Public Policies and Private Savings and Investment in Sub-Saharan Africa
An Empirical Investigation
Author: Mr. Dhaneshwar Ghura and Mr. Michael T. Hadjimichael
Publication Date:
01 Feb 1995
eISBN:
9781451922523
Language:
English
Keywords:
WP; broad money; government investment; rate of inflation
Download PDF (2.9 MB)

This paper assesses empirically the role of public policies in stimulating private savings and investment in sub-Saharan African countries, based on data for the period 1986-92. The main findings of the analysis are as follows: (i) policies effective in stimulating private savings and investment include those that keep the rate of inflation low, reduce macroeconomic uncertainty, promote financial deepening, and lower the external debt burden; (ii) measures that promote structural reforms and reduce the budget deficit (without lowering government investment) help to raise private investment; and (iii) declines in government savings are only partially offset by increases in private savings.

Abstract

This paper assesses empirically the role of public policies in stimulating private savings and investment in sub-Saharan African countries, based on data for the period 1986-92. The main findings of the analysis are as follows: (i) policies effective in stimulating private savings and investment include those that keep the rate of inflation low, reduce macroeconomic uncertainty, promote financial deepening, and lower the external debt burden; (ii) measures that promote structural reforms and reduce the budget deficit (without lowering government investment) help to raise private investment; and (iii) declines in government savings are only partially offset by increases in private savings.

I. Introduction

The experience of sub-Saharan African countries with public policy reforms and economic performance has been extensively investigated by several recent studies. 1/ The renewed interest in economic developments in these countries has been prompted by a less than satisfactory record of performance for sub-Saharan Africa as a whole during the past two decades. A recent paper by Hadjimichael and others (1995) provides a detailed assessment of economic performance in sub-Saharan Africa during 1986-93. The paper’s analysis indicates that the overall poor performance of sub-Saharan African countries masks major differences in the performance of individual countries or country groups. The divergence in the private sector and overall economic performance among various country groups has reflected primarily the differences in their policy response to the deterioration in the terms of trade--in particular, in the progress made in promoting macroeconomic stability, improving external competitiveness and alleviating the structural and institutional impediments to private sector activity. Countries that have adopted and effectively implemented broad-based structural adjustment programs (the sustained adjusters) and countries with low macroeconomic imbalances have done better than the others. The sustained adjusters achieved positive per capita real GDP growth and a reduction in inflation during 1986-93, while countries with inappropriate policies (countries with protracted imbalances) experienced a decline in per capita incomes and an increase in inflation. The implementation of appropriate policies by the sustained adjusters contributed to higher government savings and higher private investment, and was complemented by increasing inflows of foreign assistance. At the same time, the countries with positive real per capita growth during 1986-93 experienced positive government savings, increases in government investment, and strong increases in private savings and investment. In contrast, countries with negative per capita growth recorded declines in savings and investment by both the government and the private sector. The average savings and investment ratios during 1986-93 for the countries with positive per capita growth were significantly higher than those for the countries with negative per capita growth and the average for all sub-Saharan African countries; in addition, countries with positive per capita growth financed a much larger proportion of their total investment through domestic savings than did the other African countries.

The experience of sub-Saharan African countries in recent years has demonstrated that a key factor for achieving gains in real per capita incomes is an expansion in private savings and investment. Nonetheless, savings and investment ratios for sub-Saharan African countries are significantly lower than for other developing countries and still too low to support a sustainable expansion in output and employment. For example, the ratios to GDP of total savings and investment in sub-Saharan Africa during 1986-92 amounted to 11.8 percent and 18.6 percent, respectively, as compared with 24.5 percent and 25.7 percent for developing countries in general (IMF (1994)). World Bank estimates indicate that an investment/GDP ratio of about 25 percent would be needed by sub-Saharan African countries to maintain a sustainable economic growth rate of about 6 percent. In addition, recent developments in the theory of endogenous growth have established positive long-run effects on growth stemming from increases in the investment ratio, and much of the recent empirical evidence in the growth literature has found strong, positive, significant, and robust effects of increases in investment ratios on economic growth. 1/ Thus, in the period ahead, one of the key challenges facing sub-Saharan African countries would be to raise their investment ratios in an attempt to boost economic growth and close the gaps in regional economic performance. However, given the limited role played by foreign direct investment in sub-Saharan Africa and the growing demand for the limited international foreign assistance funds, more of the resources needed to finance investment in the region would have to come from domestic sources. 2/ Accordingly, public policies would need to be directed at establishing a conducive environment for the development of the indigenous private sector; such policies will also be needed to attract foreign direct investment.

In broad terms, private savings and investment depend on a range of institutional and economic factors, including economic endowments; the levels of human capital and of per capita income; the administrative, legal and institutional framework; and the stance of financial policies. In many African countries, the expansion of private investment has traditionally been impeded by a number of structural and institutional constraints, government intervention, overvalued exchange rates and high cost structures, unstable macroeconomic conditions, and limited confidence in the direction of government policies. While public policies would not be the only determinant of private savings and investment, they certainly would have a major role to play. In this context, with a view to gauging the relative importance of public policies, the principal objective of this paper is to empirically investigate the effectiveness of public policies in stimulating private savings and investment in the context of sub-Saharan Africa. This issue has received only limited attention in available empirical studies on the experience of sub-Saharan Africa. 3/ This paper uses data from a relatively large sample of countries in the region during 1986-92 to perform the empirical analysis. The results indicate that policy measures that keep the rate of inflation at low levels, lower macroeconomic uncertainty, promote financial deepening, and reduce the external debt burden are effective in raising private savings and investment. Also, private investment can be encouraged by measures aimed at lowering fiscal deficits without lowering government investment, and at promoting structural reforms.

The layout of the rest of the paper is as follows: Section II outlines the broad trends in the behavior of savings and investment in sub-Saharan Africa during 1986-92 and highlights the problems with the data. Section III discusses the channels identified in the theoretical literature through which public policies and other factors can affect private savings and investment. Section IV outlines the empirical methodology and summarizes the estimation results. The last section summarizes the conclusions and draws some policy implications.

II. Trends in Savings and Investment

National accounts estimates of private savings and investment in the context of developing countries suffer from major weaknesses, as they are usually obtained as residuals. Typically, direct measures of private savings from household and corporate surveys are not available in such countries; moreover, in most sub-Saharan African countries available national accounts data are rudimentary and subject to significant deficiencies, and cover mainly the output side of the economy. The data used in this paper were obtained from the African Economic Trends data base, compiled by the African Department of the IMF since 1985. These data are derived from national sources where available, or on estimates by the IMF staff based on a number of indicators. To ensure consistency across countries, private investment is calculated as the difference between total gross investment and government investment; as a consequence, this measure of private investment includes investment by public enterprises for which no separate data are available. Similarly, gross national savings are obtained as the sum of gross domestic investment and the external current account balance, and domestic (or domestically generated) savings are calculated by subtracting official transfers (foreign grants) from national savings. 1/ Private savings are derived as the difference between total savings and government savings; the measurement of government savings is based on budgetary data for the central government. Thus, the estimates of private savings and investment are bound to be affected by deficiencies in the data for aggregate investment, the balance of payments, and the government budget. The data weaknesses relate mainly to problems in classifying aggregate expenditure as consumption or investment; difficulties in measuring private capital flows; and distortions in the official exchange rates. 1/ Also, the measures for savings and investment would be affected by shifts in the coverage of the government sector over time, and by differences in definitions across countries. Given these deficiencies and the associated weaknesses in testing the implications of theoretical models on the behavior of savings and investment, the results of empirical investigation need to be interpreted with caution.

In addition to looking at sub-Saharan Africa as a whole, this paper considers a number of analytically interesting subgroups of countries, based on criteria related to institutional arrangements and economic performance (Table 1). The first criterion is membership in the CFA franc zone, given the limitation that this membership poses on the use of nominal exchange rate adjustments as an instrument of policy. During 1986-92, CFA franc countries had to rely entirely on internal measures to address their adjustment needs, resulting, inter alia, from a major worsening in their external environment. The differentiated economic performance of the CFA franc countries as a group in comparison with the other African countries has had a distinct influence on the performance of sub-Saharan Africa as a whole. The second criterion relates to ex post economic performance. The countries of sub-Saharan Africa are divided into two groups, depending on whether they have attained on average positive or negative (non-positive) real per capita GDP growth during 1986-92. Under the third criterion, countries are divided into three distinct groups depending on the need for, or the implementation of, appropriate adjustment policies. A small group of sub-Saharan African countries has been characterized in recent years by relatively low internal and external imbalances (countries with low macroeconomic imbalances); broadly speaking, these countries implemented appropriate policies during 1986-92 and did not need to adopt major adjustment programs, with or without support from the IMF. The remaining countries have been classified into two other groups, depending on whether or not they adopted broadly appropriate policies under IMF-supported adjustment programs for at least three years during 1986-92 (sustained adjusters and countries with protracted macroeconomic imbalances). Admittedly, this criterion may contain some judgmental considerations, particularly as some countries were not able to maintain satisfactory performance throughout 1986-92.

Table 1.

Sub-Saharan Africa: Analytical Country Groups 1/

article image

The abbreviations are as follows: SSA: total sub-Saharan Africa; CFA: CFA franc countries; NCFA: Non-CFA franc countries; PPC: countries with positive average per capita growth during 1986-92; NPC: countries with negative average per capita growth during 1986-92; LOW: countries with low macro imbalances; IMB: countries with protracted imbalances; and SUS: sustained adjusters. Only the sub-Saharan African countries covered by the African Department of the IMF are included in this study owing to the availability of a consistent data base for this group of countries. However, Angola and Liberia are excluded owing to data limitations. Djibouti, Mauritania, Somalia, and Sudan are covered by the IMF’s Middle Eastern Department.

South Africa and Zaire are excluded from the empirical investigation of Section IV.

Table 2 summarizes the evolution of total domestic and private savings and investment for 39 countries in sub-Saharan Africa during 1986-92. Several interesting patterns can be detected from these data. 1/ Both the ratios for private savings and investment followed generally upward trends during 1986-92 for sub-Saharan Africa as a whole. However, these aggregate trends masked major differences in the behavior of private savings and investment for the various country groups. Thus, for example, the positive per capita growth countries experienced generally increasing ratios of total (and private) savings and investment during 1986-92. The opposite was true for the negative per capita growth countries and the CFA franc countries. Also, broadly speaking, the movements in savings and investment were quite closely linked for the various country groups. Countries that have maintained generally high (low) ratios of private and total investment also have maintained high (low) ratios of private and total savings. The countries with low macroeconomic imbalances recorded higher average ratios of savings and investment than did the other subgroups. Also, the positive per capita growth countries registered higher average ratios of savings and investment than did the countries with negative per capita growth. The average ratio of savings to investment was quite high for the positive per capita growth countries in comparison with that of the other subgroups.

Table 2.

Trends in Savings and Investment, 1986-92 1/

(In percent of GDP)

article image
Source: IMF, Economic Trends in Africa data base, August 1993.

See Table 1 for the countries included in each subgroup of countries. Note that South Africa and Zaire are excluded from the analytical subgroups of countries.

An objective of this paper is to empirically investigate the factors that can explain this diversity in the performance of private savings and investment among the various groups of countries in sub-Saharan Africa. To this end, an econometric investigation is carried out in Section IV; to help with the interpretation of the results of this analysis, the next section provides a review of the theoretical underpinnings of the determinants of savings and investment in developing countries.

III. Theoretical Considerations

The existing theoretical models of private investment can be classified under some broad categories, including the accelerator model, the expected profits model, the neoclassical model, and the Tobin’s Q model. 2/ However, the difficulties associated with testing the implications of these models in the context of developing countries are well known. It is often noted that certain special characteristics of developing countries make the empirical testing of these theoretical models in the context of developing economies rather difficult. 1/ As data limitations preclude the estimation of structural models, empirical studies have tended to use with success semi-reduced form frameworks to investigate the determinants of investment in developing countries. In particular, studies have examined the effects of macroeconomic policies on private investment, while controlling for the economic environment; in this context, macroeconomic policies were perceived to be among the main factors that influence the dependent variables in structural models of private investment. 2/

In addition, a number of studies have investigated the factors that affect savings in developing countries, although not many of them have specifically investigated the relationship between macroeconomic stability and savings. 3/ Most of these studies use variants of the “life-cycle” model of the behavior of household savings, developed by Ando and Modigliani (1963). The basic premise of the life-cycle model is that individuals save mainly to smooth their consumption path over time in accordance with their anticipated lifetime income. Subsequent extensions of the basic life-cycle model of savings have incorporated the idea that individuals also save for bequest motives and for unexpected expenses. 4/ According to the general life-cycle hypothesis, households’ savings depend on lifetime income, wealth, and the expected returns on savings. In such a framework, government policies that affect these variables would be expected to affect savings as well. In addition, the age structure of the population, typically measured by the ratio of the population defined as dependent (under 15 and over 64 years) to total population (age dependency ratio), is another important determinant of savings. The larger the proportion of the working age population in a country, the larger the aggregate lifetime income and aggregate savings. Also, an implication of the life-cycle hypothesis is that the rate of output growth has a positive effect on savings. 1/ In addition, countries that are at a higher stage of development can devote more resources to savings.

Thus, according to life-cycle theories, two important determinants of savings are the population’s age structure and the rate of economic growth. Nevertheless, the ability of these variables to explain the behavior of savings in developing economies has been reexamined recently (e.g., Deaton (1989), and Gersovitz (1988)). 2/ It has been suggested that certain special characteristics of these economies tend to weaken the link between the life-cycle variables and savings. First, the income level of a large proportion of households in developing economies barely surpasses subsistence income. Growth in income from these low bases tend to be consumed, regardlers of other incentives for financial savings. Second, imperfections in the financial markets coupled with the uncertain nature of household income (stemming mainly from the vagaries of weather and terms of trade shocks) in developing economies tend to encourage households to save in nonfinancial assets; under such circumstances an increase in income, if saved at all, would tend to be in the form of assets such as jewelry, livestock, and land. Third, it has been noted that the relatively short life expectancy of households in developing economies coupled with the concept of extended families prevalent in these economies, tend to weaken the relationship between age dependency and savings. Unlike in industrial countries where intra-generational transfers by households are done in order to smooth consumption during the life cycle, households in developing economies tend to depend more on inter-generational transfers within the extended family to smooth consumption over time.

The following subsections review briefly the main transmission channels through which macroeconomic policies and other factors can influence private savings and investment.

1. Macroeconomic policies and uncertainty

A stable macroeconomic environment could be defined as one where inflation is low and predictable; the exchange rate is near its equilibrium level; the government budget is well managed; the budget deficit relative to GDP is at a reasonable level (consistent with a non-increasing ratio of public debt to GDP); and the use of central bank credit to finance the budget deficit is kept at a minimal level. Macroeconomic stability sends important signals to the private sector about the direction of economic policies and the credibility of the authorities’ commitment to manage the economy efficiently. Such stability, by facilitating long-term planning and investment decisions, encourages savings and capital accumulation by the private sector.

a. Macroeconomic uncertainty

Excessive volatility in key macroeconomic variables and/or uncertainty about the outlook for these variables make it difficult and costly for economic agents to extract the correct signals from relative prices--such as the real returns to investment--and thus lead to inefficient allocation of resources (Barro (1976 and 1980)). In addition, recent papers indicate that uncertainty can have large adverse effects on investment, given the irreversible nature of investment decisions (e.g., Bernanke (1983) and Pindyck (1988)). According to these papers, since capital goods are not very mobile once installed, investors would be expected to exercise their option of waiting for new information in an uncertain environment. In such an environment, even small changes in the degree of uncertainty can have large effects on private investment.

b. Inflation

The direction of the effects of inflation on savings and investment 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 lead to higher real investment. However, in the case of developing countries with underdeveloped 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 lower private investment and savings. Also, in the context of developing countries, inflation may serve as an indicator of the authorities’ commitment to a stable macroeconomic environment. The presence of high and variable inflation rates would be expected to lower the credibility of the authorities vis-à-vis the private sector and reduce the returns on private savings and investment, thereby lowering these aggregates. In addition, in the cash-in-advance models (e.g., Stockman (1981)), anticipated inflation, by raising the cost of capital, lowers capital accumulation.

c. Fiscal policy

The extent of government involvement in the economy and the government’s ability to create an environment conducive to private sector development--through, inter alia, its tax policy and investment decisions--have received considerable attention in the development literature. 1/ The empirical literature has focused on the budget deficit as a measure of the government’s borrowing requirement and on government investment. Other things being equal, a larger budget deficit will crowd out the private sector investment as a result of lower access to bank credit and/or higher real interest rates, and a more appreciated exchange rate. Government investment is used to account for the government’s contribution to capital accumulation and the adequacy of basic economic and social infrastructure. Although larger deficits may crowd out private investment, the way in which fiscal imbalances are corrected has an influence on private capital accumulation. Reductions in unproductive government expenditure would, if viewed as permanent by the private sector, reinforce the credibility of economic policies, and thus would be expected to stimulate private investment. However, if fiscal imbalances are corrected by curtailing government investment, private investment would be reduced, given the complementarity between public and private investment. Although a large fiscal deficit may be indicative of an unstable macroeconomic policy stance, some authors (e.g., Krueger (1991) and Sala-i-Martin (1991)) have argued that large deficits could be symptomatic of more fundamental microeconomic policy distortions. In this context, inappropriate microeconomic policies, by introducing distortions in resource allocation and use, slow down economic growth for a given level of resources. If the growth rate falls below a level that is politically acceptable, governments may embark on expansionary fiscal policies in an attempt to revive economic activity, thus raising the budget deficit ratio.

A growing budget deficit is likely to be associated with declining government savings. 1/ In the theoretical literature, the effects of changes in government savings on private savings are ambiguous. According to the simple Keynesian view, an increase in government consumption increases income by a “multiplier effect.” An increase in income, in turn, raises private savings. However, in an intertemporal environment where expectations play an important role, current government fiscal actions can affect the current and future time paths of macroeconomic variables, including private savings and investment. If the private sector expects an increase in future tax liabilities, current and future private consumption and savings would change, depending on the income and substitution effects of these liabilities. The income effect of an increase in tax liabilities would lower consumption in all periods, whereas the substitution effect would encourage consumption in the current period. If the income effect were to dominate the substitution effect, consumption would fall in the current period, thus raising private savings.

In such an intertemporal framework, the “Ricardian equivalence” theory (Barro (1974)) suggests that in anticipation of an increase in future tax liabilities following a decline in contemporaneous government savings, private savings increase by an equivalent amount so that national savings remain unaffected. 1/ Thus, according to this view, changes in government deficits would not be expected to affect national savings. In fact, the Ricardian equivalence theory maintains that for a given higher level of public expenditure, a temporary switch from tax finance to bonds finance will not alter real variables in the economy. This is because forward-looking economic agents, having full knowledge of the government’s budget constraint and using the same discount rate as the government, know that at some future time taxes would have to be increased. Hence, as economic agents are assumed to be concerned about the well-being of their heirs, they raise their savings to leave bequests, so that the welfare of these heirs is not reduced when taxes actually increase at some point in the future. However, in the context of developing economies, the strict conditions required for the Ricardian equivalence theory to hold (such as the existence of perfect capital markets, the absence of liquidity constraints, the lack of uncertainty about the future course of fiscal policy, and the equality between public and private discount rates) are unlikely to be met. Thus, contrary to the Ricardian equivalence theory, private savings would most probably not offset fully an increase in fiscal deficits. 2/

d. Exchange rate policy

Exchange rate policy has received significant attention in the adjustment efforts of developing countries. An objective of a devaluation is to correct an overvaluation, thereby improving external competitiveness by inducing a real depreciation. 3/4/ However, the direction of the effects on investment of changes in the real exchange rate is ambiguous. 5/ On the one hand, a real depreciation raises the cost of imported capital goods and, since a large component of investment goods is imported in developing countries, it depresses private investment, especially investment in the nontradable goods sector. On the other hand, however, a depreciation of the real exchange rate, by raising profitability in the tradable goods sector, would be expected to stimulate private investment in that sector.

2. Structural and institutional reforms

Although macroeconomic stability is a necessary condition for sustained economic growth, it must be supplemented by structural and institutional reforms. Such reforms are necessary to enhance economic incentives and improve the allocation of resources, as well as remove the impediments to private sector development. Structural policies include those aimed at improving efficiency and resource allocation, and those aimed at expanding the productive capacity of the economy (Khan (1987)). The first category of policies includes measures to reduce the wedges between prices and marginal costs, which typically arise from price controls (including those on interest rates, exchange rates, and agricultural prices), imperfect competition, subsidies and tax exemptions, distortive taxes, and exchange and trade restrictions. In recent years, many sub-Saharan African countries have implemented, in varying degrees, a broad range of structural reforms. These measures have included the restructuring and privatization of public enterprises; the lifting of controls on retail and producer prices and on marketing arrangements for agricultural products; the liberalization of the exchange and trade system; the lifting of interest rate controls, the introduction of government financial instruments, the restructuring of commercial banks, the introduction of strengthened prudential requirements, and other financial sector reforms; the implementation of tax reforms to broaden the tax base and strengthen economic incentives and promote equity; administrative reforms to strengthen the economic management capacity of the public sector; the introduction of legal reforms; and the simplification of the administrative requirements or procedures for private sector activities. 1/

Although, in principle, structural reforms make it possible to raise output for a given level of economic resources without reducing consumption, in practice the rewards from such policies may be felt with a lag, as factors of production move slowly from one sector (or activity) to another. In addition, special interest groups that benefit from the rent opportunities provided by existing price distortions may stand in the way of implementing structural reforms.

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, including human and physical capital. The lack of political freedom, therefore, is expected to lower private investment. 1/

3. Financial intermediation

McKinnon (1973) and Shaw (1973) have emphasized the crucial role played by financial deepening in increasing the rate of domestic savings. 2/ In addition, if financial deepening contributes to an increase in the expected profitability of capital, it would also be expected to encourage investment. Efforts toward financial liberalization include the lifting of ceilings on deposit and lending rates and making them more responsive to market conditions. 3/ From the theoretical literature, the effect of a change in interest rates on savings and consumption is ambiguous because the income and substitution effects of such a change work in opposite directions. An increase in the returns to savings raises the stream of future income and wealth, and thus is expected to raise current consumption. At the same time, the higher returns on savings are expected to encourage economic agents to raise savings because postponing current consumption would imply the possibility of larger future consumption out of current income. If the substitution effect of a rise in the returns to savings dominates the income effect, then an increase in interest rates would be expected to raise savings.

Although financial liberalization is expected to have a positive effect on savings, it could be envisaged that such liberalization, by easing the liquidity constraints of households and business firms and lowering uncertainty as regards the future income stream, could potentially reduce the rate of domestic savings in the short term. This would occur especially if the easing of liquidity constraints made it less costly to finance consumption for a given level of aggregate income. However, if the liberalization efforts were also successful at raising real income (say, by raising the level of efficiency in the economy), the adverse effects on the rate of savings, if any, would be mitigated.

The empirical literature for developing countries does not provide a definitive answer to the ambiguity regarding; the effects of changes in interest rates on the rate of savings. 1/2/ The lack of a significant and robust relationship between real interest rates and savings in the context of developing countries may reflect more a measurement problem than a substantive one with the existing theory concerning the role of financial intermediation. It is well known that savings data for developing countries are not very reliable. Also, during the time period used by most existing studies, interest rates in many developing countries were controlled, and thus they adjusted very slowly relative to economic fundamentals. Although the empirical evidence suggests that interest rate policies have small effects on savings rates, maintenance of negative real interest rates for prolonged periods could lead to a flight out of financial savings.

4. External debt

A number of channels have been identified in the literature for the negative impact on private investment and savings of large ratios of external public debt to exports. Three of these channels are particularly relevant for the countries of sub-Saharan Africa. First, the resources used for servicing the debt crowd out public investment, which in turn discourages private investment, given the complementarity between these two types of investments. Second, the external debt ratio could be indicative of a “debt overhang,” whereby the presence of high debt ratios leads economic agents to anticipate future tax liabilities for its servicing (Borensztein (1990a and 1990b) and Eaton (1987)). An increasing external debt ratio could induce these agents to transfer funds abroad, thus raising the implicit domestic cost of capital. Finally, it has been argued that uncertainty as regards the future stance on economic policies in response to an uncertain debt service profile would also have deleterious effects on private capital formation.

5. Terms of trade changes

The effects of changes in the terms of trade on savings are also ambiguous in the theoretical literature. 1/ In the models proposed by Harberger (1950) and Laursen and Metzler (1950), a deterioration in the terms of trade, by lowering real income, lowers savings. Subsequent authors, using intertemporal models, have challenged the Laursen-Metzler-Harberger effect. For example, in a model proposed by Obsfeld (1982), savings actually increase with a deterioration of the terms of trade. In this model, the economy has a target level of real wealth; with a fall in real wealth following a deterioration of the terms of trade, savings are raised to maintain the targeted level of real wealth. Svensson and Razin (1983) and Persson and Svensson (1985) have noted that a change in the terms of trade has ambiguous effects on savings, depending on whether this change was of a temporary or a permanent nature, or on whether or not this change was anticipated.

6. Foreign assistance

The effects of an increased inflow of foreign assistance on private savings and investment are ambiguous. The increased availability of foreign assistance, by lowering liquidity constraints, could potentially encourage consumption, thereby lowering savings. Also, a branch of the literature on the real exchange rate, building on the “Dutch disease” effect, has noted that foreign aid can have an undesirable impact on economic performance. 2/ When a large part of foreign aid is spent in the nontraded goods sector, the ensuing upward pressure on the domestic price of nontraded goods causes the equilibrium real exchange rate to appreciate, thus lowering external competitiveness. But the resulting improvement in the profitability of the nontraded goods sector induces labor to move out of export-oriented activities into service-oriented activities. A fall in labor supply in the export-oriented sector puts upward pressure on labor costs, thus lowering the profitability of this sector. The resulting decline in external competitiveness hurts export performance and, in turn, could depress investment in the tradable goods sector. Despite the Dutch disease effect, an inflow of foreign assistance is not always undesirable. Foreign assistance could also have beneficial effects, if it helps the development of economic and social infrastructure, thereby complementing private sector activities. The extent to which foreign aid is beneficial would depend on whether the positive effects of aid inflows were to dominate their adverse effects.

IV. Empirical Framework and Results

1. Empirical framework

A regression framework is used to gauge empirically the effects of public policies on the rates of private savings and investment, using pooled data for a sample of countries in sub-Saharan Africa during 1986-92. Each variable of interest (private savings and investment) is regressed on indicators of public policies, while controlling for the effects of other variables. However, as the data sample includes pooled time-series and cross-country data, a number of adjustments need to be made to correct for problems arising from country heterogeneity, time effects, and heteroscedasticity. 1/ The estimation is carried out using a feasible generalized least squares (GLS) procedure.

The empirical specification for the private investment equation takes the form:

PIY=f(Ei,Pi,U,S),(1)

where

article image

Table 3 gives the definitions and sources of the variables used in the regressions. The economic variables (Ei) include per capita economic growth (YGPC); external debt as a ratio to GDP (DETX); and overseas development assistance as a ratio to GDP (ODAY). The policy variables (Pi) include the rate of inflation (INF); fiscal deficit as a ratio to GDP (DEFY); government investment as a ratio to GDP (GIY); broad money as a ratio to GDP (BMY); and the percentage change in the real effective exchange rate (RERG). The variables that capture the effects of uncertainty (U) are the standard deviation of the rate of inflation (INFSD), and the standard deviation of the percentage change in the real effective exchange rate (RERGSD). In addition, qualitative variables are used to capture the effects of structural and institutional reforms, and to account for the effects of certain special institutional arrangements among subgroups of countries and of progress toward political liberalization (S). In particular, dummy variables are used for the CFA franc countries (CFA); countries with protracted macroeconomic imbalances during 1986-92 (IMB); sustained adjusters during 1986-92 (SUS); 1/ and an index of political freedom and civil liberties (FREE). 2/

Table 3.

Definitions of the Variables Used in the Regressions 1/

article image

See Table 1 for a list of countries included in this paper. All data (except for the ones indicated) are from the Economic Trends in Africa (AETA) data base.

The empirical specification for the private savings equation takes a form similar to that for private investment:

PSY=g(Es,Ps,U,S),(2)

where

article image

The economic and demographic variables (Es) include real per capita GDP expressed in U.S. dollars (RYPC$); YGPC, DETX, and ODAY, which are defined above; and the dependency ratio (DEPEND). The policy variables (Ps) include the rate of inflation (INF); government savings as a ratio to GDP (GSY); and a dummy variable to account for the effects of financial deepening (FINANCE). 3/ The set of variables that measure uncertainty (U), and structural and institutional reforms (S) are the same as those used to estimate equation (1).

2. Preliminary evidence

Before proceeding with a discussion of the regression results, an examination of the correlation coefficients between pairs of variables used in the regressions is informative. Table 4 gives the sample and subsample means of the variables used in the regressions and Table 5 gives the correlation matrix. Several of these correlation coefficients are statistically significant. As expected, the ratios to GDP of private savings and investment are significant and positively correlated, with a coefficient of 0.56. Therefore, it is not surprising that a number of the variables that are significantly correlated with private savings are also significantly correlated with private investment. Thus, both the ratios of private savings and investment are positively correlated with: per capita growth, the level of real per capita GDP, and the ratios to GDP of government savings, broad money and foreign assistance; and negatively correlated with: the rate of inflation, the budget deficit as a ratio to GDP, the standard deviation of inflation, and the standard deviation of the percentage change in the real effective exchange rate. However, the correlation between private savings and investment is not spectacularly high, indicating perhaps a low average degree of financial intermediation for sub-Saharan Africa as a whole.

Table 4.

Sample and Subsample Means of the Variables Used in the Regressions 1/

article image

All means are unweighted averages during 1986-92; thus, they would differ from those given in Table 3. The symbols ***, **, and * beside the estimated means for the subsamples indicate that these means are statistically different from their relevant overall sample means at the 0.01, 0.05, and 0.10 levels of significance, respectively.

See Table 3 for definitions of variables.

The abbreviation SSA refers to sub-Saharan Africa. See Table 3 for the definitions of the other abbreviations.

Table 5.

Matrix of Correlations Coefficient for Pairs of Variables Used in the Regressions 1/

article image

See Table 3 for the definition of the variables. The symbols ***, ** and * beside the estimated coefficients denote statistical significance at the 0.01, 0.05, and 0.10 level, respectively.

As regards the analytically interesting subgroups of countries, in comparison with the sample means for all countries, the countries with positive growth had a significantly higher private investment ratio, government savings ratio, foreign assistance ratio, external debt ratio, and index of political freedom. 1/ They also had a lower budget deficit ratio and dependency ratio, and a more depreciated real effective exchange rate. The countries with protracted imbalances had a significantly lower private investment ratio, private and government savings ratios, per capita growth rate, broad money ratio, and index of political freedom. These countries also had a higher budget deficit ratio and external debt ratio, and a larger standard deviation of the percentage change in their real effective exchange rate. In addition, the sustained adjusters had a significantly higher government investment ratio, broad money ratio, external assistance ratio, and standard deviation of inflation; and a lower budget deficit ratio, real per capita GDP, and dependency ratio. The average real effective exchange rate of the sustained adjusters as a group was also more depreciated. Finally, in comparison with sub-Saharan Africa as a whole, the countries of the CFA franc zone were characterized by lower levels of government investment and savings ratios, per capita real GDP growth, inflation, the standard deviations of inflation and the percentage change in the real effective exchange rate, external debt ratio, external assistance ratio, and index of political freedom. They also had a more appreciated real effective exchange rate.

It should also be noted that certain explanatory variables (such as the rate of inflation, the standard deviation of inflation, and the standard deviation of the percentage change in the real effective exchange rate) are highly correlated, and thus, contain largely the same information; in order to avoid the problem of multicollinearity, these variables are not included together in the same regression.

3. Estimation results for private investment

The regression results for private investment are summarized in Table 6. 1/ The estimated regressions explain a large proportion of the variation in the rate of private investment, as indicated by the estimated “goodness of fit” statistic suggested by Buse (1973). 2/ Studies that have used panel data to examine the determinants of private investment have typically been successful in explaining a large fraction of the variations in the rate of private investment. 3/ It is clear from the results that macroeconomic policies and uncertainty play an important role in explaining the behavior of the rate of private investment in sub-Saharan Africa. It should be noted, however, that once account is taken of the effects of macroeconomic policies, per capita GDP growth has no independent effect on the rate of private investment, supporting a similar result by Özler and Rodrik (1992), despite the fact that the simple correlation between per capita growth and the rate of investment is positive and significant (Table 5). 4/

Table 6.

Estimates of the Private Investment Equation 1/

(Estimation method: generalized least squares)

article image

See Table 3 for definitions of variables. 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. All the regressions include dummies for the years.

Based on the coefficient of determination suggested by Buse (1973) for heteroscedastic models.

F-value 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.

Number of observations used in estimation. Owing to data limitations on the real effective exchange rate, only 32 countries are included in the regressions. Data are available for the period 1986-92; however, one observation is lost with the one-period lag.

The effects of macroeconomic uncertainty, as measured by the standard deviations of the rate of inflation and the percentage change in the real effective exchange rate, are, as expected, negative and strong, confirming similar results by Serven and Solimano (1993) for a group of developing countries, and Larrain and Vergara (1993) for East Asia. 1/ Thus, highly variable real effective exchange rates and inflation rates, which are indicative of uncertainties about the returns on investment or about the direction of the future course of economic policies, have deleterious effects on private investment. This finding confirms the theoretical predictions of irreversible investment models (e.g., Bernanke (1983)). The correlation matrix indicates that the standard deviation of the percentage change in the real effective exchange rate is positively and strongly correlated with the rate of inflation, the standard deviation of inflation, the government deficit ratio, and the external debt ratio. Thus, monetary and fiscal policies geared toward macroeconomic stabilization are expected to have beneficial effects on the rate of private investment by lowering the degree of uncertainty as regards the future returns to private investment. Furthermore, the effect of inflation is negative and strong, indicating that inflationary policies are generally detrimental to efforts toward private capital accumulation. This result tends to support the predictions of the cash-in-advance models (e.g., Stockman (1981)), and the empirical findings of Greene and Villanueva (1991), and Özler and Rodrik (1992). In the context of African countries, Oshikoya (1994) found the effects of inflation to be negative for a group of three low-income economies but positive for a group of four middle-income economies.

The estimation results provide support for the complementarity between government and private investment, confirming similar results by a number of other studies. 2/ Thus, government investment, by providing positive externalities, fosters private capital accumulation. In addition, the budget deficit ratio--a variable not typically considered by previous empirical studies--is found to have a significant negative effect on the rate of private investment, indicating a crowding-out effect (regression (1)). However, this effect is not robust to the various specifications. For example, in regressions (2) and (4), the budget deficit ratio loses its statistical significance, although in regression (2) the coefficient for the budget deficit ratio is still significant at the 5 percent level for the one-tail test. As regards regression (4), it must be noted that the measure of uncertainty (RERGSD) is highly correlated with the budget deficit ratio. Thus, the regression that includes both DEFY and RERGSD suffers partly from multicollinearity and cannot detect an independent effect of DEFY. It can be concluded with some degree of confidence, however, that increases in budget deficits may be detrimental to private capital accumulation. Also, the results for estimated coefficients for the budget deficit ratio and the government investment ratio combined imply that although lowering the deficit would be beneficial to private investment, achieving this objective by cutting government investment expenditure would be counterproductive.

The regression results confirm the beneficial effects of financial deepening on private investment, as depicted by the positive, significant, and robust effect of the ratio of broad money to GDP (BMY). The real interest rate was used as an additional variable to capture the effects of financial liberalization and deepening. This variable, however, was highly correlated with the rate of inflation (with a correlation coefficient of 0.96), indicating that during 1986-92, nominal interest rates in sub-Saharan African countries adjusted rather slowly to changes in economic fundamentals and that, on average, changes in inflation dominated movements in the real interest rates. 1/

In regressions (1) and (2) where no uncertainty variables are included, increases in the external debt ratio are found to have a significant adverse effect on private investment, supporting similar results by a number of other studies. 2/ This adverse effect could occur through a number of transmission channels, as indicated in Section III, including the debt overhang effect. It appears, however, that the “uncertainty” channel dominates the impact of an increase in the debt ratio on private investment, as suggested by the results in regression (3) and (4). Once the standard deviation of inflation or the standard deviation of the percentage change in the real effective exchange rate is included in the regression, the debt ratio loses its significance, owing in part to multicollinearity. Thus, when account is taken explicitly of the effects of uncertainty, DETX does not have an independent effect on private investment.

The level of the real exchange effective seems in all regressions to have an insignificant effect on private investment, confirming similar results by Cardoso (1993), Larrain and Vergara (1993), and Servén and Solimano (1993). This result could simply reflect the theoretical ambiguity underlying the effects of real depreciations, with a positive effect on investment in the tradable goods sector canceling the adverse effect on investment in the nontradable goods sector. This result could also reflect the fact that investment reacts favorably to real depreciations with a long lag, an effect that cannot be adequately captured by the short time span covered by the data set used in the current study. Indeed, using data covering a longer time span for Chile, Solimano (1989) has found that private investment reacts adversely to a real depreciation in the short run, but positively over the medium term.

The regression results also suggest that progress in implementing structural reforms encourage private investment. The coefficient of the dummy variable for the sustained adjusters is significantly positive in all regressions, while the coefficients for the dummy variables for the CFA franc countries and the countries with protracted macroeconomic imbalances are significantly negative and large in absolute terms. As account has already been taken of differences in the policy stance, the macroeconomic environment, and the level of uncertainty, these dummy variables are most likely capturing the effects of the differentiated progress registered by these country groups in alleviating structural and institutional impediments to private sector development. Furthermore, the coefficient of the variable indicating progress toward political freedom and institutional reforms (FREE), is positive, albeit not robust; 1/ this effect appears to be significant only when account is taken of macroeconomic uncertainty.

Finally, the regression results indicate that the effect of foreign assistance (ODAY) on private investment, an impact that has typically been ignored in the empirical literature, is nonlinear and negative at the sample mean. This somewhat puzzling result is in contrast to the simple correlation between PIY and ODAY which is positive and significant. Thus, it would appear that once account is taken of the effects of other variables, increases in ODAY have adverse effects on PIY.

4. Estimation results for private savings

The regression results for private savings (PSY) are summarized in Table 7. Again, as in the case of private investment, the estimated regressions explain a large proportion of the variation in private savings, as indicated by the estimated “goodness of fit” statistic suggested by Buse (1973). The results suggest that macroeconomic policies and uncertainty play an important role in explaining the behavior of private savings in sub-Saharan African countries. It should be noted, however, that once account is taken of the effects of macroeconomic policies, per capita GDP growth has no independent effect on private savings, although the simple correlation between per capita growth and private savings is positive and significant (Table 5). In addition, the age dependency ratio, another important determinant of savings in life-cycle models (besides income growth), has an insignificant effect on the rate of private savings. A number of studies have investigated the effects of demographic factors on savings, but the evidence to date is not definitive and remains somewhat controversial. 1/ It appears that the results are highly dependent on the sample and time period considered, the specification used, and the variables included or excluded.

Table 7.

Estimates of the Private Savings Equation 1/

(Estimation method: generalized least squares)

article image

See Table 3 for definitions of variables. 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 0.01, 0.05, and 0.10 levels, respectively. All the regressions include dummies for the years.

Based on the coefficient of determination suggested by Buse (1973) for heteroscedastic models.

F-value 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.

Number of observations used in estimation. Owing to data limitations, only 36 countries are included in regressions (1)–(3), and 31 countries are included in regression(4). Data are available for the period 1986–92; however, one observation is lost with the one–period lag.

Thus, the two important variables that explain the behavior of savings in life-cycle models have no statistically significant role to play in explaining the behavior of private savings for this sample of mostly very low income economies, confirming the views advanced by Deaton (1989). It must be noted, however, that the lack of significant effects may be due to two other factors. First, the time span considered is short; demographic variables change very slowly over time, thus making it difficult to pick a significant effect with a short time series even with cross section data in which the majority of countries have essentially similar demographic characteristics. Second, as regards the potential for simultaneity bias stemming from income growth, taking the lagged values of the latter may not have adequately removed this bias.

The effects of macroeconomic uncertainty on private savings, as measured by the standard deviations of the rate of inflation and the percentage change in the real effective exchange rate (INFSD and RERGSD), are negative and strong. Thus, highly variable real effective exchange rates and inflation rates, by raising the uncertainty about future returns on savings or about the direction of future course of economic policies, have deleterious effects on private savings. As in the case of private investment, monetary and fiscal policies that foster macroeconomic stability are expected to have beneficial effects on private savings by lowering uncertainty. The effect of inflation is negative and strong, indicating that inflationary policies are generally detrimental to private savings efforts. The papers by Aghevli and others (1990), Gupta (1987), Lahiri (1989) and Schmidt-Hebbel, Webb, and Corsetti (1992) have also investigated the effects of inflation on savings in developing countries, but their results are mixed and generally statistically insignificant.

The effect of government savings (GSY), an indicator of fiscal policy, is negative, significant, and robust. According to the Ricardian equivalence theory, a fall in government savings would be fully offset by an increase in private savings so that aggregate savings remain unchanged. For the sub-Saharan African countries considered in this paper, however, only a partial offset is found. This result could be interpreted either as evidence against full Ricardian equivalence, or indicative of partial Ricardian equivalence, or a Keynesian income effect. 1/

The estimation results provide support for the beneficial effects on private savings of financial liberalization and deepening, as depicted by the positive and significant coefficient of the dummy variable FINANCE in all regressions. This result is in sharp contrast to that presented by Schmidt-Hebbel, Webb, and Corsetti (1992), who found a significant negative relationship between household savings and the beginning-of-period ratio of broad money to income in a diverse group of developing countries; they interpret their result as evidence in favor of an increase in consumption (a decline in savings) following an easing of liquidity constraints or an increase in real wealth. In addition, the regression results confirm the adverse effects of a rise in the external debt ratio (DETX) on private savings. Several channels for such effects were indicated in Section III. Unlike for the case of private investment, it appears that the debt overhang channel has a dominating effect for the case of private savings, because DETX has a strong robust effect on private savings, even when account is taken of the effects of macroeconomic uncertainty. The residual adverse effect of DETX on private savings could be due to an anticipation of future tax liabilities in the presence of large debt ratios that encourage capital flight. 2/ Thus, to the extent that private savings are an important source of funds for private capital accumulation, private investment would also be indirectly but strongly affected by an increase in DETX. 3/

There is some indication that progress in implementing structural reforms contributes to higher private savings. As in the regressions for private investment, the coefficients for the dummy variables for CFA franc countries and the countries with protracted macroeconomic imbalances are significantly negative. However, neither the sign nor the significance level of the estimated coefficient for the dummy variable for the sustained adjusters is robust. Also, the variable indicating progress toward political freedom and institutional reforms (FREE) is not statistically significant.

Changes in the terms of trade did not have a robust effect on private savings, although this effect is positive. 1/ Also, the estimation results indicate that private savings rates are higher in countries at higher levels of development (measured by the real GDP per capita, expressed in U.S. dollars (RYPC$)). 2/ The ability of an economy to mobilize domestic savings to finance investment activities depends in part on its level of development; thus, countries that are at low levels of development can devote fewer domestic resources to finance private investment. Finally, the effect of foreign assistance (ODAY) on private savings is nonlinear and positive at the sample mean, a result that is in contrast not only to the effect of foreign assistance on private investment, but also to the effects found for aggregate savings by numerous other studies.

5. Beta coefficients

An indication of the relative importance of the independent variables in explaining the variability in private investment and savings can be obtained by the beta coefficients reported in Table 8. The variability of the real effective exchange rate exerted the largest relative impact on private investment during 1986-92, a result that lends strong support to the dominating effect of uncertainty proposed by the irreversible investment models. Uncertainty also exerted a relatively large adverse impact on private savings. Of the variables measuring the effects of policies, government investment and broad money exerted the largest relative impact on private investment. The results clearly indicate that although lowering the fiscal deficits would be beneficial to private capital accumulation, achieving this objective by cutting government investment would be largely counterproductive. Also, efforts toward financial liberalization and deepening would have a large beneficial impact on both private savings and investment, as indicated by the relatively large positive beta coefficients of BMY and FINANCE. A result that is in sharp contrast with the one found in sub-sections 3 and 4 is the impact of foreign assistance (ODAY) on both private savings and investment; the beta coefficients show a positive impact on private investment and a negative impact on private savings, in contrast to the opposite effects when evaluated at sample mean. 1/ In addition, the level of development (RYPC$) exerted the largest impact on PSY, indicating that policies that help to raise the level of real per capita income would also be expected to have a large positive impact on private savings. Furthermore, the external debt ratio (DETX) exerted a relatively large negative impact on private savings. The effects of the explanatory other variables were relatively small.

Table 8.

Estimated Beta Coefficients

article image

See Table 3 for definitions of variables.

The coefficients are from Table 6. All estimated coefficients are from regression (1), except for the coefficients for INFSD and RERGSD which are from regressions (3) and (4), respectively.

The beta coefficient of an explanatory variable X, for example, is obtained simply by multiplying the estimated coefficient of X by the standard deviation of X, and then dividing the resulting product by the standard deviation of the dependent variable.

The coefficients are from Table 7. All estimated coefficients are from regression (1), except for the coefficients for INFSD and RERGSD which are from regressions (3) and (4), respectively.