Sub-Saharan Africa's Integration in the Global Financial Markets

The paper uses a unique database covering 44 countries in sub-Saharan Africa (SSA) countries between 2000 and 2007 to study the determinants of the allocation and composition of flows across countries, as well as channels through which private capital flows could affect growth. In our sample, the degree of financial market development is an important determinant of the distribution of capital flows across countries as opposed to property rights institutions. The fairly consistent positive association between net capital flows and growth for SSA countries contrasts with the more pessimistic results of recent studies, though our data do not allow us to make conclusive inferences about a causality relationship.

Abstract

The paper uses a unique database covering 44 countries in sub-Saharan Africa (SSA) countries between 2000 and 2007 to study the determinants of the allocation and composition of flows across countries, as well as channels through which private capital flows could affect growth. In our sample, the degree of financial market development is an important determinant of the distribution of capital flows across countries as opposed to property rights institutions. The fairly consistent positive association between net capital flows and growth for SSA countries contrasts with the more pessimistic results of recent studies, though our data do not allow us to make conclusive inferences about a causality relationship.

I. Introduction

For decades, Africa has been heavily dependent on official development assistance (ODA) to finance persistent balance of payments deficits. With low incomes and generally weak economic policies, most countries in sub-Saharan Africa (SSA) were unattractive to private investors, and thus received little private capital, except for foreign direct investment (FDI) in sectors focused on natural resource extraction. This has changed in the last decade.

Abundant global liquidity combined with improved economic policies and prospects in many SSA countries led to a surge in private capital flows, with sharp increases in all forms of private capital inflows—FDI, portfolio investment, and private debt flows—to SSA (Figure 1). As a result, in 2007 for the first time ever, private capital inflows increased almost six-fold since 2000 to reach an estimated US$84 billion, double the amount of ODA to SSA. The onset of the financial crisis radically changed the picture, with a drying up and, in some cases such as South Africa, a large reversal of capital flows.

Figure 1.
Figure 1.

Private Capital Inflows to Sub-Saharan Africa, 2000-07.

Citation: IMF Working Papers 2009, 114; 10.5089/9781451872613.001.A001

Source: African Department Database and IMF staff calculations.

These developments raise important questions and policy challenges for SSA countries. How dependable are these capital flows? Have they contributed, or are they likely to contribute to, improved economic growth prospects? How can governments enhance prospects for continued capital inflows, particularly in the face of the recent global financial crisis, which have seen a sharp decline in global liquidity and a reduced global appetite for risk? To begin to answer these questions, one must first understand better the factors driving the recent surge in private capital flows to SSA.

That is the objective of the paper: to identify factors behind the composition and distribution of recent private capital flows to sub-Saharan Africa, and to begin investigating their possible contribution to growth. Building on the observation that the distribution and composition of capital inflows varies substantially across SSA countries, we analyze the impact of key country characteristics in attracting different amounts and types of private capital.

Other studies have highlighted the importance of slow-moving determinants such as the level of economic development or growth performance, de jure capital account openness and the quality of institutions. These country characteristics are also going to have both a direct and indirect impact on growth in recipient countries, along with private capital flows, making it difficult to identify the direction of causality. To complicate matters further, the empirical literature is ambiguous as to whether capital flows have a positive impact on economic growth. To partly address these issues, we use an instrumental variables approach to study the determinants of private capital flows. On the growth impact, following Rodrik and Subramanian (2009), we use a two-step approach to investigate possible channels of transmission through which capital flows could influence growth: in order to have even a second-order impact on growth, capital flows should, as a necessary condition, at least have a first-order impact on variables known to influence growth.

On the determinants side, we find that capital market development, rather than basic property rights institutions, is more closely associated with higher private capital inflows. This may be a reflection of the increasing importance of portfolio and debt flows, relative to FDI, that has been observed in the recent period in SSA.

Looking at the growth impact, we find strong positive correlations between private capital inflows and real GDP growth in the recipient countries. However, in part because of the limited time-period of our sample, we are unable to prove a causal relationship whereby inflows increase growth.

The paper is related to recent research on determinants of capital flows, and on the growth impact of capital flows. As documented below, both strands of the literature are quite well-populated, but, to our knowledge, this study is the first to analyze these issues for the whole of sub-Saharan Africa, in the context of the recent surge in global private capital flows. It exploits a unique data set for 44 sub-Saharan African countries1, including annual data on private capital inflows (broken down by FDI, portfolio, and private debt) between 2000 and 2007. Appendix I provides further details about the capital flows database and how it compares to similar, published datasets, and Appendix II presents the definitions and sources of other variables used in the empirical analysis.

The paper is organized as follows. Section II looks into recent trends in private capital flows to SSA. Section III discusses the determinants of private capital flows, and section IV investigates the association between these capital flows and growth. Section V concludes.

II. Recent trends in financial globalization in SSA

In the period of the recently-ended boom in global financial markets, inflows of private capital became the most important source of external finance for sub-Saharan Africa. The increase in private capital flows to the region was in line with that in other regions and all low-income countries (Dorsey et al., 2008, World Bank, 2007). On average, in the period between 2001 and 2007 total gross inflows to the region increased by 32 percent annually—the rate accelerating to 44 percent after 2005. Private capital flows to SSA have risen almost 6-fold since 2000, from about $15 billion to about $84 billion in 2007 (the latest year for which data is available). Recorded private capital outflows are on the whole much smaller than inflows, so that in 2007 net capital flows to sub-Saharan Africa amounted to about US$76 billion. As of 2007, private capital flows represented 10 percent of the region’s GDP, about twice the volume of official development assistance (about US$40 billion).

A confluence of factors was responsible for the surge in private capital. A significantly improved macroeconomic performance across much of the continent (Figure 2), and for some countries the implementation of the Multilateral Debt Relief Initiative (MDRI), coincided with increases in global liquidity and higher oil and commodity prices (Figure 3).

Figure 2.
Figure 2.

Global Liquidity “Push” Factors

Citation: IMF Working Papers 2009, 114; 10.5089/9781451872613.001.A001

Source: African Department Database and IMF staff calculations.
Figure 3.
Figure 3.

Domestic “Pull” Factors

Citation: IMF Working Papers 2009, 114; 10.5089/9781451872613.001.A001

Source: African Department Database and IMF staff calculations.Note: Excluding Zimbabwe.

These inflows represent a potential source of vulnerability, as countries become increasingly dependent on potentially volatile private capital flows. While the overall volatility of capital flows fell in the latter years in the sample, the volatility of FDI has actually increased. (Table 1).2 In addition, for low income SSA countries in particular, there has been a strong correlation between capital inflows and current account balances, with inflows increasing as the current account improves and decreasing as the current account deteriorates, magnifying the resulting swings in the overall balance of payments (Figure 6). On the other hand, the surge in private capital inflows has contributed to a steady increase in international reserves of SSA countries, providing these countries with increased scope to finance balance of payments deficits.

Figure 4.
Figure 4.

Recipients of Private Capital Flows

Citation: IMF Working Papers 2009, 114; 10.5089/9781451872613.001.A001

Figure 5.
Figure 5.

Top Ten Recipients of Private Capital

Citation: IMF Working Papers 2009, 114; 10.5089/9781451872613.001.A001

Note: All rankings based on averages for 2000-07.Source: African Department data base and IMF staff calculations.
Figure 6.
Figure 6.

External Vulnerability

Citation: IMF Working Papers 2009, 114; 10.5089/9781451872613.001.A001

Table 1.

Private Capital Flows: Volatility

article image
Source: IMF African Department database, 2008

III. Determinants of Private Capital Flows to SSA

This section investigates the determinants of private capital flows to sub-Saharan Africa during 2000-2007, with particular focus on the role of domestic capital market development. Capital flows to sub-Saharan Africa, particularly portfolio flows, have been concentrated in a relatively small number of countries with relatively sophisticated financial markets. In these countries capital has flown into equities where the stock market was sufficiently developed (e.g. Kenya, Nigeria), government bonds in countries that had sovereign issues (Botswana, Ghana), and short-term government securities (Uganda, Tanzania and Zambia) when few other investment vehicles open to non-residents were available. The availability of investment instruments appears to have dominated the existence of capital controls. For example, in Tanzania legal loopholes allowed international investors to purchase significant amounts of government securities although in principle foreigners were not allowed to hold government debt (IMF, 2008).3

We use a new indicator of capital market development in sub-Saharan Africa that captures the structure of capital markets. Capital markets, along with money, foreign exchange and derivatives markets, form part of a country’s “essential financial markets” (Stone et al., 2008, Lukonga, forthcoming). In particular, we consider government securities markets, both for treasury bills and treasury bonds, as well as corporate bond markets and equities markets (see Lukonga, forthcoming). We construct a simple index, which is the sum of four dummy variables taking the value of 1 if the market in question exists and is well-developed. The index covers all sub-Saharan African countries except Somalia. Using this index, simple tabulations show that the average level of financial market development is fairly low in sub-Saharan Africa, with only 10 countries or 21 percent of the sample countries having complete capital markets (Table 2).4 However, as discussed earlier, these countries are also among the largest recipients of private capital inflows in sub-Saharan Africa, along with a few oil exporters (Angola, Congo, Chad and Equatorial Guinea).

Table 2.

Capital Market Development in Sub-Saharan Africa

article image
Sources: Lukonga (forthcoming) and authors’ calculations.Notes. X denotes that the market is fairly or very well developed* denotes that the market nominally exists or is partially developedN denotes that the market does not exist.

While the role of institutional development in attracting international capital flows has been emphasized in recent studies, very few have attempted to isolate the role of financial versus basic property rights institutions. The measures of financial market development used in the literature and that would be available for low-income countries, such as ratios of money and credit to GDP, have their limitations. They measure outcomes, such as the ability of the financial (mostly banking) sector to provide transaction services (in the case of the ratios of money or broad money to GDP) or the capacity of the banking system to channel savings, rather than the extent of financial market development as measured by the availability of investment instruments.

Our empirical results, detailed below, indicate that, along with other factors such as market size (particularly for portfolio inflows) and oil exporting status (particularly for FDI flows), the index of capital market development is a significant and robust determinant of international capital flows to sub-Saharan African countries.

A. Recent Literature

The starting point of much of the recent literature on determinants of capital flows has been an attempt to explain the Lucas Paradox, or why private capital doesn’t seem to flow from rich to poor countries. Standard neoclassical theory predicts that capital should flow from capital-rich countries (with low marginal productivity of capital) to capital-scarce countries (with a high marginal product of capital). In practice however, the bulk of observed flows, particularly in the modern period, consists of flows among developed economies. Compared to the early 20th century era of financial globalization, the current wave of capital flows is characterized by massive diversification of flows between high-income economies and a relative marginalization of less-developed economies (Obstfeld and Taylor, 2004, Schularick, 2006, and Schularick and Steger, 2008).

Some authors posit that the paradox is due to country-specific fundamentals that tend to lower the marginal product of capital in developing countries, such as missing factors of production including human capital, poor economic policies, and poor institutions (Alfaro, Kalemli-Ozcan and Volosovych, 2005a and 2005b, Faria et. al, 2006, Faria and Mauro, 2004, Schularick and Steger, 2008). In this vein, a recent study shows that the extent of domestic financial market development and the quality of domestic financial institutions has played a significant role in attracting capital flows to emerging markets (IMF, 2007). The second set of explanations centers around international capital market imperfections, such as asymmetric information, sovereign risk and capital controls. For example, Portes and Rey (2005) find that informational frictions, proxied by geographic distance, affect the amount and direction of capital flows, and Lane (2004) also finds that credit market frictions are a determinant of debt flows in 1970-1995.

The studies that focus on the role of institutions in attracting foreign private capital highlight the importance of basic institutions, particularly property rights protection. Weak property rights, owing to poor quality institutions, affect entrepreneurs’ investment decisions. Risks of expropriation, armed conflict, and malfunctioning legal systems all have a potentially negative impact on the returns from those investments. Empirically, property rights institutions have been proxied by an index derived from the Kaufman, Kray and Mastruzzi World Governance Indicators—WGIs, or by an index derived from the ICRG indicators. 5 Using those proxies, better quality institutions are found to be positively and significantly correlated with foreign equity flows (Alfaro et al. 2005a, 2005b), and with a higher share of equity in a country’s total stock of external liabilities (Faria and Mauro, 2004, and Faria et al., 2006).

As mentioned earlier, few studies focus on the role played by the extent of domestic financial market development in attracting international capital flows. While the literature tends to link capital flows (financial globalization) to improvements in domestic financial systems (e.g. Chinn and Ito, 2005, 2002, Mishkin, 2006), in principle the direction of causality is not clear. Chinn and Ito (2005), and Klein (2005) find empirical evidence supporting the notion of a threshold effect: financial openness contributes to equity market development only when a threshold level of general development of legal systems and financial development is attained. However, although property rights and financial institutions could potentially affect capital flows differently, no study of the determinants of capital flows has empirically attempted to properly disentangle the effects of both.

Ju and Wei (2007, 2006) develop theoretical models showing that financial institutions and property rights institutions can have different effects on capital flows. Whereas an inefficient financial system can be bypassed by international capital flows, as they could come in the form of FDI instead of portfolio and debt flows, high expropriation risk cannot be bypassed. Conversely, the stronger domestic property rights, the more the country will benefit from international capital flows.

B. Estimation strategy and Model

Cross-country regressions have been found appropriate to investigate the fundamental (and thus slow-moving) determinants of why some countries attract more inflows, or more of a certain type of inflows, than other countries. In particular, we are interested in testing whether financial institutions have a separate and different effect on capital inflows than property rights institutions. That effect may also differ depending on the type of capital flows. Empirically, Acemoglu and Johnson (2005) show that property rights and contracting/financial market development institutions do affect economic outcomes differently. While property rights institutions have a first order effect on economic outcomes such as long-run GDP growth, investment rates, and financial depth, contracting institutions seem to mostly affect the degree of financial market development. As others, we use instrumental variables to tackle the issue of reverse causality and endogeneity associated with the measures of institutions (Acemoglu and Johnson, 2005, Alfaro, Kalemli-Ozcan and Volosovych, 2005a and 2005b, Dell’Ariccia et al., 2008, Faria et. al, 2006, Faria and Mauro, 2004).

Building on the recent literature, the reduced-form linear equation to be estimated is:

Ki=μ+αFi+βPi+χCi+ui(1)

Where Ki is log of average inflows to country i, Fi is a measure of financial institutions and Pi a measure of property rights, Ci is a set of exogenous control variables, and ui is an independent error term. α, β, and γ are coefficients capturing the impact of financial development, property rights institutions and other controls on capital flows, where α and β are of particular interest, and μ is a constant.

Given that the measures of financial and property rights institutions are endogenous, simple OLS estimates of equation (1) will be biased. As indicated in Acemoglu and Johnson (2005), the α and β coefficients may capture reverse causality, or the effect of some other omitted characteristic. Second, as financial market development and property rights are measured with error, there may be a downward attenuation bias, and, finally, if contracting and property rights institutions are correlated the effect of the type of institution that is measured with greater error will load onto the other variable. Following Acemoglu and Johnson (2005), the estimation strategy therefore relies on estimating equation (1) with two-stage least squares (2SLS), using valid and separate instruments for financial and property rights institutions. The two first stage equations will thus be of the form:

Fi=v1+δ1Li+η1Ei+χ1Ci+ε1iPi=v2+δ2Li+η2Ei+χ2Ci+ε2i(2)

Where Li and Ei are instruments for financial institutions and property rights institutions, respectively. The key exclusion restriction is that Cov(ui, Li) = Cov (ui, Ei) =0, ui being the error term in the second-stage equation (1).

C. Empirical Analysis

This study utilizes a unique data set on capital inflows to SSA, prepared by the staff of the International Monetary Fund, described in Appendix I.

Choice of variables

The following set of variables was chosen to estimate equations (1) and (2) (see Appendix II for definitions and sources. Descriptive statistics are presented in Appendix III, Table 1.

  • Financial institutions are proxied by the simple index of capital market development described earlier (Lukonga, forthcoming). The more developed capital markets, the higher capital flows.

  • Property rights institutions are proxied by the rule of law score from the World Governance Indicators (Kaufmann, Kray and Mastruzzi, 2008). Among the six governance dimensions used by Kaufmann et al, 2008, this is the closest proxy for property rights protection.6 Better property rights protection should lead to higher capital flows.

  • FollowingAcemoglu and Johnson (2005), we use British legal origin as the instrument for financial market development, taken from La Porta et al. (1999). Acemoglu, Johnson and Robinson, 2002, Glaeser and Schleifer, 2002, and La Porta et al., 1999, show that in the case of former European colonies, legal origin is a valid instrument for contracting institutions/financial market development and property rights, respectively. Legal origin is shown to be positively and significantly associated with financial institutions, with civil law countries exhibiting lower financial development than common law countries (Glaeser and Schleifer, 2002, La Porta et al. 1999).

  • The instruments for property rights institutions are the degree of ethno linguistic fragmentation and the country’s latitude, taken from La Porta et al. (1999), which have been found to be associated with worse property rights and regulations.7Acemoglu and Johnson, 2005, Acemoglu Johnson and Richardson (2002), Faria et al. 2006, and Faria and Mauro, 2004, use settler mortality in 1500 or population density in 1500 and show that they are valid instruments for property rights institutions. For sub-Saharan African countries, however, settler mortality and population density in 1500 are available only for a subset of countries (Acemoglu and Johnson, 2002).8 Given the better country coverage we have chosen to proceed with ethno linguistic fragmentation and latitude.

  • A dummy variable to control for oil-exporting countries (the dummy variable takes the value of 1 if the country is an oil exporter). Oil-exporting countries attract more FDI, other things equal.

  • A dummy variable to control for market size (the dummy variable takes the value of 1 if the country is South Africa or Nigeria). Larger and more sophisticated economies attract more capital flows.

Similar studies also include a number of additional explanatory variables, though it is more difficult both to argue that they are exogenous and to find valid instruments for them. We use these variables in our baseline regression.

  • Human capital, proxied by the share of secondary school enrollment in the population, from the WDI. This variable features in explanations of the Lucas Paradox that focus on missing factors of production (Alfaro et al., 2005a, 2005b). The expected impact of higher human capital on capital flows is expected to be positive.

  • A measures of de jure capital account openness, proxied by the Chinn-Ito measure of capital account openness, based on the IMF’s AREAER (Chinn and Ito, 2007). We use the average value of the Chinn-Ito variable for sub-Saharan African countries over 1999–2005.9 The effect of capital account openness on capital flows is expected to be positive.

  • A measure of trade openness, proxied by the sum of exports and imports over GDP. Particularly at an initial stage, capital mobility develops around greater openness to trade flows. Arguably, countries that are more open to trade may also be politically more open to greater integration with the rest of the world.

  • Macroeconomic performance, proxied here by the real GDP growth rate. In a panel study of 47 industrial and emerging countries, IMF (2007) find that expected macroeconomic performance, proxied by lagged real GDP growth, is a significant determinant of private capital inflows. IMF (2008) similarly finds that real GDP growth and lower fiscal deficits are significantly correlated with higher private capital inflows. Alfaro (2005b) uses inflation volatility as a proxy for macroeconomic performance. However, it is difficult to establish causality as inflows directly affect macroeconomic outcomes (also see below). Further, macroeconomic outcomes have been shown to be shaped by institutions, making it difficult to study the independent effect of institutional development on capital flows.

Correlations

Appendix Table III.2 shows pairwise correlations between different measures of capital flows, while Appendix Table III.3 shows the pairwise correlations corresponding to the baseline regressions, and Appendix Table III.4 focuses on the pairwise correlations of the institutional variables and the instruments.

Appendix Table III.2 indicates that, for our sample, the choice of scaling variable affects the pairwise correlations of the inflows with the independent variables. This confirms the impression from Figure 5 above, that the picture of the allocation of private capital inflows across sub-Saharan countries is very different whether one looks at overall volume of inflows, or at the size of the inflows relative to a country’s population or GDP. Whereas total inflows as a share of population or GDP will be the variable of interest when studying the economic impact of the inflows on individual countries (see Section IV below), the investigation of the factors determining the allocation of private capital flows to sub-Saharan Africa across the continent should focus on the total volume of inflows. A simple explanation is that private investors are more likely to decide on allocating specific amounts of money to various markets or projects, than to fix investment limits relative to a country’s population or GDP. Therefore, for the remainder of this section the variable of interest will be the log of private capital inflows and their components, FDI on the one hand and portfolio and debt flows on the other, averaged over 2000-2007 for each sub-Saharan African country excluding Somalia and Zimbabwe.

Looking at the potential determinants of private capital flows to SSA and their composition in Appendix Table III.2 (columns 1-4), overall inflows as well as portfolio and debt inflows are correlated with the index of financial sector development, secondary school enrollment, the market size dummy variable, and the log of initial GDP per capita. There are however interesting variations for the determinants of FDI flows, to which GDP per capita and secondary school enrollment are not significantly correlated, but the proxy for property rights, and the dummy for oil producers are. The measure of trade openness and the Chinn-Ito capital openness variable are never significantly correlated with private inflows or their components. However, capital account openness is correlated with the degree of financial sector development, as in Chinn and Ito (2005), and with secondary school enrollment.

Appendix Table III.3 illustrates the challenges associated with finding truly exogenous determinants of private capital flows and their composition, as a number of potential explanatory variables are correlated with each other. In particular, the measure of financial market development is highly correlated with the proxies for property rights institutions, as discussed above. Further, educational achievement is significantly correlated with the institutional quality variables and economic development. This is not surprising, as it can be understood as an outcome of good institutions and higher degree of economic development. This strong degree of endogeneity and the fact that we are interested in disentangling the impact of institutions on capital inflows argues that, like the variables capturing macroeconomic performance, it is probably not a useful explanatory variable given our focus on the impact of institutions per se on capital inflows. The same holds true for the proxy for economic development, log of GDP per capita in 2000 (Alfaro et al., 2005b, show that log real GDP per capita has no additional explanatory power once a measure of institutional quality is introduced).

The pairwise correlations of the potential instruments—British legal origin, population density in 1500 and ethno linguistic fragmentation—with the institutional variables and with the capital flows variables are encouraging for the instrumental variables approach (Appendix Table III.4). The potential instruments for property rights institutions are uncorrelated with the inflows variables but are significantly correlated with the property rights proxies. Although British legal origin is significantly correlated with total inflows in addition to financial institutions, it is not correlated with property rights nor its potential instruments.

Regression results

Table 3 presents the results of simple OLS estimations of various specifications of equation (1), as a benchmark. The coefficient on the financial market development index is positive and significant, and robust across all specifications. For portfolio and debt flows it is even the only variable that is significant, maybe highlighting the particular importance of capital market development for non-FDI flows. The coefficient on property rights institutions (rule of law score) is never significant and has the wrong sign. The coefficients on the controls for country size and oil exporters are large and generally significant, but, while oil exporters do receive more FDI than other countries, they do not have any particular advantage with portfolio and debt inflows. Real GDP growth is the only other variable that is significant for total inflows and FDI. However, as discussed earlier, real GDP growth is not exogenous, and, to study the impact of institutions further we use the specification in column (3) as our baseline. The log of real GDP per capita in 2000, secondary enrollment, and capital account and trade openness are never significant.10 The results obtained are similar when breaking the sample into two sub-periods, capturing the period before and after the acceleration of private capital inflows. This is not surprising considering that the determinants of inflows are fairly slowly-moving country characteristics. They are also robust to excluding South Africa and Nigeria, indicating that the results are not driven by the two largest recipients, and to excluding all oil exporters (Results available upon request). Indeed, the impact of financial market development on inflows is even larger in both cases.

Table 3.

Baseline OLS Estimation Results

article image
Note: ***,**, and * indicate significant at 1, 5, and 10 percent. t-Statistics are reported in parentheses. Robust standard errors.Source: Authors’ calculations.

Table 4 presents the results of the 2SLS estimation. The first stage equations are shown in Panel A, and the second stage in Panel B. Columns (1) and (2) report results using Ethno linguistic fractionalization as the instrument for property rights institutions, and columns (3) and (4) using latitude. Given the very low R-squared obtained in the latter case corresponding results are not shown for the components of the inflows.

Table 4.

Instrumental Variables Estimation Results

article image
Note: ***,**, and * indicate significant at 1, 5, and 10 percent. t-Statistics are reported in parentheses. Robust standard

In columns (1) and (2) the instrument for rule of law is ethnolinguistic fractionalization. In columns (3) and (4) the instrument is latitude.

Source: Authors’ calculations.

Starting with the first-stage regressions, British legal origin is shown to have a strong positive and significant impact on financial development, but not on the rule of law, while the instrument for property rights has a stronger and mostly significant impact on property rights (greater ethno linguistic fragmentation has a negative impact on rule of law, as does closeness to the Equator), than on financial institutions. Market size has a positive and significant impact on capital market development, and oil exporting status has a negative and significant impact. However, market size is negatively related to rule of law—although this is likely to be mainly because of the relatively low ratings of Nigeria and South Africa regarding property rights protection. Oil exporting status is also negatively related to rule of law, a result attributed to the rent-seeking culture that appears to take hold in resource-rich countries (Auty, 2001, Auty and Gelb, 2001, Karl, 1999, 1997).

Comparing the second-stage equations with the OLS estimates, the impact of financial institutions is overall weaker (smaller coefficients) but remains strong and significant for total inflows and FDI inflows, as well as the share of FDI in total inflows. Interestingly, the property rights variable is never significant and continues to have the “wrong” sign in most cases, except for FDI inflows and the share of FDI (column 2), perhaps indicating that non-FDI investors believe property rights are not critical for their relatively liquid investments. Oil resources is a strong determinant of FDI inflows, but not other types of flows, whereas it is market size that seems to matter for portfolio and debt inflows.

When excluding South Africa and Nigeria, or oil-rich countries, from the estimation, results are broadly similar, with the coefficient on financial institutions larger particularly in the FDI regressions. This indicates that, in non-oil sub-Saharan African countries excluding South Africa, FDI inflows are significantly larger, the more financial markets are developed. Perhaps surprisingly, property rights protection is insignificant. This may be explained by the fact that, as FDI is harder to expropriate, even countries with poor property rights protection receive FDI. Alternatively, for these relatively small countries, it may be possible for foreign investors to effectively negotiate greater property rights for their FDI than is generally available in the country for other investors.

Given the small sample size, the results must be interpreted with caution, but overall they support the importance of unbundling institutions and measuring them more precisely. Results also support the notion that, for capital flows, financial institutions as proxied by capital markets development matter more than basic property rights institutions: capital including in the form of FDI will flow less to countries with less developed financial markets, against the view presented in Albuquerque (2003) that FDI should flow more to financially constrained countries.

IV. Do private capital flows have an impact on growth in SSA?

This section undertakes a preliminary analysis of the relationship between capital inflows and growth in sub-Saharan Africa. Significant capital flows to African countries are a relatively recent phenomenon and, as noted above, have so far been restricted to a limited number of markets. Given the general inconclusiveness of the empirical evidence on the links between growth and capital flows from emerging markets that have been on the radar of international investors much longer than most African markets, it is perhaps too early to expect robust econometric results with the particular sample in this paper.

While it is easy to make a theoretical case for the positive impact of capital flows on growth, the empirical verification of this impact has proved much more challenging.11 Alternate explanations to this empirical puzzle have predicated the gains from capital flows on the threshold development of certain domestic institutions, and/or proposed that such flows generate certain indirect or collateral benefits even when their direct impact on growth is difficult to gauge. Klein (2005) finds support for a model in which institutional quality intermediates the impact of capital account liberalization on growth by affecting the extent to which savings are protected from expropriation and the premium borrowers pay for funds from abroad. In countries with better institutions, capital account openness has a significant positive impact on growth. Bekaert, Harvey, and Lundblad (2005) find that the largest growth response to equity market liberalizations accrues to countries with better legal systems, above-average financial development, and better-quality institutions in terms of investor protection and accounting standards. Using industry-level data, Prasad, Rajan, and Subramanian (2007) indicate that for countries that have above-median levels of financial development, foreign capital aids the relative growth of those industries dependent on finance. But for countries below the median for financial development, the effect of foreign capital inflows is diametrically opposite. Using firm-level data, Alfaro and Charlton (2007) find a significant relationship between entrepreneurial activity and both de jure and de facto international capital integration. Their results suggest that foreign capital may improve access to capital either directly or through improved domestic financial intermediation

Recent evidence, however, indicates that countries that grow more rapidly are those that rely less and not more on foreign capital, and in turn foreign capital tends to flow to countries that experience not high, but low productivity growth (Prasad, Rajan and Subramanian (2007); Gourinchas and Jeanne (2007)). One possible explanation for this puzzle is trade: fast-growing countries also are the ones that experienced a takeoff in their export sector (Gourinchas and Jeanne, 2007). If capital inflows lead to higher prices for nontraded goods owing to absorptive capacity constraints, and thus to overvaluation, they will have a negative impact on exports and hence growth (Kose and others, 2006; Prasad, Rajan, and Subramanian, 2007). Another explanation is that inflows from abroad that supplement domestic savings will have a limited impact in an economy where the binding constraint is the scarcity of viable investment opportunities (Rodrik and Subramanian, 2009).

We estimate the gains from financial globalization using the same cross-section sample of average flows for the 2000-07 period as in the preceding section. However, in this section we also make use of data on net flows. For most African countries the differences between net and total capital flows are not often vast, nevertheless the economic impact of such flows depends on the actual infusion of foreign capital rather than gross flows. Also, we scale net flows to the size of the economy by alternatively dividing them by population or GDP.

A. Correlation Analysis

We begin the correlation analysis with a simple mapping between economic growth and capital flows in sub-Saharan Africa, while not drawing any inferences about the direction of causation. 12 The scatter plots indicate a generally positive correlation between growth and capital flows in our sample.13 Consistent with the findings in the literature, the relationship is more pronounced when only FDI flows are considered.

Figure 7.
Figure 7.

Sub-Saharan Africa: Capital Flows and Real GDP Growth

Citation: IMF Working Papers 2009, 114; 10.5089/9781451872613.001.A001

Note: All variables are the average value of the 2000-07 period. Where normalized over population the charts exclude the outliers Equatorial Guinea and Seychelles.

Table 5 provides further evidence of the positive, significant correlation between capital flows to sub-Saharan Africa and real GDP growth in the recipient countries. This relationship however does not hold when capital flows are disaggregated—while FDI is positively correlated with growth, portfolio and debt-creating flows are generally inversely related to growth, though not significantly so. This finding is also consistent with evidence from other studies(Dell’Aricia and others, 2008)

Table 5.

Correlations: Net Capital Flows and Real Growth

article image
Note: * indicates significant at 10 percent.Source: Authors’ calculations.

B. Testing channels of transmission of private capital flows on growth

Another approach to gauging the likely impact on capital inflows on growth is to test the various channels described in the literature on financial globalization and growth. Three broad channels are identified as linking capital inflows and growth:

  • Overvaluation: to the extent that capital inflows cause currency appreciation they hurt exports and thus dampen growth. This channel is well-documented in Rodrick (2007), but also discussed in Prasad et al. (2007).

  • Savings and investment: where foreign capital supplements domestic savings it fosters growth. However, this would only be true for economies where the main constraint on growth is a lack of savings. Where the main hindrance to economic growth is the scarcity of viable investment opportunities additional foreign capital will only add to domestic distortions (Gourinchas and Jeanne, 2007; Rodrick and Subramanian, 2009).

  • Institutional development: a number of studies argue that capital inflows carry indirect (collateral) benefits for recipient countries via the development of domestic institutions (Becker and others, 2007; Kose and others, 2006), or more particularly financial sector institutions (Mishkin, 2006).

The testing methodology employed here utilizes correlations as well, via bivariate regressions. In the first stage, the bivariate model regresses capital flows on the channel of interest. If the first stage results indicate that the channel is significantly correlated with capital flows then in the second stage, the model regresses the predicted value of the channel from the first regression, i.e. that part of the channel variable correlated with capital flows on growth. If the coefficient in the second regression is significant and with the expected sign, it implies that the channel is likely to be operational for this particular sample of countries. Evidence that a particular channel could be at work is not sufficient to demonstrate the impact of private capital flows on growth, but it is a necessary first step.

Thus,

Yi=α1+βKi+u1(1)
gi=α2+ηY^i+u2(2)

Where Yi is the channel variable of interest, Ki are capital flows normalized over GDP or population, gi is the real GDP growth rate and Ŷi is the predicted value of the channel variable of interest derived from the first regression. All variables are in their level values so that sign and significance of β and η are of greater interest than their magnitude. The results are presented in Table 6 below.

Table 6.

Channels of Transmission Between Capital Flows and Growth

article image
Note: *** and ** indicate significant at 1 and 5 percent, respectivelly.Source: Authors’ calculations.

The findings indicate that in most instances capital flows are not significantly correlated with the channel variables in our sample. Except when capital flows are scaled to the size of the recipient country’s population — both savings and investment are positively and significantly related to capital flows in this instance.14 This implies that, for SSA countries, the impact of private capital flows on growth could be partly intermediated through the impact of net capital flows on domestic savings and investment. Another implication would also be that African economies are not necessarily investment-constrained.15

In an unexpected outcome the institutional variable is negatively related to growth, a result that holds even when oil-exporters are excluded from the sample (not shown).

C. How do private Capital Flows Affect Growth?

Going beyond bivariate correlations, we first present a simple cross-country multivariate regression model, making use of a averages for the 2000-07 period. 16 Although this method is not without shortcomings, the limited time coverage of our dataset and year-on-year volatility dampens the usefulness of panel estimation. However, our analysis is subject to the considerable limitations of the ordinary least squares estimator in growth models — especially those arising from endogenous variables.

We fit a simple ordinary least squares specification to estimate the relationship between capital flows and growth.

gi=α+βKi+ηXi+ui

Where gi is real GDP growth, Ki are capital flows normalized over GDP or population, Xi is a set of control variables, and ui is an independent error term.

In our choice of control variables we are guided by the “channels” analysis from the preceding section, since our interest in primarily in the impact of capital flows on growth. Thus, our multivariate model is essentially a simultaneous test of the correlations between of the various channel variables, (plus capital flows directly), and growth. As discussed above, other factors shown in the literature on financial globalization to influence growth include savings and investment, exchange rate overvaluation, and institutions. The exchange rate overvaluation measure is constructed as the ratio of the nominal exchange rate to the purchasing power parity exchange rate.17 The measure of institutions used here is the number of procedures to enforce a contract (Contracts) from the World Bank’s Doing Business database. We also control for level of economic development by including the log of initial per capita real GDP growth. The results are documented in Table 7.

Table 7.

Estimation Results: Linear Growth Model with Capital Flows

article image
Note: ***, ** and * indicate significant at 1, 5 and 10 percent, respectivelly. t-statistics in parenthesis.

The results indicate that capital flows are positively and significantly (in all but one instance) associated with growth. Initial per capita GDP comes in with a negative and significant coefficient in all specifications indicating support for growth convergence. Domestic savings and investment are the other significant right-hand side variables that also come in with the expected sign. Contract enforcement, the proxy for institutional development, is not significant in any of the specifications. Overvaluation has the expected sign but is never significant. Estimating the equation with measures of net FDI instead of total net private capital yields similar results, with an even stronger association between the FDI variable and growth.

Aside from the positive correlations and partial evidence of a possible through the savings-investment channel, our analysis is rendered incomplete by the short time period since these inflows began. While the limited number of observations in our sample required us to focus on average growth and inflows over the period 2000-2007, most of the inflows came late in this period. This is an issue which requires further study, as additional observations (years) become available. As described in Section II, the observed rise in private capital flows and growth during the sample period is an important necessary condition, in particular because other studies even fail to uncover such an association for different country groups and/or sample periods. However, it is also far from sufficient for making inferences about causality.

V. Conclusion

As a result of the global financial crisis and recession, private capital flows to all regions of the world have declined. Even when the international economy recovers and financial markets return to health, it is likely that investors will remain, for some time, more risk-averse than they were previously. At the same time, donor flows to low-income countries, including those in sub-Saharan Africa, are likely to be adversely impacted by these global developments.

Our results were inconclusive regarding the impact of private capital flows on growth in SSA countries. Additional research will be needed before a definitive statement can be made in this regard, once additional observations are available. However, most SSA countries are likely to remain dependant on financial inflows in some form to help finance current account deficits—deficits that will be the counterpart of development-oriented investments in these economies. As donor flows are likely to be restricted in the coming years, due to the fiscal problems the traditional donor countries find themselves confronting, SSA countries will need private capital flows to finance these current account deficits. The alternative would require smaller or no current account deficits, meaning reduced development-oriented investments in infrastructure and other key areas, with adverse consequences for growth potential.

All this means that policy makers in SSA will need to pay increasing attention to efforts to attract private capital to their countries. Our results indicate that one of the most important things policy makers can do to attract private capital is to strengthen domestic capital markets. Deep and well-functioning money, foreign exchange, equity and (where relevant) derivative markets are key to attracting private capital. Governments in SSA countries would be well-advised to redouble their efforts to develop and strengthen these markets, to put their countries in a position to attract private capital in the future. At the same time, an important of the current financial crisis should also be to put in place strong monitoring, regulatory and supervisory frameworks, to avoid a buildup of balance-sheet vulnerabilities.

Appendix I. Capital Flows Data

The information on private capital flows is based on detailed balance of payments data obtained from AFR country desk economists in late 2007 for 44 sub-Saharan African countries. Data for 2000-2006 are actual, and data for 2007 are preliminary estimates. The data distinguishes between FDI, portfolio (debt and equity) and private debt-creating flows, and between inflows and outflows. The definition of capital inflows adopted for this study differs from other papers using balance of payments data as a measure of capital flows (see Appendix Box I.1 for precise definitions used in the paper).

Balance of payments recording of capital flows

From the perspective of the financial account one usually thinks of changes in liabilities as positive (inflows) and changes in assets as negative (outflows). In practice, both changes in liabilities and assets are reported as net of any disinvestment and consequently both can have any sign. In the BOP accounts, an increase (decrease) in liabilities to nonresidents is entered as positive (negative) while an increase (decrease) in net foreign assets held by residents is entered as negative (positive). If a resident sells an asset held abroad the transaction is recorded as a reduction in net foreign assets held by residents, bringing an inflow of capital. For this study, capital inflows (outflows) are taken to be the sum of increases (decreases) in liabilities to non-residents and of decreases (increases) in net foreign assets held by residents in a year. This allows to capture all transactions that give rise to a inflow of capital into a country on the liability and asset side, and avoids negative “inflows”. Net flows of capital are calculated as the sum of the inflows (positive sign) and outflows (negative sign).

This data differs from other, published data sources with respect to its timeliness, coverage, and level of detail:

  • The 2000-2007 period covers the most recent episode of private capital inflows to sub-Saharan Africa, which took off around 2005-2006.

  • It covers 43 sub-Saharan African countries.

  • It allows to distinguish between different types of private capital flows (FDI, portfolio flows and debt-creating flows). In contrast with other studies (e.g. Faria and Mauro, 2004) that add together FDI and portfolio equity flows, we keep them separate and group portfolio debt and equity (either category is too small to stand on its own). We are interested in the determinants of various types of flows, particularly as regards the development of financial markets, and straight FDI doesn’t require any particular type of financial market infrastructure. However, portfolio debt and equity investment require the existence of bond and equity markets.

Other data sources that cover all sub-Saharan African countries include the IMF’s WEO, IFS, Balance of payments data, and the World Bank’s global development finance data. Another widely used data base is the Lane-Milesi-Ferretti data, but the most recent updated version goes until 2005, and covers only a sub-set SSA countries. Another group of studies use current account data as a proxy for capital flows (Prasad et al., 2007; Gourinchas and Jeanne, 2007) – which have the advantage of providing wide coverage in terms of years and countries but make it impossible to disaggregate between official and private, transfers and different types of capital flows.

The Fund’s World Economic Outlook (WEO) database and the World Bank’s Global Development Finance (GDF) database provide the most comprehensive coverage of capital flows to sub-Saharan countries. These data sets have few missing countries and few missing observations for the countries that are in the data sets. However, the comprehensiveness of the WEO and the GDF databases is achieved at a price as each data set relies on Fund and Bank staff estimates to fill in the lacunae left from official statistics (Dorsey et al., 2008).

Other data sets, such as the Fund’s International Financial Statistics and the Balance of Payments Statistics databases rely on official reporting without the use of staff estimates to fill in the gaps. The more detailed and comprehensive coverage of the international transactions at the country level is available in the Balance of Payment Statistics, published once a year by the IMF. Although information for most of the sub-Saharan African countries is included in the publication, the most recent database shows a complete set of annual data for just 18 out of 44 countries in the region, mostly the middle-income countries, two oil-exporters and the largest low-income economies among them. The other countries have missing data either for the capital flows components or some years, specially at the end of the period covered in the study, which is crucial for the econometric analysis. The work of the country desk economists was essentially to make sure that all series were available and the components were consistent with the main aggregates. We checked also consistency across the different published databases, finding discrepancies mostly in the last few years of the sample and few revisions to historical numbers in the early 2000’s for the countries with data in the various publications.

Looking into the World Economic Outlook database, there is a full coverage of the countries in the region, but it is difficult to separate actual figures from the forecasted ones or estimates (specially for countries with lags in reporting official data from the authorities) and programs with the Fund requiring accurate estimates on the missing BOP data, embedded in the Program. We used the WEO data, which are also reported by country desk economist in the Fund, as a starting point, and asked the desk economists to look into the details and components that are normally not used in the aggregated analysis of the external accounts offered in the WEO studies and data.

In brief, we have not created a separate database but rather taken the available information and refined it in a way that allows us to cover as many countries in the region as possible, regardless of the size of the country and the level of development, weighing properly in the model those countries with no capital flows (reporting zeros) and including those null figures as information in the dependent variable.

Appendix II. Variables Definitions and Sources

Appendix II.

Variables Definitions and Sources

article image
article image

Appendix III. Determinants of Capital Flows: Descriptive Statistics and Correlations

Appendix Table III.1.

Descriptive Statistics for Regression Variables

article image
Source: Authors’ calculations.
Appendix Table III.2.

Bivariate Correlations of Capital Flows with Independent Variables

article image
Note: pairwise correlation coefficient reported in first row, significance level reported in second row.Source: Authors’ calculations.
Appendix Table III.3.

Bivariate Correlations of Baseline Regression Variables

article image
Note: Pairwise correlation coefficients reported in first row, significance level reported below.Source: Authors’ calculations.
Appendix Table III.4.

Bivariate Correlations of Measures of Institutions and Instruments

article image
Note: Pairwise correlation coefficients reported in first row, significance level reported in the second row and number of observation in the third row.Source: Authors’ calculations.

Appendix IV. Private Capital Flows and Growth

Appendix Table IV.1

Private Capital Flows and Growth: Summary Statistics for Multivariate Regression

article image
Source: Authors’ calculations.
Appendix Table IV.2

Private Capital Flows and Growth: Pairwise Correlations for Multivariate Regression

article image
Note: * indicates significant at 10 percent.Source: Authors’ calculations.

References

  • Acemoglu, D., and S. Johnson, 2005, “Unbundling Institutions”, Journal of Political Economy, Vol. 113, No. 5, pp. 949995.

  • Acemoglu, D., Johnson, S., and D. Robinson, 2002, “Reversal of Fortune: Geography and Institutions in the Making of the Making of the Modern World Income Distribution”, Quarterly Journal of Economics, pp. 12311294.

    • Search Google Scholar
    • Export Citation
  • Albuquerque, R., 2003, “The Composition of International Capital Flows: Risk Sharing Through Foreign Direct Investment,” Journal of International Economics, Vol. 61, pp. 353383.

    • Search Google Scholar
    • Export Citation
  • Alfaro, L., and A. Charlton, 2007, “International Financial Integration and Entrepreneurial Firm Activity,” NBER Working Paper No. 13118 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation
  • Alfaro, L., S. Kalemli-Ozcan, and V. Volosovych, 2005a, “Capital Flows in a Globalized World: The Role of Policies and Institutions,”NBER Working Paper No. 11696 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation
  • Alfaro, L., S. Kalemli-Ozcan, and V. Volosovych, 2005b, “Why Doesn’t Capital Flow from Rich to Poor Countries? An Empirical Investigation,” NBER Working Paper Series No. 11901 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation
  • Auty, R. M., ed., 2001, Resource abundance and Economic Development: Improving the Performance of Resource-Rich Countries (Oxford: Oxford University Press).