations is not a feasible option because some of these countries are ESAF users. One intuitively appealing option would be to use the absolute value of inflation, on the grounds that what really matters for growth is price stability and that negative inflation rates are as indicative of growth-deterring distortions as are high rates. A second option, which Sarel (1996) employs, is to replace all negative inflation rates with 0.1 percent, which corresponds to the smallest positive inflation rate in his sample. In this chapter, both alternatives were tried. Since the resulting coefficient estimates were very similar, in the rest of the analysis the absolute value of inflation is used. Moreover, the inclusion of a dummy for negative inflation rates (see above) mitigates, to some extent, problems arising from using absolute values of inflation. A few studies have dealt with negative inflation observations by specifying the inflation term as the logarithm of (1 + inflation/100), but such a transformation does not go very far in addressing the evident nonlinearity of the relationship (see Chapter 6).
9Other measures of inflation nonlinearities, such as the square of the rate of inflation or the variance or standard deviation of inflation rates, were also tried but were generally not significant in the estimated equation. The major advantage of the method used here is that it is informative about more than just whether nonlinearities exist, and permits the identification of a range of inflation at which such nonlinear effects begin to occur.
10The search was conducted using half-percentage-point increments in *, and the break point was chosen at the rate that maximized the adjusted R2 of the growth regression.
11Actual debt-service payments in relation to exports of goods and services. Although the scheduled debt-service ratio would be a more accurate indicator of the debt burden, these data are available only from 1989 in the World Bank’s World Debt Tables database.
12The ECONSEC variable is an unweighted average of separate indices measuring these five factors. Data for these indices were not available for all the countries in the sample. The gaps were filled using fitted values from a regression of the ECONSEC variable on the size of bank deposits relative to broad money. Clague and others (1995) argued that the willingness of people to put money in the banking system must reflect confidence in the enforcement of contracts, which, in turn, is an important element of risk. They found that the indices measuring risk are highly positively correlated with the ratio of bank deposits to broad money.
13Other studies such as Jaeger (1992), Ghura (1995), Ghura and Hadjimichael (1995), and Hadjimichael and others (1995) used a similar proxy for weather-related shocks.
14As can be the case in samples that are not sufficiently large, the results of the Hausman specification tests appear unreliable in this sample. Also, the power of the test is heavily dependent on the quality of the instrumental variables used. Here, three regional dummy variables for Africa, Western Hemisphere, and Asia were used as instruments for the ratio of investment to GDP in constant prices (INVGDP), the dummies representing independent region-specific effects.
15The question of causality in the inflation-growth relationship is discussed more extensively in Chapter 6. Hausman tests of exogeneity give unreliable results. When a variable such as the logarithm of the inflation rate (LINFL) or the budget balance (BUDBAL)) is considered jointly with EXTRA5, the test suggests that both, or all three, are exogenous—apparently because the poor performance of the instrumental variables results in a large variance of the coefficient estimate. However, when these variables are considered individually, they appear endogenous (at a 5 percent significance level). This is true even for a variable such as the weather dummy, which a priori would be considered exogenous. See Maddala (1992) for a discussion of the properties of the Hausman test.
16Nonoverlapping five-year averages were also tried in place of annual observations. However, some of the policy variables in the specification in Table 5.3OPENIND, GCONS, and also WAR—turned out not to be statistically significant. This may be (at least in part) because the averaging of data over standard five-year periods tends to blur the turning points in the policy stance and thus to weaken the observed link between policies and growth.
17The sample comprises all 136 countries in the World Bank Economic Growth Project (used by Bruno and Easterly, 1995), minus 36 high-income countries (including all industrial countries) and transition economies, plus 5 ESAF countries not included in the original sample. Of these 105 countries, 21 countries (mainly small island economies in the Caribbean and the Pacific) had to be dropped owing to lack of data.
18The six Asian and European transition economies—Albania, Cambodia, Kyrgyz Republic, Lao P.D.R., Mongolia, and Vietnam—were excluded from this part of the analysis because of lack of data for much of the period under consideration.
19Because of the presence of a time-invariant regressor (the logarithm of initial-period GDP), it was not possible to implement a “fixed-effects” estimation procedure.
20The R2 rises to 0.35 when five-year average data are used in the specification in Table 5.3.
21The specification search was carried out by sequentially dropping the regressor with the lowest t-statistic. A fuller presentation of these results is contained in the appendix (Table 5.8; the final specification corresponds to equation 5 in the table). The finding of highly nonnormal residuals weakens the power of the F- and t-tests, but the large number of observations (over 1,000) goes some way toward mitigating this problem.
22Ghura and Hadjimichael (1995) are unusual in finding, using a sample of sub-Saharan African countries over the period 1981–93, that life expectancy at birth has a stronger empirical relationship with growth than either primary or secondary school enrollment ratios.
23The size of the estimated coefficient is in line with other empirical work (see, for instance, Barro and Sala-i-Martin, 1995).
24These results differ from those of Sarel (1996), who found that inflation rates below the point at which the kink occurs have no significant effect on growth. One reason for finding a positive relationship between inflation and growth at low inflation rates and a negative effect from deflation could be that, in this sample, a large number of countries experiencing very low or negative inflation are in the CFA franc zone. These countries also experienced low growth rates for most of the sample period under consideration, in large part because of the misalignment of their exchange rates.
25Regional dummies were also added to the final specification (results are shown in the appendix. Table 5.8). The Africa and Latin America dummies were found to be significantly negative, while the Asia dummy was positive but insignificant. An interesting result is that, with the inclusion of the regional dummies, the coefficients on the life expectancy and inflation terms become smaller, and the coefficient on economic security loses statistical significance (although its size remains more or less unchanged). This suggests that the regional dummies are, in part, acting as proxies for effects associated with inadequate investment in human capital, inflation, and poor governance.
26A dummy variable for ESAF countries was found to be statistically insignificant (see the appendix, Table 5.8).
27The sample means for each five-year period and the contributions of different factors to growth in ESAF and non-ESAF countries in the corresponding period are shown in the appendix, fables 5.9 and 5,10, respectively.
28These results are in line with the findings of several previous studies, including Ghura and Hadjimichael (1995) and Sachs and Warner (1995). As indicated in Table 5.4, other things being equal, on the basis of the underlying tendency for growth rates in poorer countries to converge to those of richer ones, during 1981–95 ESAF countries might have had annual per capita growth some 1¼ percentage points higher than that of non-ESAF countries.
29A detailed description of the methodology and data used to construct the scores is contained in Appendix 4.1 (Chapter 4).
30The results were similar when the regressions were run using annual observations, with the structural policy score for each five-year subperiod repeated for each year in the subperiod.
31Another possible reason for weak results, of course, is measurement error.
32The finding may also reflect insufficient variation in the index for public enterprise reform for it to have significant explanatory power: as noted in Chapter 4, progress in this area has generally been weak across the sample.
33With 4 individual indices, 11 different combinations were possible: 6 combinations with 1 indices each, 4 combinations with 3 indices each, and I combination with all 4 indices. The regression results were virtually the same using unweighted averages and first principal components; hence, only regressions based on the latter are reported (see the appendix, Tables 5.10 and 5.11).
34Principal-component analysis is a computational process to determine how much independence there is in the set of variables under consideration. A trivial case would be if all the variables moved proportionately; in that case, any one of the individual variables would suffice to describe the behavior of all the variables. In the more realistic case where all the variables of interest do not move proportionately, the question whether one can describe the effect of the variables by finding a linear function of them that has the highest variance. The first or largest principal component provides such a summary statistic. Another interpretation of principal-component analysis is that it permits the data to dictate the way the individual indicators are combined to form the overall index, rather than by imposing equal weights to the scores as in the case of a simple average.
35The regression took the form:growth = constant + (β1*S1 + β2*S2 + β3*(S1*S2),where S1 and S2 are different structural reform indices and where β3 > 0 suggests positive complementarities in that the effect of an increase in S1 (for example) on growth is equal to β1 + β3*S2.
36The latter would require weighing the joint negative effect on growth of undertaking two sets of reforms against the costs (to growth) of postponing one set of reforms. Analysis of sequencing issues requires more precise information on the timing of specific reforms and is beyond the scope of a broad overview such as this.
37The fragility of the statistical support probably also reflects the high degree of collinearity among many of the indicators of structural policies. This is not surprising, especially when viewed in a political economy context, since reform-minded governments tend to implement reforms on a broad front.
38These results are consistent with the findings of an earlier study on the response of growth and investment to adjustment policies (Goldsbrough and others, 1996).

    Other Resources Citing This Publication