Sovereign Wealth Funds and Long-Term Investments in Sub-Saharan Africa

Transcription

1 Public Disclosure Authorized Policy Research Working Paper 7903 WPS7903 Public Disclosure Authorized Public Disclosure Authorized Sovereign Wealth Funds and Long-Term Investments in Sub-Saharan Africa Boubacar Diallo Fulbert Tchana Tchana Albert G. Zeufack Public Disclosure Authorized Macroeconomics and Fiscal Management Global Practice Group & Africa Region Office of the Chief Economist December 016

2 Policy Research Working Paper 7903 Abstract This paper explores the landscape, contributions, and determinants of sovereign wealth funds long-term investments in Sub-Saharan Africa. The study finds that of all regions, Africa receives the lowest share of investment from sovereign wealth funds, and the landscape is dominated by Asian funds. The investment strategies of sovereign wealth funds established by African countries tend to be to invest less domestically and more abroad, contrary to Asian funds. In addition, using an enriched simple mean-variance portfolio model with an exponential utility function, the analysis shows that the investment rate of return and political connections have a positive and significant effect on sovereign wealth fund investments, and risk exerts a negative but not significant effect. The paper confirms these results empirically, using a database that includes 6 sovereign wealth fund investments over Hence, sovereign wealth funds investing in Africa care more about high returns and the political interests of their country of origin than the risk of their investment. This paper is a product of the Macroeconomics and Fiscal Management Global Practice Group and the Office of the Chief Economist, Africa Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at The authors may be contacted at diallobs@kfupm.edu.sa, ftchanatchana@worldbank. org, and azeufack@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team

3 Sovereign Wealth Funds and Long-Term Investments in Sub-Saharan Africa Boubacar Diallo Fulbert Tchana Tchana Albert G. Zeufack Keywords: Sovereign Wealth Funds, Long-Term Investments, Africa, Portfolio choice JEL classification: G11, R1 The authors would like to thank Havard Halland for his valuable comments and discussions. Assistant Professor, Department of Finance and Economics, College of Industrial Management (King Fahd University Petroleum and Minerals-KFUPM): Senior Economist, Macro-fiscal Management, the World Bank Group: Chief Economist, Africa Region, the World Bank Group:

4 Introduction Sub-Saharan Africa (henceforth referred to as Africa for simplicity) has been on a twodecades-long growth spurt and, despite the recent slowdown, the continent remains the third fastest-growing region in the world, with GDP growth averaging 4.6 percent in 014. Africa s commodities-driven growth has interestingly been underpinned by improved macroeconomic and fiscal stability, with inflation stabilized below 10 percent in most countries, lower debt, and significantly lower fiscal and external balances. Africa s brightest prospects have led to a renewed interest from investors, most of whom have been awakened to the continent s huge infrastructure deficit. Sovereign Wealth Funds (SWFs) have also joined the bandwagon. SWFs are government-owned investment funds set up for a variety of macroeconomic purposes. Generally speaking, there are three types of SWFs: stabilization funds, savings funds, and pension or reserve investment funds. These funds invest in natural resources as well as in infrastructure projects. While much has been said about SWFs as political investments or as stabilization mechanisms to manage growing natural resource rents, their role as a source of long-term finance for development, especially in Africa, has not been fully explored. Indeed, if SWFs could invest only 1 percent of their assets in Africa, this would amount to about $40 billion over the period , which would broadly correspond to the missing half of required investments in infrastructure needed to meet the Millennium Development Goals (MDGs) (Zoellick, 010). SWFs contribution to long-term finance is critical, as traditional foreign direct investments (FDI) cannot play this role. In this paper, we theoretically and empirically analyze SWF investments in Africa. Since SWF activities remain minimal in Africa, the question is what would it take for SWFs to contribute more to long-term financing of development in Africa? In terms of the literature, Triki and Faye (011) count at least 15 SWFs, and analyze how African economies can benefit from SWFs and use them as a channel to tap into international financial markets. They use data taken from different sources (Sovereign brands survey, JPMorgan) to discuss the landscape of African SWFs, as well as SWF interventions on the continent, and find that African Funds are small, suffer from poor governance structures, and mainly focus on stabilizing local economies. However, this paper does not analyze the determinants of SWF investments, nor the empirical relationship between SWF investments, economic growth and financial development. Turkisch (011) analyzes the opportunities and barriers for SWF investments in Africa, and suggests that the introduction of SWFs in African economies facilitates up to 50 percent of the investment needed for infrastructure. However, this paper also identifies the main barriers for SWFs in Africa, which include macroeconomic risks and high volatility of returns, the lack of technological knowledge, and the size and liquidity of African economies. Moreover, the latter study did not use a theoretical model or empirical estimates to obtain these findings. This lack of studies motivated us to explore the landscape and determinants of SWF investments in Africa. In doing this, we use an enriched simple mean-variance portfolio model to derive testable results. First, we theoretically show that the determinants of SWF investments in Africa are rate of return, risk, and networks. More precisely, we find that the rate of return has a positive and significant effect on SWF investments, while risk exerts a negative effect.

5 Second, we find empirically using a database of 6 SWF investments in Africa over the period that: i) about half of SWF investments in Africa are commodity-fueled funds, which raises some concerns given the end of the commodity super-cycle; ii) Real Estate and Hotels are the most popular sector for SWFs in Africa, totaling 4 percent of investment, while only 10 percent of their investments are in infrastructure 1 ; iii) SWFs that invest in Africa mainly originate from Asia, and Africa is the region that received the smallest share of SWF investments. Third, in terms of econometric specification, we regress the logarithm of the investment share on return rate, risk and networking parameters using cross-country data over the period Our empirical results show that rate of return has a positive and significant effect on SWF investments, while volatility exerts a negative but not significant effect. Most importanly, we also find that SWFs with a political bias are more likely to invest in Africa. Our results remain robust to the introduction of the following control variables: the Linaburg-Maduell transparency index, school, inflation, macroeconomic variables (government consumption and budget balance), trade, religion, corruption, property rights, and the use of legal origin as controls. The rest of the paper is organized as follows. Section 1 investigates some facts of SWF investments in Africa. Section presents the econometric investigation of the determinants of SWF investments. Finally, section 3 provides policy implications and concludes. 1 Stylized facts 1.1 What do we know about SWFs in Africa? Using data from the Sovereign Wealth Institute (SWI) transaction database, we document information on SWFs in Africa according to several variables. Table 1 shows African SWFs, their capitalization, type, and origin. In terms of numbers, the world market share of sovereign wealth funds reached a record amount of $6,831 trillion in September 014 (see Sovereign Wealth Fund Institute), with $4,098.6 trillion being oil and gas related, and $,73.4 trillion for other industries and commodities. The mean of African SWFs in terms of capitalization is $16. billion, with a minimum of $0.003 billion, and a maximum of $6.9 billion. The percentage of African SWFs in total capitalization represents 0.4 percent, and 0.4 percent if we take only the capitalization of SWFs related to oil and gas. This suggests that African SWFs are less capitalized than other emerging SWF countries. For example, SWFs from Asia are 40 percent capitalized, while 35 percent is capitalized in the Middle East. In column (3), we indicate the type of SWFs. The majority of African SWFs are stabilization and development funds. In terms of origin, column (4) shows that African SWFs are mostly related to oil windfalls. Following the literature, we also investigate the transparency of African SWFs, using the Linaburg-Maduell transparency index. This index, developed by Carl Linaburg and Michael Maduell, is based on 10 principles that depict SWF transparency to the public. The minimum rating takes a value of 1, while the maximum takes a value of 10. Overall, we find that African SWFs are not transparent. According to SWI, 1 This is in sharp contrast with Asia where the significant majority of SWF investments are in infrastructure. Other sectors of interest in Africa include extractive industries (16 percent) and manufacturing (16 percent). 3

6 Table 1: African Sovereign Wealth Funds in billion of US$ Country Capitalization Type Origin Botswana 6.9 Future Generations Diamonds Angola 5 Development Oil Nigeria 1.4 Stabilization Oil Congo, Dem. Rep Stabilization Oil Senegal 1 Development Non-commodity Mauritius 0.5 Development Non-commodity Gabon 0.4 Future Generations Oil Mauritania 0.3 Stablization Oil and Gas Sudan 0.15 Stabilization Oil Equatorial Guinea 0.08 Future Generations Oil Ghana 0.07 Stabilization Oil Chad 0.03 Stabilization Oil Sao Tomé Principe 0.01 Development Oil Namibia n/a Development Minerals Sources: Sovereign Wealth Fund Institute available at The information for the following SWFs: Sao-Tome-Principe, Sudan, Chad and Congo, are taken from Triki and Faye (011). and the available data, only Nigeria s SWF is quite transparent with a value of 9, followed by Botswana s SWF with a value of 6. According to the available information taken from SWI, African SWFs are all managed by local managers. In Table, we present African SWFs capitalization over the period , using the information given by the SWI in terms of the international capitalization of SWFs around the world. Given that the share of African SWFs is equal to 0.4 percent, this allows us to estimate the capitalization of African SWFs over the period The mean capitalization of African SWFs over the period is $11.1 billion, while the minimum is $8 billion, and the maximum reaches $15 billion. Table : African SWFs capitalization over the period in $ billion Year World capitalization African SWFs capitalization 007 3, , , , , , , Sources: SWF world capitalization data are taken from the Sovereign Wealth Fund Institute, and the African SWFs capitalization data are based on the authors calculations. 4

7 1. Data: SWF investments in Africa In this section, we present SWF investments in Africa. We first look at the sources of funding for SWFs that invest in the continent. Second, we take into account the geographical origin of the SWFs, and finally, we classify SWF investments in Africa by sector. We collected data from different sources. We use data taken from Bortolotti et al. (009, 015), who investigate the relationship between SWFs and their investment patterns and performance levels. Their database includes the total number and value of investments, the number and value of domestic investments, the number and value of listed equity investments, the number and value of unlisted equity investments, and the number and value of real estate investments of SWFs over the period January 1986 to September 008, for 35 funds. We also used data taken from Chhaochharia and Laeven (009), who analyze the investments of SWFs by taking into account common cultural traits. They collected information on a total of 9,95 equity holdings by SWFs over the period , comprising 10,8 equity holdings across 51 countries at year-end 007. We use a database taken from the Sovereign Wealth Center and the SWI, and another database taken from Triki and Faye (011). Finally, in order to have more information about SWF investments in Africa, we went to different SWF websites, and recorded their acquisitions in Africa. We then used different online sources to find the amounts invested in these companies by SWFs, particularly in financial and economic newspapers. Indeed, this strategy allowed us to track the country destinations of five SWF investments over the period , and then calculate the share of investments in terms of GDP received by African countries. To investigate SWF investments and economic growth, we use per-worker GDP data for 30 African countries over the period taken from the WDI. 3 We also use private credit as a measure of the level of financial development, as well as stock market capitalization as an alternative measure to analyze the relationship between SWF investments and growth through financial development. The datasets of these variables are also taken from WDI. SWFs that invest in Africa mainly obtain their funding from extractive industries. 47 percent of SWFs that invest in Africa obtain their finding from extractive industries, therefore 53 percent are from other sources (non-commodity). SWFs that invest in Africa mainly originate from Asia. The investments of SWFs in Africa are dominated by Asian funds, with 49 percent of the total, followed by African funds with 38 percent (this includes Lybia and Algeria). However, the domination by Asian SWFs is due to China, which represents 48 percent. The share of investments of SWFs from the Middle East and the Americas in Africa is 6 percent, while only 1 percent of SWF investments are from Europe. Istithmar Investment Fund invested in Kenya, Djibouti, Rwanda, South Africa, and Tanzania. The Mubadala Fund invested in Guinea, and Nigeria. The Brazilian SWF invested in Mozambique, Democratic Republic of Congo and Angola. The Chinese SWF invested in Nigeria, Sudan, Guinea, South Africa, and Ethiopia. The Singaporean SWF (Temasek) invested in Nigeria, Tanzania, and South Africa. 3 The World Development Indicators are publicly available at 5

8 Africa is the region that received the smallest share of SWF investments. In Figure 1, we show the percentage of the world SWF investments by target region. Using a 009 database taken from Monitor-FEEM SWF, we show that SWFs have primarily invested in Asia, with 9 percent of total investment, followed by Europe with 7 percent, and the Middle East with 18 percent. America received 17 percent of total investment in 009, while other regions attracted only 10 percent. This suggests that African countries attracted less SWF investment than any other region. In fact, Africa s share is certainly marginal. Figure 1: World SWF investments by Target Region Source: Authors calculation based on Monitor-FEEM SWF data The real estate and hotel sector is by far the largest share of SWF investments in Africa by sector. Following Triki and Faye (011), we are able to track the sectors in which SWFs invest. For 50 SWF investments in Africa, we observe that 4 percent are located in real estate and hotels, followed by extractive industries and industrial, with 16 percent, infrastructure with a value of 10 percent, and finally the financial and banking sectors with 8 percent. Figure presents the investments of SWFs in Africa by sector, while Figure 3 presents the percentage of SWF investments by sector around the world. Using Monitor-FEEM SWF data over the period , this shows that world investments by SWFs are located in real estate, with a value of 16 percent, the financial sector represents 18 percent of investments, while the industrial and energy sectors represent 17 and 9 percent, respectively. This suggests that SWF investments in Africa by sector are different from other regions in the world. 6

9 Figure : Percentage of investments by sector in Africa Figure 3: Percentage of world investments by sector Source: Authors calculation based on Triki and Faye (011) and Monitor-FEEM SWF data Source: Authors calculation based on Triki and Faye (011) and Monitor-FEEM SWF data 1.3 Are SWF investment strategies typically different in Africa? This section compares the strategies of African SWFs to SWFs from Asia and the Middle East in terms of domestic and foreign investments. Given the lack of data to track all African SWFs we use a case-study analysis of three SWFs from three different regions: investment strategies of the Nigerian SWF (Nigerian Sovereign Wealth Fund Authority), the Malaysian SWF (Khazanah Nasional) and the United Arab Emirates SWF (Mubadala Development Company). We were able to track 9 investments realized by the Nigerian SWF over the period , using data taken from the Sovereign Wealth Center. We found that between 005 and 014, six of these investments were located in Africa with an amount of $19 million, and three were in the United States, with an amount of $00 million. This suggests that the Nigerian SWF invests more in the United States with 61 percent of its investment than in Africa with only 39 percent for regional investments. 4 Using the data taken from Bortolotti et al. (009) over the period , we found that the Malaysian SWF have invested an amount of $10 billion in 108 projects. However, 90 projects were located domestically, with an amount of $8 billion, and the rest were in foreign countries. This suggests that the Malaysian SWF investments are located domestically with 80 percent. It also implies that in terms of investment strategy, the Nigerian SWF and the Malaysian SWF are very different. Finally, using the same database we found that Mubadala Development Company invested an amount of $30 billion over the period The Nigerian SWF invested $75 million in 005 in Goldman Sachs, $75 million in 007 in Credit Suisse, and $75 million in 007 in VBS Global Asset Management. 7

10 in 6 projects. Only 30 percent of these investments were located domestically, with an amount of $9 billion, which corresponds to 0 projects. These findings confirm the fact that African SWFs are more likely to invest abroad than domestically or regionally. 1.4 Do SWFs improve long-term financing and do their investments lead to growth in Africa? We investigate the relationship between SWF investments and economic growth and long-term financing. We find that the average GDP growth rate is positively correlated to SWF investments. This suggests that countries receiving SWF investments grow faster, and that these investments positively affect economic growth. Figure 4: Economic growth and SWF investment Average growth rate AGO UGA BFA GHA ZMB MRT BEN MUS MWI NAM BWA CMR NER SEN KEN BDI MDG SWZ SYC TGO CIV ZWE ZAF RWA TZA ETH SDN SWF investments on GDP NGA Source: Authors calculation However, we find that there is a weak negative correlation between SWFs and the level of financial development measured by private credit, while there is a positive correlation between private credit and the interaction term between SWF investments in Africa and private credit. This result is not surprising according to the results of Turkisch (011), who showed that the size and liquidity of African financial markets are barriers for SWF investments, and also that the small size of most African stock markets renders big investments difficult. This suggests that SWFs invest in Africa where the level of financial development is important. In addition, we verified the relationship between economic growth and SWF investments and the level of financial development by calculating the interaction between SWF investments and private credit. Our result shows that there is a positive correlation between economic growth and the interaction term. This implies that SWF investments affect the average growth rate in countries with a high level of financial development. To the best of our knowledge, these findings are the first in 8

11 the existing literature to empirically establish a positive relationship between SWF investments, financial development, and economic growth in Africa. The correlations among the variables are shown in Table 3. The average GDP growth rate is positively and significantly correlated with SWF investments at the 1 percent level. This result implies that SWF investments may increase economic growth for African countries. The correlation between growth rate and the interaction term between SWF investments and private credit is also positive and significant at the 10 percent level, while SWF investments is negatively but insignificantly correlated with private credit. The correlation between the level of financial development measured by private credit and the interaction term is positive and statistically significant different from zero at the 1 percent level. This suggests that SWFs invest in countries with a good financial system, and the combination of these investments and the financial development increases economic growth. 9

12 Table 3: Correlations Matrix Growth Growth SWF invest. Credit SWF inv. Credit SWF inv (0.008) Credit (0.43) (0.780) SWF inv. Cred (0.093) (0.000) (0.00) Source: Authors calculations. ***(1%), **(5%) and *(10%) 10

13 The Theoretical Model In order to study the relationship between SWFs and long-term investments in Africa, we built a portfolio choice theoretical model. We assume that the manager of each SWF invests in portfolios that consist of three types of assets: two risky assets, and one risk-free asset. The first risky asset represents investments in Africa by SWFs, while the second risky asset captures investments outside Africa. Let R p be the random return on this portfolio, E( R p ) and V( R p ) be the mean and the variance of the return on portfolio, respectively. Let θ 1, θ, and θ 3 also be the percentages of a portfolio invested in Africa, outside Africa, and in risk-free assets, respectively. R 1 and R, the random returns of the two risky assets, respectively. Let also R 1 and R be the expected returns of the two risky assets, and R f the return of the risk-free asset. Finally, let σ 1 and σ be the standard deviations of the two risky assets, respectively. We assume that there are no investment costs, and that the covariance between the two risky assets is equal to zero. Each SWF manager augments the return of the risky asset by a preference to the region represented by X 1, with 0 φ 1, being the weight given to this preference. This allows us to analyze the foreign investment bias of SWF investments in Africa. It also captures historical and cultural variables, specifically religion, language or legal origin, for example. This modeling feature is motivated by the findings of Huberman (001) and Grinblatt and Keloharju (001), who found that investors prefer to invest in familiar investment opportunities as opposed to unfamiliar investments, as well as Guiso et et al. (009), who found that historical and cultural variables, particularly religion, affect trust in people from other countries, and that these differences in trust affect people s financial decisions. We assume that the SWF manager maximizes expected negative exponential utility and that the manager has a constant attitude towards risk: U( R p, X 1 ) = e (φ R p+(1 φ)θ 1 X 1) (1) where R p = θ 1 R 1 + θ R + (1 θ 1 θ )R f. This utility function is increasing and concave in R p, and X 1. The Jensen inequality allows us to determine the expected negative exponential utility given by: E(U( R ( ) p, X 1 )) = E e [φ R p+(1 φ)θ 1 X 1] = e E[φ R p+(1 φ)θ 1 X 1] () with E( R p ) = R p = θ 1 (R 1 R f ) + θ (R R f ) + Rf and E(X 1 ) = X 1 and V(X 1 ) = σ X. If there is no correlation between the random return of portfolio, R p, and the networking parameter X 1, then the log-normalization directly implies that: E(U( R p, X 1 )) = e {φθ 1(R 1 R f )+φθ (R R f )+φr f +(1 φ)θ 1 X 1 1 [φ θ 1 σ 1 +φ θ σ +(1 φ) θ 1 σ X]} (3) The objective of the manager is to maximize a modified expected negative exponential utility rate given by (3). The maximization problem becomes: 11

14 max φθ 1 (R 1 R f )+φθ (R R f )+φr f +(1 φ)θ 1 X 1 1 [ φ θ θ 1,θ 1σ1 + φ θσ + (1 φ) θ1σ X ] (4) The first-order conditions directly imply that: θ 1 = φ(r 1 R f )+(1 φ)x 1 φ σ 1 +(1 φ) σ X θ = φ(r R f ) φ σ θ 3 = 1 θ 1 θ (5) The percentages of portfolio θ 1, θ, and θ 3 depend on the standard deviations σ 1, σ and σ X ; as well as the expected returns R 1, R, and R f and the networking parameter measured by X 1. Result The percentage of investments in Africa, θ 1, depends negatively on risk in Africa, σ 1 and risk of network, σ X, but positively on excess return in Africa (R 1 R f ) and the networking parameter, X 1. Proof. These properties follow directly from differentiating equation (5). The result is quite intuitive. First, the increase in risk in Africa reduces the amount invested by SWFs. The excess return that indicates the return rate on an investment relative to the return rate on the risk-free investment increases SWF investments in Africa. It also shows that an increase in the networking parameter positively affects SWF investments in Africa. Empirical investigation In terms of econometric specifications, we estimate the following equation: 5 log(1 + θ 1ij ) = α 1 + α (R 1ij R f ) + α 3 σ1ij + α 4 X j + α 5 Y i + ɛ ij (6) where i and j denote SWF and African country, respectively. θi,j is the share of investments of SWF i in African country j, (R 1ij R f ) is the excess return defined as the difference between the expected returns of African country j and the risk-free asset. σ 1ij is the risk of investment in African country j. X j and Y i are country j and SWF i characteristics, respectively. ɛ ij is the error term. α 1 is the intercept and, α, α 3,α 4 and α 5 are constant coefficients. If the rate of return on investment in Africa has a positive and significant effect on SWF investments, then α should be positive and statistically significant different from zero. The risk of investment in Africa is expected to exert a negative and significant effect on SWF investments, thus α 3 should be negative. The coefficients of the networking variables, α 4 and α 5, are also expected to have 5 See the Appendix for the derivation of equation (6) from the first equation of the system given by (5). 1

15 a positive and significant effect on SWF investments. The introduction of country and SWF characteristics captured by X j and Y i allows us to examine if there exists cultural bias for SWF investments in Africa, and if the drivers of these investments are financial rather than political. Indeed, Chhaoccharia and Laeven (009) found that cultural bias is more pronounced for SWFs and they tend to invest in countries with similar cultural traits. Most importantly, they showed that religious closeness is positively and significantly related to SWF investments. Knill et et al. (01) find that political relations are important in why SWFs invest. However, for Avendano and Santiso (009), SWF investments do not exhibit political bias, and their investments are not different from Mutual Fund investments. Our dependent variable is the logarithm of one plus the share of each investment in terms of total capitalization for 6 SWFs in Africa over the period In order to estimate the empirical model, we use the following proxies. We first calculate the excess return using the average GDP growth rate over the period taken from the new Penn-World Table 8.0 of Feenstra et al. (013). The United States is used as a proxy in terms of investment in a risk-free asset, R f. The expression R j, which denotes the expected return in African country j, is measured as the average per-worker GDP growth rate. 6 To measure the risk of investments in Africa, we use the standard deviation of the average per-worker GDP growth rate over the period Our networking variables X j and Y i are proxied by religion and political fund, respectively. Specifically, Bortolotti et al. (015) use the question 11 of Truman (008): Are decisions on specific investments made by the Managers? to construct the index of political fund. They classify all funds with managerial independence score of less than 1 as political funds and all others as non political funds. They then define a dummy variable that takes the value of 1 for political fund and 0 if not. 7 The religion variable includes the share of Christians and Muslims in a given country using the new database of the PewResearch Center. 8 In terms of control variables, we use the following variables: Private credit and bank costs to take into account the level of financial development and bank efficiency, respectively. Private credit is calculated as the credit provided by the banking sector, including all credit to various sectors on a gross basis, with the exception of credit to the central government, which is net and bank costs are measured by overhead costs to total assets (%), defined as operating expenses of a bank as a share of the value of all assets held, and cost to income ratio (%), defined as operating expenses of a bank as a share of the sum of net-interest revenue and other operating income. 9 We also control our results using school, inflation, macroeconomic policies such as budget balance and trade taken from the World Bank WDI. 10 School is measured by the total enrollment in secondary education, regardless of age, expressed as a percentage of the population of official secondary education age. Inflation, consumer price index, measured by the consumer price index, reflects the annual percentage change in the cost to the average consumer of acquiring a 6 We also use the average per-capita GDP growth rate to test the robustness of our results. 7 The database of political SWFs are from Bortolotti et al. (015). 8 The database are publicly available at 9 The data for financial development depth and bank efficiency come from the Global Financial Development Database (GFDD), publicly available at 10 The World Development Indicators are publicly available at 13

16 basket of goods and services that may be fixed or changed at specified intervals, such as yearly (the Laspeyres formula is generally used); budget balance as % of GDP is cash surplus or deficit revenue (including grants) minus expenses and the net acquisition of non-financial assets. In the 1986 GFS manual, non-financial assets were included under revenue and expenditure in gross terms and trade calculated as the sum of exports (% of GDP) and imports (% of GDP). In terms of institutional variables we use corruption taken from Laporta et al. (008) and the International Country Risk Guide (ICRG). Higher values in corruption reflect more developed institutions. 11 To take into account the level of innovation of a given country we introduce the proximity to the world technology frontier, defined as the ratio of the GDP per-worker of the country in 1985 to the GDP of the US at the same period. Finally, we add several SWF characteristics as controls. The following variables are used: Transparency, capitalization taken from the SWI. 1 To investigate the foreign investment bias for SWFs that invest in Africa, we define a dummy variable that takes a value 1 if the country origin of SWF is a member of the Organization of the Islamic Conference (OIC) and 0 if not. More particularly, this dummy variable allows us to analyze if SWFs from muslim countries invest more or less in Africa. 13 Table 4 shows the summary statistics of our variables. The average of the logarithm of the investment share in Africa by SWFs is 0.83, with a standard deviation equal to The maximum is 3.93, and the minimum is 0. In terms of transparency, the average is 7.5, with a maximum of 10, and a minimum of 1. The average of SWF capitalization in $ billion is However, we observe a high dispersion between SWFs in terms of capitalization, confirmed by the value of the standard deviation, which is equal to The average of the rate of return is -0.81, while the average of the risk is 0.. The average of the dummy variable political fund is The maximum of political SWF is 1, while the minimum is 0. In terms of institutional variables, we find that the averages of the level of the corruption is The maximum level corruption is 0.5, while the minimum is equal to The maximum of the variable property rights takes a value equal to 3, while the minimum is. The averages of private credit is percent of GDP. The minimum of the level of financial development measured by private credit is 4.40 percent of GDP, while the maximum is percent. In terms of correlation we find that the logarithm of the share of SWF investments in Africa is negatively correlated with the Linaburg-Maduell transparency index. This suggests that transparent SWFs are less likely to invest in Africa. However, we find a positive correlation between political fund and political index with the logarithm of the investment share, suggesting that political SWFs are more likely to invest in Africa. 11 Their dataset is publicly available at and 1 The database of SWF transparency measured by the Linaburg-Maduell index and capitalization are publicly available at 13 The list of members are publicly available at 14

17 Table 4: Summary statistics Variables Obs Mean Std. Dev. Min. Max. Logarithm of one plus the investment share Return Risk Political fund Christians (%) Muslims (%) Transparency Capitalization (in $ billion) Private credit School Inflation Budget balance Trade Corruption Source: Authors calculations..1 Results This section presents the results of our specifications. The dependent variable is the logarithm plus one of the share of investments in Africa of SWF i. In each specification, we introduce the intercept, and we deal the heteroskedasticity problems in regressions by using White s consistent standard errors for statistical inference. The specifications are estimated using ordinary least squares (OLS). In column (1) of Table 5 we regress the logarithm of one plus investment share of SWFs in Africa on rate of return and risk in Africa, as well as our networking variables captured by political and religion motives of the fund. This allows us to test if SWF investments are politically biased or not. The introduction of religion takes into account the cultural bias of SWF investments in Africa. Our main results show that return on investment has a positive and significant effect on SWF investment, suggesting that the rate of return enhances the investment of SWFs in Africa. The coefficient associated with risk enters negatively but is not significant. This result suggests that risk decreases the investment of SWFs in Africa, but does not affect SWF investments. Most importantly, we find that the coefficient of political fund enters positively and statistically significantly different from zero at the 1 percent level. This result suggests that SWFs with political views are more likely to invest in Africa. Our results are in line with Knill et al. (01), but contradict those of Avendano and Santiso (009), who showed that SWF investments are not politically biased. For the second type of networking variable, namely religion, we find that the coefficients associated with the share of christians and muslims enter (weakly) positively and negatively significantly different from zero at the 10 percent level, respectively. This implies that SWFs from christian countries tend to invest more in Africa, a finding in line with Chhaoccharia and Laeven (009). Our results suggest that longterm SWF investments in Africa depend more on rate of return. They also imply that political 15

18 SWFs tend to invest more in Africa. Using the Linaburg-Maduell transparency index for SWF transparency as control in column (), we find our coefficient of interest, α, remains positive and significant at the 1 percent level, while risk enters negatively and statistically significantly different from zero at the 1 percent level. In terms of networking variables, we find that the dummy variable political fund remains positive and significant at the 10 percent level, while the share of christians has a positive and significant effect on the investment share of SWFs in Africa at the 5 percent level. In column (3), we add the capitalization of SWFs in $ billion as a control. Our coefficient of interest, α, remains positive and significant at the 5 percent level, while α 3 remains negative but insignificant. This suggests that return rate increases SWF investments, while risk does not affect these investments. The coefficients associated with the dummy variable political fund, the share of christians and muslims, enter positively and statistically significantly different from zero at the 1 percent and 10 percent levels, respectively. These findings confirm our theoretical model and empirical strategy, and show that there are political and cultural biases for SWF investments in Africa. In column (4), we find an interesting result after controlling for the dummy variable called Islamic fund. The rate of return has a positive and significant effect on SWF investments, while risk exerts a negative and insignificant effect. We also find that SWF investments are politically biased since the coefficient of political fund remains positive and significant at the 1 percent level. Also, the coefficient associated with the share of christians enters positively and significantly different from zero at the 10 percent level. But the dummy variable Islamic fund enters negatively and insignificantly. To take into account the level of financial development and bank efficiency of countries in Africa, we add private credit, overhead costs and cost to income ratio as controls in columns (5)-(8). The coefficients associated with the return rate and political fund remain positive and significant, while risk remains negative but is not significant. The share of christians remains positive and significant at the 10 percent level, while the share of muslims remains negative. This suggests that SWFs from christian countries tend to invest more in Africa but not for those from muslim countries. The introduction of bank efficiency measures confirms our results listed above. They suggest that the return rate enhances SWF investments, and political SWFs invest more in Africa. Surprisingly, the coefficient associated with bank cost to income ratio enters positively and statistically significantly different from zero at the 5 percent level. This suggests that SWFs tend to invest more in countries with less efficient banking systems, a finding in contraction with Turkisch (011). It could be that our results are driven by human capital. To control for this fact, we add school enrollment in column (1) as a control of Table 6. The coefficient associated with the return rate remains positive and significant at the 5 percent level, while risk remains negative and insignificant. Also, political fund remains positive and significant at the 1 percent level. As expected, human capital enhances SWF investments in Africa as the coefficient of school has a positive and significant effect on the share of investments at the 10 percent level. This suggests that SWFs tend to invest in African countries with good education system. We continue to check the robustness of our findings by adding inflation, macroeconomic policies such as budget balance, openness measured by trade, innovation captured by the proximity to the technology 16

19 Table 5: SWFs and long-term investments in Africa (1) () (3) (4) (5) (6) (7) (8) Return (0.036) (0.010) (0.041) (0.048) (0.034) (0.047) (0.0) (0.00) Risk (0.156) (0.007) (0.161) (0.60) (0.418) (0.04) (0.143) (0.344) Political SWF (0.000) (0.080) (0.001) (0.00) (0.000) (0.000) (0.001) (0.001) Christians (0.070) (0.05) (0.074) (0.094) (0.085) (0.07) (0.110) (0.174) Muslims (0.060) (0.783) (0.073) (0.986) (0.080) (0.079) (0.050) (0.10) SWF transparency (0.158) SWF capitalization (0.903) Dum. Islam SWF (0.673) Private credit (0.40) Overhead costs (0.756) (0.031) Cost to income ratio (0.04) (0.00) Cons (0.189) (0.01) (0.197) (0.441) (0.509) (0.37) (0.706) (0.497) Number of observations R Note that (***, ** and * ) indicate significance at the 1%, 5% and 10% levels, respectively. p-values are in parenthesis. The dependent variable is the logarithm of the share of investments in Africa by an SWF. Column (1) regresses the logarithm of the investment share in Africa of an SWF on rate of return, risk and the networking parameters captured by political fund and religion (catholic, muslim and protestant). Column () controls for the transparency of SWFs measured by the Linaburg-Maduell index. Columns (3) and (4) add SWF capitalization and the dummy variable Islamic fund for SWFs from Muslim countries, respectively. Columns (5)-(8) control for the level of financial development measured by private credit and bank efficiency captured by overhead costs and cost to income ratio. 17

20 frontier, and transparency measured by corruption as controls. In column (), we regress the logarithm of investment share of SWFs on rate of return, risk in Africa, political fund, the share of christians and muslims, and inflation. Our results show that return on investment has a positive and significant effect on SWF investment, suggesting that the rate of return enhances the investment of SWFs. The coefficient associated with risk enters negatively but is not significant. This result suggests that risk decreases the investment of SWFs in Africa. Political fund remains positive and significant at the 1 percent level. The share of christians enters positively and statistically significantly different from zero at 10, while the share of muslims remains negative and significant at the 10 percent level. These findings show that the excess return that indicates the rate of return on an investment relative to the rate of return on the risk-free investment increases SWF investments and that there exists political and cultural biases for these investments in Africa. Budget balance is introduced in column (3), and all coefficients of interest enter significant with the expected signs. Most importantly, the rate of return has a positive and significant effect on SWF investments at the 10 percent level. To take into account the degree of openness we control for trade in column (4). Our main results remain unchanged. It could be that SWFs invest in countries close to the technology frontier in terms of innovation. In addition, we add into our specifications the proximity of the country to the technology frontier (where the leader is the United States). The coefficient associated with the rate of return remains positive and significant. More specifically, the magnitude of the coefficient associated with the rate of return increases and remains positive and significant at the 1 percent level. Also, the networking parameter captured by political motive of the fund remains positive and significant at the 1 percent level. This suggests that SWFs with a political bias are more likely to invest in Africa. We also find that the coefficient associated with the share of christians has a positive and significant effect on SWF investments. Interestingly, the coefficient associated with the proximity to the technology frontier also enters positively and statistically significantly different from zero at the 5 percent level. This shows that the share of SWF investments in Africa increases with the level of technological innovation. In terms of institutional variables, we use the level of corruption. The results are shown in columns (6) and (7). The signs and significance of our coefficients of interest remain unchanged. Our main results remain robust to the introduction of inflation, macroeconomic variables, trade, corruption and the proximity to the technology frontier as controls, and confirm at the same time the results listed above, and our theoretical implications. 3 Conclusion and policy implications Using available data on SWF in Africa, we observe that 47 percent of SWFs that invest in Africa are from minerals (extractive commodity), while 53 percent are from non-commodity sources. In addition, we find that African funds invest less domestically, in contrast to Asian funds. In addition, we show using an enriched mean-variance portfolio model that the determinants of SWF investments in Africa are excess return, risk and networks. We test our theoretical predictions empirically, using an econometric specification regressing the logarithm of the in- 18

21 Table 6: SWFs and long-term investments in Africa (1) () (3) (4) (5) (6) (7) Return (0.013) (0.037) (0.086) (0.036) (0.005) (0.09) (0.000) Risk (0.578) (0.5) (0.053) (0.178) (0.890) (0.1) (0.505) Political fund (0.000) (0.000) (0.001) (0.001) (0.000) (0.000) (0.00) Christians (0.074) (0.073) (0.15) (0.107) (0.069) (0.076) (0.370) Muslims (0.10) (0.065) (0.09) (0.059) (0.093) (0.066) (0.311) School (0.08) (0.36) Inflation (0.345) (0.35) Budget balance (0.100) (0.186) Trade (0.639) (0.467) Prox. technology frontier (0.06) (0.91) Corruption (0.931) (0.80) Cons (0.166) (0.44) (0.056) (0.41) (0.08) (0.191) (0.586) Number of observations R Note that (***, ** and * ) indicate significance at the 1%, 5% and 10% levels, respectively. p-values are in parenthesis. The dependent variable is the logarithm of the share of investments in Africa by an SWF. Column (1) controls for human capital measured by school enrollment. Column () regresses the logarithm of the investment share in Africa of an SWF on rate of return, risk, political fund, religion and controls for inflation. Columns (3) controls for macroeconomic policies captured by budget balance. Column (4) takes into account the degree of openness measured by trade. Columns (5) introduces the proximity for a given country to the world technology frontier. Column (6) uses corruption as a control variable. 19

22 vestment share of SWFs on rate of return, risk, and networking indicators using cross-country data over the period for 6 SWFs. Empirical results show that rate of return has a positive and significant effect on SWF investments, while risk exerts a negative and but not significant effect. Most importantly, we find that political SWFs are more likely to invest in Africa. These results are robust to the introduction of the following control variables: the Linaburg-Maduell transparency index, the political structure of the SWF, religion, corruption, private credit, bank efficiency, and the proximity to the technology frontier. To increase the quantity of SWF investments in Africa, policy makers should set up good governance and transparency policies by promoting a low level of corruption and high levels of property rights. African countries should reform their financial sectors by rendering their stock markets more attractive. Finally, strong governance and a higher level of coordination among public entities positively affect the efficiency and effectiveness of public investment programs. 0