Growth in Sub-Saharan Africa

The paper investigates empirically the determinants of economic growth for a large sample of sub-Saharan African countries during 1981-92. The results indicate that (i) an increase in private investment has a relatively large positive impact on per capita growth; (ii) growth is stimulated by public policies that lower the budget deficit in relation to GDP (without reducing government investment), reduce the rate of inflation, maintain external competitiveness, promote structural reforms, encourage human capital development, and slow population growth; and (iii) convergence of per capita income occurs after controlling for human capital development and public policies.

Abstract

The paper investigates empirically the determinants of economic growth for a large sample of sub-Saharan African countries during 1981-92. The results indicate that (i) an increase in private investment has a relatively large positive impact on per capita growth; (ii) growth is stimulated by public policies that lower the budget deficit in relation to GDP (without reducing government investment), reduce the rate of inflation, maintain external competitiveness, promote structural reforms, encourage human capital development, and slow population growth; and (iii) convergence of per capita income occurs after controlling for human capital development and public policies.

I. Introduction

Aggregate economic performance in sub-Saharan Africa during the last decade has remained unsatisfactory, in contrast to robust performance of developing countries elsewhere. Both domestic and external factors have contributed to this disappointing overall performance. The external environment, characterized by sharp declines in world commodity prices and substantial losses in the terms of trade, has been generally unfavorable. For many countries, the effects of these adverse external developments have been compounded by unfavorable weather. Also, virtually all countries in the region have been confronted with deep-rooted developmental constraints--rapid population growth, low human capital development, and inadequate infrastructure--which have constituted major impediments to private sector development and the supply response of the economies in general. In addition, ethnic conflicts, political instability, adverse security conditions, and protracted civil wars have aggravated the economic performance of several countries. Furthermore, governance concerns in several sub-Saharan African countries have been compounded by the legacy of repressive regimes in several African countries, as well as by bloated and inefficient public administrations, ineffective judicial systems, and complex administrative and institutional frameworks. Finally, inappropriate economic policies pursued by several countries have also contributed to the weak aggregate economic performance.

Nevertheless, the overall poor growth performance of sub-Saharan Africa has masked the important progress made by several countries, particularly since the mid-1980s, in lowering internal and external imbalances and addressing structural rigidities, and thus, at establishing the necessary conditions for resuming sustainable growth. On average, countries that have adopted and effectively implemented broad-based macroeconomic and structural reforms have done better than others. 1/ A recent paper by Hadjimichael and others (1995), using cross-section data during 1986-92, has demonstrated that after population growth and unfavorable weather, inappropriate macroeconomic policies were the second most important contributing factors to the poor per capita growth performance of sub-Saharan African countries. A number of other studies have provided evidence in support of the beneficial impact on growth of a stable macroeconomic environment. 2/

The current paper extends the empirical investigation of the determinants of growth in sub-Saharan Africa in three different ways. First, following Knight, Loayza, and Villanueva (1993), an extended version of the Mankiw, Romer, and Weil (1992) framework is applied to panel data for 29 countries in the region during 1981-92. 1/ Second, following the papers by Khan and Kumar (1993) and Khan and Reinhart (1990), the contribution of private and government investment to growth is investigated. Finally, the effects on growth of macroeconomic policies, structural reforms, changes in the terms of trade, human capital, the weather, and political freedom are investigated. The results indicate that growth is stimulated by public policies that lower the budget deficit in relation to GDP (without reducing government investment), reduce the rate of inflation, maintain external competitiveness, promote structural reforms, encourage human capital development, and slow population growth. Also, an increase in private investment has a relatively large positive impact on per capita growth. In addition, convergence of per capita income occurs after controlling for human capital development and public policies.

The rest of this paper is organized as follows: Section II discusses briefly some theoretical considerations in the growth literature. Section III presents the empirical model and Section IV outlines the empirical framework and summarizes the estimation results. The last section summarizes the main conclusions and draws some policy implications.

II. Some Theoretical Considerations

In the Solow-Swan growth model, steady state growth depends on technological progress and population growth, both of which are exogenous to the model; in the absence of technological progress, steady state per capita output does not grow. In this framework, an increase in the savings rate raises per capita economic growth in the short run; however, owing to diminishing returns to capital, per capita output in the long run grows at the rate of exogenously given technological progress. As such, economic policies do not affect steady state economic growth, although they can affect the level of output or its growth rate when the economy is in transition from one steady state to another. An important prediction of neoclassical growth models is that the output levels of countries with similar technologies should converge to a given level in the steady state. A number of recent papers have shown that the unconditional convergence hypothesis does not appear to be consistent with the empirical evidence. Nevertheless, support for conditional convergence is usually found when account is taken of the effects on per capita growth of the rate of investment, population growth, and public policies. 2/

The dependence of growth on exogenous technological progress in the neoclassical model, as well as the apparent inconsistency of convergence to hold with actual data have prompted investigation of alternative growth models. Endogenous growth models are able to generate a linkage between public policies and growth in the long run by assuming aggregate production functions that exhibit non-decreasing returns to scale. 1/ The remainder of this section provides a brief review of the main theoretical considerations in the current literature on growth.

1. Human capital

Recent endogenous growth models have shown that human capital accumulation can be an important source of long-term growth, either because it is a direct input into research (Romer (1990)) or because of its positive externalities (Lucas (1988), and Becker, Murphy, and Tamura (1990)). For example, Lucas (1988) proposes a model in which investment in human capital enhances the productivity of both the recipients of such capital and the society at large, thereby creating positive externalities. The decision of individuals to invest in human capital enhances technological progress. Hence, policies that enhance public and private investment in human capital affect long-run economic growth. Nevertheless, using a neoclassical framework, Mankiw, Romer, and Weil (1992) have shown that the contribution of human capital to growth is also consistent with the predictions of the Solow-Swan model.

2. Macroeconomic stability

Macroeconomic policies may affect economic growth either directly through their effect on the accumulation of factors of production, namely capital, or indirectly through their impact on the efficiency with which factors of production are used. Macroeconomic stability--reflected in low and stable inflation, sustainable budget deficits, and appropriate exchange rates--sends important signals to the private sector about the direction of economic policies and the credibility of the authorities regarding their commitment to manage the economy efficiently. Such stability, by facilitating long-term planning and investment decisions, encourages savings and private capital accumulation. The lack of macroeconomic stability, by creating an atmosphere of uncertainty, makes it difficult for economic agents to extract the correct signals from relative prices, such as the real returns to investment, and thus leads to inefficient allocation of resources (Barro (1976 and 1980)). An appropriate level of the real exchange rate, and an appropriate structure of relative prices and economic incentives in general, are key ingredients of a stable macroeconomic environment.

The effect of inflation on growth is ambiguous in the theoretical literature. According to the Tobin-Mundell effect, higher anticipated inflation leads to a lower real interest rate and causes portfolio adjustments away from real money balances toward real capital; hence, a higher anticipated inflation would be expected to raise real investment and growth. However, in the case of developing countries with underdeveloped domestic capital and financial markets, the portfolio adjustments would most likely be from real money balances to real assets, which are not usually included in private investment, or to assets denominated in foreign currency through capital flight. Thus, higher anticipated inflation in these countries would be expected to lower private investment and growth. In the cash-in-advance models (e.g., Stockman (1981)), anticipated inflation, by raising the cost of capital, lowers capital accumulation and economic growth.

Fiscal policy and the extent of government involvement in the economy have received considerable attention in the development literature. 1/ Other things being equal, a higher budget deficit will crowd out the private sector, as a result of lower access to bank credit, a higher real interest rates, and a more appreciated real exchange rate. Government investment has also been used in empirical studies as a direct proxy of the government’s contribution to capital accumulation, as well as an indicator of the adequacy of basic economic and social infrastructure. Some studies have included government consumption to allow for the concern of supply side theories that higher government spending creates expectations of future tax liabilities that in turn distort incentives and lower economic growth (e.g., Kormendi and Meguire (1985)). Endogenous growth models have shown that fiscal policy can have significant effects on economic growth in the long run. For example, in a model that assumes constant returns to scale with respect to government inputs and private capital combined but diminishing returns with respect to private capital alone, Barro (1989 and 1990) has shown that high levels of government taxation distort savings decisions, which in turn lower economic growth in the steady state.

3. Trade policy

Outward-oriented trade policies are conducive to faster growth because they promote competition, encourage learning-by-doing, improve access to trade opportunities, raise the efficiency of resource allocation, and enhance positive externalities resulting from access to improved technology (Grossman and Helpman (1989a, 1989b, and 1991), Khan (1987), Lucas (1988), and Romer (1986 and 1990)). 2/ Also, Krueger (1974) has argued that the existence of import quotas diverts productive resources to rent-seeking activities and reduces growth.

4. Structural policies

Structural and institutional reforms are essential to enhancing economic incentives and improving the allocation of resources, as well as removing the impediments to private sector development. 1/ Policies aimed at improving the efficiency of economic resources involve measures to reduce the wedges between prices and marginal costs, which typically arise from price controls, imperfect competition, subsidies and tax exemptions, distortive taxes, and exchange and trade restrictions (Khan (1987)). Policies to expand capacity include reforms aimed at raising savings and investment.

Other structural reforms include liberalization of administrative procedures for private sector activity and legal reforms. A common characteristic of countries with limited political rights and civil liberties is a lack of well-defined property rights and market-friendly legal institutions. The absence of these rights and institutions lowers the security for life and property, and as a consequence reduces the rate of accumulation and the efficiency of factors of production.

III. The Empirical Model

The paper assumes a Cobb-Douglas production function of the following form: 2/

Y=A0(ApKp)α(AhKh)β(ALL)1αβ,(1)

where Y is real output; L is labor; Kp and Kh are the physical and human capital stock, respectively; A0 is an overall index of technology and efficiency in the economy; and Ap, Ah, and AL are the physical and human capital-augmenting and labor-augmenting technology, respectively. Defining:

A=AL(A0ApαAhβ)1/1αβ,(2)

equation (1) can be rewritten as:

Y =KpαKhβ(AL)1αβ ,(3)

where the A encompasses all the factor-augmenting and the economy-wide levels of technology and efficiency; A now represents the overall labor-augmenting technology and efficiency level. 1/

Labor and labor-augmenting technology are assumed to grow according to the following functions,

L =L0ent ,(4)

and

A =A0e(gt + xθ) ,(5)

where n is the exogenous rate of growth of the labor force; t is a time index; g is the exogenous rate of technological progress; X is a vector of policy and other factors that can affect the level of technology and efficiency in the economy; and θ is a vector of coefficients related to these policy and other variables.

Let Sg and Sh be the fractions of income invested in physical and human capital, respectively; for simplicity it is assumed that both types of capital stock depreciate at the same rate δ. Thus, physical and human capital are accumulated according to the following functions:

dKpdt=SpY-δKp ,(6)

and

dKhdt=ShY-δKh .(7)

Next, let kp and kh be the stock of physical and human capital in terms of effective labor units, that is, kp = Kp/A.L and kh - Kh/A.L. Let also y be the level of output per effective unit of labor, that is, y = Y/A.L. Rewriting the production and accumulation functions in terms of quantities per effective labor unit gives:

y =kpαkhβ ,(3)
dkpdt =Spy -(n + g +δ)kp ,(6)

and

dkhdt =Shy -(n + g +δ)kh .(7)

In the steady state, the levels of physical and human capital per effective labor unit are constant. Thus, setting (6ʹ) and (7ʹ) to zero and solving the resulting equations gives:

kp* =(Sp1βShβn + g +δ)1/(1 -α -β) ,(8a)

and

kh* =(SpαSh1αn + g +δ)1/(1 -α -β) ,(8b)

Substituting (8a) and (8b) in (3ʹ) and taking natural logarithms gives the steady state output per effective labor unit:

ln(y*) = -(ξ1 -ξ)ln(n + g +δ) +(α1 -ξ)1n(Sp) +(β1 -ξ)ln(Sh) ,(9)

where ξ = (α + β).

An empirical counterpart of (9) can be obtained by taking the natural logarithm of y = Y/A.L, and substituting for A from equation (5):

1n(YL) = 1n(A0) + gt + xθ -(ξ1 -ξ)ln(n + g +δ)+(α1 -ξ)1n(Sp) + (β1 -ξ)ln(Sh) .(10)

The terms ξ/1-ξ, α/1-ξ, and β/1-ξ in the above equations are the elasticities of per capita income with respect to population growth, and the fraction of income invested in physical and human capital, respectively. This model predicts that the sum of the elasticities with respect to Sp and Sh is equal to that on (n + g + δ).

Finally, following Mankiw, Romer, and Weil (1992), the transition of actual output per effective labor unit to its steady state level is approximated by:

d1n(y)dt =λ[1n(y*) - 1n(y)] ,(11)

where λ = (n + γ + δ) (1 - ξ) is the speed of convergence, y is the actual output per effective labor unit, and the other variables are defined as before. Equation (11) implies that: 1/

1n(y) = (1 -eλt)1n(y*) +eλtln(y0) ,(12)

where yo is output per effective labor unit at time to. Subtracting yo from both sides of (12) and substituting ln(y*) from equation (10) gives:

1n(y) - 1n(y0) = (1 -eλT)[-(ξ1 -ξ)ln(n + g +δ) +(α1 -ξ)ln(Sp)+(β1 -ξ)1n(Sh) + xθ - 1n(y0) + gt + 1n(A0)] ,(13)

where T is the length of time under consideration.

An empirical counterpart of equation (13) for the i-th sub-Saharan African country considered in this study is written as follows:

YGPCi,t =η01n(Y01) +η11n(PGi,t + g +δ) +η21n(PIYi,t) +η31n(GIYi,t)+η41n(HCIi) +θ1INFLi,t +θ2BDYEi,t +θ3RERGi,t+θ4XGi,t+θ5STRUC +θ6TTGi,t +θ7DRYi,t+θ8PRi,t+ui+vt+ei,t,(14)

where YGPC is per capita real GDP growth rate; YO is a measure of initial income; PG is population growth rate; PIY and GIY are the ratios of private and government investment to GDP, respectively; HCI is an indicator of human capital development, measured alternatively by gross primary and secondary school enrollment ratios, and life expectancy at birth; INFL is the rate of inflation; BDYE is the ratio of the government budget deficit (excluding grants) to GDP; RERG is the percentage change in the real effective exchange rate; XG is the growth of export volume; STRUC is a dummy variable for countries that implemented significant structural reforms during 1986-92; TTG is the percentage change in the external terms of trade; DRY is a proxy for inadequate rainfall; PR is an index of political rights; and ui, vt, and ei, t are the country-specific, time-specific, and overall error terms, respectively. Note that the variables HCI and YO vary only across countries. Table 1 gives the definitions and sources of the variables. Following Mankiw, Romer, and Weil (1992), it is assumed that (g + δ) = 0.05. Small variations of this assumed figure do not alter the results of estimation in a significant way.

Table 1.

Definitions and Sources of Variables 1/

article image

See Table 2 for a list of countries included in this study. The data (except for the ones indicated) are from the Economic Trends in Africa (ETA) data base of the IMF.

From equation (14), the elasticities of per capita income with respect to population growth, and the fraction of income invested in public and human capital (i.e., ξ/1-ξ, α/1-ξ and β/1-ξ) are obtained from the expressions η1/-η0, η2/-η0, and η3/-η0 respectively. Also, the prediction that the sum of the elasticities on Sp and Sh is equal to that on (n + g + δ) can be tested by the null hypothesis: η1 + η2 + η3 = 0. Further, if the null hypothesis: η2 - η3 = 0 were not rejected, it would imply that the elasticities of growth with respect to the ratios of private and government investment are equal. In addition, the speed of convergence is obtained by the formula:

λ = -1n(1 + T.η0)/T(15)

IV. Empirical Results

Equation (14) is estimated with panel data for 29 countries in sub-Saharan Africa. Using annual data for the period 1981-92, four observations are constructed for each country by taking 3-year nonoverlapping averages of the variables during the subperiods 1981-83, 1984-86, 1987-89, and 1990-92; thus there are a total of 116 observations. 1/ The time period (1981-92) was chosen because data for several explanatory variables, such as private and government investment and comparable indicators of macroeconomic policies, are available only since 1980 for the group of countries considered. 2/ The choice of countries was dictated by the availability of data for the complete set of variables for each country during the full period 1981-92 (see Table 2 for a list of countries included). Given the panel nature of the data set, the error term for equation (14) accordingly has three components: ui and vt which capture country- and time-specific effects, respectively, and eit which is an error term common to all countries.

Table 2.

Period Average of the Variables by Country, 1981–92

article image

See table 1 for definitions and sources of variables.

Only sub–Saharan African countries with complete data series during 1981–92 for all the variables used in the regressions are included in this study.

Countries with average per capita growth rates greater than or equal to 1 percent during 1981–92.

CFA franc countries (CFA).

Countries with average per capita growth rates less than 1 percent but greater than or equal to 0 percent during 1981–92

Countries with average per capita growth rates less than 0 percent but greater than or equal to–1 percent during 1981–92.

Countries with average per capita growth rates below–1 percent during 1981–92.

Unweighted averages.

Non–CFA franc countries.

In order to deal with time effects, the data are processed to remove the time means from the series, and the resulting model is estimated without an intercept. In regard to the treatment of country heterogeneity, the use of the least squares dummy variables (LSDV) procedure is quite common. However, with the inclusion of time-invariant variables (YO, HCI, and STRUC) in the regression analysis, the LSDV procedure cannot be implemented because the vector of dummy variables is perfectly collinear with these variables. Instead, the country-specific effects are captured by the inclusion of country-specific information imbedded in the indicators for the level of human capital development, the stance of economic policies, changes in the terms of trade, the degree of political freedom, and a proxy for the weather. In addition, dummy variables for subgroups of countries (CFA8183, CFA8486, CFA8789, and CFA9092, and STRUC) are used to account for the possibility of fixed effects stemming from a priori information regarding country characteristics and institutional arrangements.

Table 2 gives the averages of the variables for each country during 1981-92. For convenience, the sample of countries is classified into four subgroups: high growth countries (HGCs), with per capita growth greater than or equal to 1 percent; medium-to-low growth countries (MLGCs), with per capita growth less than 1 percent but greater than or equal to 0 percent; weak growth countries, with per capita growth less than 0 percent but greater than or equal to -1 percent; and very weak growth countries, with per capita growth less than -1 percent. The data indicate that, in general, countries with relatively higher growth rates had higher investment ratios, lower population growth rates, higher primary and secondary school enrollment ratios and life expectancy at birth, lower inflation rates, lower budget deficit ratios, higher export volume growth, a somewhat more depreciated real effective exchange rate, a higher degree of political rights. However, the countries with faster growth rates did not necessarily experience a more favorable external environment (in terms of changes in their terms of trade) than those with poor growth performance. Nevertheless, these broad trends are not clear cut at the individual country level, with the data averaged over a 12-year period. For example, Lesotho had an average ratio of private investment of 37 percent during 1981-92, and yet experienced negative average growth during the same period. Also, Uganda, with a high average rate of inflation experienced positive per capita growth.

Table 3 gives the matrix of correlation coefficients between pairs of variables. A number of the conventional and policy variables are significantly correlated with per capita growth. The empirical linkage between private investment and growth is stronger than that for government investment. It is interesting to note that although the measures of human capital development--life expectancy at birth (LIFE) and primary and secondary school enrollment ratios (PRI and SEC) --are highly correlated, LIFE has a stronger empirical relationship with growth than either PRI or SEC; thus, the variable LIFE is used in the regressions. In addition, there is a significant positive correlation between increases in political rights and per capita growth. Furthermore, there is no support for unconditional convergence as the correlation coefficient between per capita growth and the initial income level is positive, albeit insignificant. Also, the simple correlations between growth, on the one hand, and inflation, the percentage change in the real effective exchange rate, and the change in the terms of trade, on the other are not significant. As regards the CFA franc countries, their significantly lower rates of inflation during 1981-92, did not translate into higher growth rates (than the other sub-Saharan African countries); in fact, this group of countries registered on average significantly lower per capita growth rates during 1987-92.

Table 3.

Matrix of Correlation Coefficients for Pairs of Variables 1/

article image

See Table 1 for definitions and sources of variables. Panel data are used to calculate these correlation coefficients. The symbols ***, ** and * beside the estimated coefficients denote statistical significance at the 0.01, 0.05, and 0.10 levels, respectively.

The regression results are summarized in Table 4. All regressions are corrected for heteroscedasticity by a feasible generalized least squares (GLS) procedure implemented in two steps. First, an ordinary least squares (OLS) procedure was used to estimate each regression equation with pooled data; the residuals from this step were used to calculate the standard deviation for each country. Second, the country-specific standard deviations were used to scale all the included variables and an OLS procedure was applied again to the pooled transformed data to obtain the feasible GLS estimators. It should be noted that under the GLS estimation procedure used in this paper, the conventional coefficient of determination (R2) loses its usual interpretation (for details, see Judge and others (1985, pp. 31-32)): Buse (1973) has suggested an alternative measure for the adjusted goodness of fit for GLS models, calculated as the proportion of weighted variation in the dependent variable explained by the regression.

Table 4.

Estimates of the Growth Equation 1/

article image

The numbers in parentheses below the estimated coefficients are the absolute values of the t–ratios. The symbols ***, **, and * beside the estimated coefficients denote statistical significance at the 0.01, 0.05, and 0.10 levels, respectively.

See Table 1 for definitions of variables used.

ADJ–RSQ is an adjusted goodness of fit for heteroscedastic models (Buse (1973)).

F1 is the statistic for the test of the null hypothesis that the joint effect of all the variables included on the right hand side of the estimated equation is zero.

Using annual data during 1981–92, for each of the 29 countries, four observations are constructed by taking 3–year nonoverlapping averages of the variables during the subperiods 1981–83, 1984–86, 1987–89, and 1990–92.

F2 is the statistic of the null hypothesis that either the sum of the coefficients for In(TIY) and In(LIFE), or the sum of the coefficients for In(PIY), In(GIY) and In(LIFE) is equal to the coefficient on In(PG+.05).

From regression (1), the hypothesis of unconditional convergence is rejected, confirming the results by Barro and Sala-i-Martin (1992), and Mankiw, Romer and Weil (1992) for a diverse group of countries; and Ghura (1995a) for sub-Saharan African countries. Regressions (2) and (3) give the estimation results of the Solow-Swan model, excluding human capital. These results imply a slow but statistically insignificant rate of convergence of about 0.5 percent per year. 1/ Also, population growth exerts a relatively large adverse impact on per capita growth. Further, the effect of investment is positive and significant, as expected. From regression (3), the impact of the private investment ratio is larger than that of the government investment ratio, confirming a similar result by Khan and Reinhart (1990). However, on the basis of an F-test, the hypothesis that the estimated coefficients on these two forms of investment are equal could not be rejected. The impact of a one-standard-deviation increase in the private investment ratio is estimated to raise growth by about 1 percentage point, and a one-standard-deviation increase in the government investment ratio is estimated to raise growth by about 0.5 percentage point. Finally, from regressions (2) and (3), which do not control for human capital, the null hypothesis (given by F2 in Table 4) of the equality of the absolute values of the coefficients of physical capital and population growth is rejected, indicating perhaps the lack of an important component of capital, namely human capital.

Regression (4) reports the results of the Solow-Swan-type growth model augmented by human capital. The coefficient on initial income is now negative and statistically significant, implying that after controlling for human capital, poorer countries tend to grow faster than the less poor ones (conditional convergence), confirming a similar result by Barro (1991) for a diverse set of countries. 2/ The estimated speed of convergence is about 1 percent a year. The measure of human capital is positively and significantly correlated with per capita growth, confirming similar results by Barro (1989), and Mankiw, Romer, and Weil (1992) for a diverse group of countries, and Ghura (1995a) for the case of sub-Saharan African countries during 1970-90. Again, population growth lowers per capita growth with an elasticity which is much larger than that reported by Mankiw, Romer, and Weil (1992), and Knight, Loayza, and Villanueva (1993). Thus, it appears that increases in population have a much larger adverse impact on per capita growth in sub-Saharan African countries than in other regions. One way to attenuate this adverse effect would be to raise investment in human capital, as implied by the significant negative correlation between human capital development and population growth (Table 3). The estimated coefficient on human capital from regression (4) is 0.055 (with a standard deviation of 0.02); the data imply that a one-standard-deviation increase in life expectancy (which corresponds to about 7 years for the data sample used in this paper) would be expected to raise, on average, per capita growth by about 0.7 percentage point a year. This result is indicative of the potential gains from improvements in human capital in the context of sub-Saharan African countries, given the existing weaknesses.

Finally, in regression (4), the null hypothesis of the equality between the absolute values of the sum of the coefficients on physical and human capital on the one hand and that on population growth on the other (given by F2 in Table 4) cannot be rejected. Nevertheless, the estimated share in total income of human and physical capital from the restricted regression is 0.89, which is substantially higher than the share of 0.67 found by Mankiw, Romer, and Weil (1992) in the context of a diverse group of countries. 1/

The results of regression (5) indicate that the policy environment matters for growth. The estimated coefficients on the budget deficit ratio and the changes in the real effective exchange rate are negative and the coefficient on export volume growth is positive, and are all highly significant. Thus, countries experienced faster growth rates (than other countries in the sample) if they had: lower budget deficit ratios; 2/ higher export volume growth rates; 3/ pursued deeper structural reforms; and faster convergence of the actual real effective exchange rates toward the respective equilibrium levels. During the period under investigation, most sub-Saharan African countries experienced large deteriorations in their terms of trade, which caused the equilibrium real exchange rate to depreciate, other things being equal. In this context, a real depreciation (that is, a decline in the real effective exchange rate) can be considered as a narrowing of the gap between actual and equilibrium real exchange rates. 1/

The coefficient on the variable STRUC has a positive and significant effect on growth, supporting the view that broadly-based structural reforms alleviate the impediments to private sector development and stimulate economic growth. The fact that the variable STRUC is significant after controlling for the effects of the other variables indicates that it is capturing the independent effects of structural reforms. However, inflation did not have a significant independent direct effect on per capita growth when included with the other policy-related variables, although its effect is negative; it is argued below that the impact of inflation was registered indirectly through its effect on the volume of investment. It must be noted that the null hypothesis that the estimated coefficients on the policy-related variables are jointly equal to zero is rejected. This result is indicative of the potential benefits of implementing a comprehensive package of policies rather than piecemeal policy actions. Broad-based adjustment policies are more likely to be self-reinforcing and durable than isolated policy reforms, thus enhancing their credibility and the potential response of the private sector. 2/

Regressions (6)-(7) investigate the effects of some additional factors on growth. The external environment seems to have exerted a statistically significant influence on growth in sub-Saharan Africa during 1981-92. The estimated effect of changes in the terms of trade is positive and significant, supporting the notion that terms of trade losses, have contributed in part to the poor growth performance of sub-Saharan African countries during 1986-92. 1/ However, the adverse effects of terms of trade losses on per capita growth seem to have been offset to some extent by declines in the real effective exchange rate. It is interesting to note that once account is taken of other determinants of growth, the effect of the variable measuring political rights (PR) is not significant, contrary to the significant effects reported by Fosu (1992) and Savvides (1995) for the case of Africa. Also, inadequate rainfall appears to have reduced per capita growth on average in sub-Saharan Africa during 1981-92.

The last regression (8) attempts, albeit in a somewhat crude way, to investigate the channels through which economic policies affect growth--through their effects on either the volume or the efficiency of investment. As noted by Kormendi and Meguire (1985), if a policy variable works mainly through the efficiency channel, the inclusion of the investment ratio in the growth equation would raise the significance of the coefficient of the policy variable, but would not change (substantially) the value of its coefficient. However, if a policy variable works mainly through the volume of investment channel, the inclusion of the investment ratio would lower the significance and magnitude of the coefficient of the policy variable. The results indicate that both channels are at work for the group of countries included in this study. The significant effect of inflation on growth in regression (8) suggests that this influence is effected through the volume channel. 2/ The effects of the other policy-related variables, however, were registered mainly through the efficiency channel. 3/

A final result of interest concerns the dummy variables for the CFA franc countries. 1/ The average per capita growth rate of the CFA franc countries was about 2 percentage points a year lower than the average for sub-Saharan Africa as a whole during 1987-89, despite the fact that they had significantly lower levels of inflation than the other countries in the region. An analysis of the data indicates that the CFA franc countries, as a group, had significantly lower average government investment rates, and measures of human capital development during 1987-89 (than the average for sub-Saharan Africa). Also, in contrast with the other countries in the region, the decline in the real effective exchange rate of the CFA franc countries during 1987-89 was very modest in comparison to the recorded reduction in their terms of trade. Since the regressions control for all of these factors, the dummy for the CFA franc countries could be capturing the effects of the low confidence by the private sector in the thrust of government policies, as well as the impact of structural rigidities.

V. Conclusions and Policy Implications

The determinants of economic growth have been widely investigated by a number of recent studies. This paper has extended the investigation to the case of 29 sub-Saharan African countries during 1981-92. The main findings of this paper can be summarized as follows: (i) there is evidence in favor of conditional convergence of per capita income; (ii) an increase in private investment has a relatively large impact on per capita growth; (iii) macroeconomic policies affect per capita growth through their effects both on the volume and on the efficiency of investment; (iv) growth is stimulated by public policies that lower the budget deficit in relation to GDP (without reducing government investment), reduce the rate of inflation, maintain external competitiveness, promote structural reforms, encourage human capital development, and lower population growth; and (v) adverse exogenous factors (deteriorations in the terms of trade and droughts) have significant negative effects on per capita growth.

Notwithstanding the significant effects of exogenous factors on sub-Saharan African growth, the results of this study indicate that growth in the region can be enhanced by policies that encourage macroeconomic stability, remove tax and other price distortions, and alleviate the impediments to private sector development through structural reforms. The variables measuring the effects of macroeconomic policies and structural reforms are jointly statistically significant in influencing growth, indicating the benefits of implementing a comprehensive package of policies rather than piecemeal policy actions. As increases in private investment stimulate growth, the government should formulate and implement appropriate policies to encourage private sector development. Broad-based adjustment policies are more likely to be self-reinforcing and durable than isolated policy reforms, thus enhancing their credibility and the potential response of the private sector. In addition, a stable macroeconomic environment would also stimulate private sector savings and raise the efficiency and volume of private investment, thus speeding up the process of achieving sustainable growth.

The results also indicate that while lowering the budget deficit is beneficial to economic growth, doing so by cutting government investment is counterproductive. Thus, alternative ways of lowering the budget deficits would be needed. As many sub-Saharan African countries are characterized by narrow tax bases, weaknesses in tax administration, and a proliferation of tax exemptions, there is significant potential for raising tax receipts by broadening the tax base, improving the tax administration, and rationalizing the tax system. Such reforms would allow an increase in government revenue and investment without necessarily raising tax rates that would tend to undermine private investment. Increased government expenditure on education and health would help raise human capital and contribute to growth, both directly and also indirectly by slowing down population growth.

Growth in Sub-Saharan Africa
Author: Mr. Dhaneshwar Ghura and Mr. Michael T. Hadjimichael