Policy Research Working Paper 7903

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

Boubacar DialloFulbert Tchana Tchana

Albert G. Zeufack

Macroeconomics and Fiscal Management Global Practice Group &Africa RegionOffice of the Chief EconomistDecember 2016

WPS7903P

ublic

Dis

clos

ure

Aut

horiz

edP

ublic

Dis

clos

ure

Aut

horiz

edP

ublic

Dis

clos

ure

Aut

horiz

edP

ublic

Dis

clos

ure

Aut

horiz

ed

Produced by the Research Support Team

Abstract

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.

Policy Research Working Paper 7903

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 http://econ.worldbank.org. The authors may be contacted at diallobs@kfupm.edu.sa, ftchanatchana@worldbank.org, and azeufack@worldbank.org.

This paper explores the landscape, contributions, and determinants of sovereign wealth funds’ long-term invest-ments 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 sig-nificant 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 26 sovereign wealth fund investments over 1985-2013. 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.

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): diallobs@kfupm.edu.sa/dialloboubacar240@gmail.com‡Senior Economist, Macro-fiscal Management, the World Bank Group: ftchanatchana@worldbank.org.§Chief Economist, Africa Region, the World Bank Group: azeufack@worldbank.org.

Introduction

Sub-Saharan Africa (henceforth referred to as “Africa” for simplicity) has been on a two-

decades-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 2014. 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 sig-

nificantly 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 pur-

poses. 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. In-

deed, if SWFs could invest only 1 percent of their assets in Africa, this would amount to about

$420 billion over the period 2010-2020, which would broadly correspond to the missing half

of required investments in infrastructure needed to meet the Millennium Development Goals

(MDGs) (Zoellick, 2010). 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 (2011) 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

(2011) 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.

2

Second, we find empirically using a database of 26 SWF investments in Africa over the period

1985-2013 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 42 percent of investment, while only

10 percent of their investments are in infrastructure1; 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 1985-

2013. 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 2 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 in-

formation 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 2014 (see Sovereign Wealth Fund Institute), with $4,098.6

trillion being oil and gas related, and $2,732.4 trillion for other industries and commodities.

The mean of African SWFs in terms of capitalization is $16.2 billion, with a minimum of $0.003

billion, and a maximum of $6.9 billion. The percentage of African SWFs in total capitalization

represents 0.24 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,

1This 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

Table 1: African Sovereign Wealth Funds in billion of US$

Country Capitalization Type Origin

Botswana 6.9 Future Generations DiamondsAngola 5 Development OilNigeria 1.4 Stabilization OilCongo, Dem. Rep. 1.39 Stabilization OilSenegal 1 Development Non-commodityMauritius 0.5 Development Non-commodityGabon 0.4 Future Generations OilMauritania 0.3 Stablization Oil and GasSudan 0.15 Stabilization OilEquatorial Guinea 0.08 Future Generations OilGhana 0.07 Stabilization OilChad 0.03 Stabilization OilSao Tome Principe 0.01 Development OilNamibia n/a Development Minerals

Sources: Sovereign Wealth Fund Institute available at http://www.swfinstitute.org/. The information for the followingSWFs: Sao-Tome-Principe, Sudan, Chad and Congo, are taken from Triki and Faye (2011).

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 2, we present African SWFs capitalization over the period 2007-2013, 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 0.24 percent, this allows us to estimate

the capitalization of African SWFs over the period 2007-2013. The mean capitalization of

African SWFs over the period 2007-2013 is $11.21 billion, while the minimum is $8 billion, and

the maximum reaches $15 billion.

Table 2: African SWFs capitalization over the period 2007-2013 in $ billion

Year World capitalization African SWFs capitalization

2007 3,259 82008 4,149 102009 4,032 102010 4,418 112011 4,842 122012 5,198 12.52013 6,106 15

Sources: SWF world capitalization data are taken from the Sovereign WealthFund Institute, and the African SWFs capitalization data are based on theauthors’ calculations.

4

1.2 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. (2009, 2015), 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 2008, for 35 funds. We also used data taken from Chhaochharia and

Laeven (2009), who analyze the investments of SWFs by taking into account common cultural

traits. They collected information on a total of 29,295 equity holdings by SWFs over the period

1998-2007, comprising 10,282 equity holdings across 51 countries at year-end 2007. We use a

database taken from the Sovereign Wealth Center and the SWI, and another database taken

from Triki and Faye (2011). 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 2000-2013, and then calculate the

share of investments in terms of GDP received by African countries.2

To investigate SWF investments and economic growth, we use per-worker GDP data for 30

African countries over the period 2000-2013 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.

2Istithmar Investment Fund invested in Kenya, Djibouti, Rwanda, South Africa, and Tanzania. The MubadalaFund invested in Guinea, and Nigeria. The Brazilian SWF invested in Mozambique, Democratic Republic of Congoand Angola. The Chinese SWF invested in Nigeria, Sudan, Guinea, South Africa, and Ethiopia. The SingaporeanSWF (Temasek) invested in Nigeria, Tanzania, and South Africa.

3The World Development Indicators are publicly available at http://www.worldbank.org/.

5

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 2009 database

taken from Monitor-FEEM SWF, we show that SWFs have primarily invested in Asia, with 29

percent of total investment, followed by Europe with 27 percent, and the Middle East with 18

percent. America received 17 percent of total investment in 2009, 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 (2011), we are able to track the sectors in which

SWFs invest. For 50 SWF investments in Africa, we observe that 42 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 2

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

2000-2008, 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

Figure 2: Percentage of investments by sec-tor in Africa

Source: Authors’ calculation based on Triki and Faye(2011) and Monitor-FEEM SWF data

Figure 3: Percentage of world investmentsby sector

Source: Authors’ calculation based on Triki and Faye(2011) 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 2005-2014, using

data taken from the Sovereign Wealth Center. We found that between 2005 and 2014, six of

these investments were located in Africa with an amount of $129 million, and three were in the

United States, with an amount of $200 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. (2009) over the period

1986-2008, 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 1986-2008

4The Nigerian SWF invested $75 million in 2005 in Goldman Sachs, $75 million in 2007 in Credit Suisse, and$75 million in 2007 in VBS Global Asset Management.

7

in 62 projects. Only 30 percent of these investments were located domestically, with an amount

of $9 billion, which corresponds to 20 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 invest-

ments. This suggests that countries receiving SWF investments grow faster, and that these

investments positively affect economic growth.

Figure 4: Economic growth and SWF investment

AGO

BDIBENBFA

BWA

CIV

CMR

ETH

GHA

KENMDG

MRTMUSMWINAMNER

NGARWA

SDN

SEN

SWZSYCTGO

TZAUGA

ZAF

ZMB

ZWE

-50

510

15A

vera

ge g

row

th r

ate

0 .05 .1 .15 .2 .25SWF investments on GDP

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 (2011), 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

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

Tab

le3:

Corr

ela

tion

sM

atr

ix

Gro

wth

SW

Fin

vest

.C

red

itS

WF

inv.×

Cre

dit

Gro

wth

1.00

0

SW

Fin

v.

0.48

1∗∗∗

1.00

0(0

.008

)C

red

it-0

.223

-0.0

541.

000

(0.2

43)

(0.7

80)

SW

Fin

v.×

Cre

d0.

317∗

0.72

3∗∗∗

0.53

3∗∗∗

1.0

00

(0.0

93)

(0.0

00)

(0.0

02)

Sourc

e:A

uth

ors

’ca

lcula

tions.

***(1

%),

**(5

%)

and

*(1

0%

)

10

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 Rp be the random return on this portfolio, E(Rp) and V(Rp) be

the mean and the variance of the return on portfolio, respectively. Let θ1, θ2, and θ3 also be the

percentages of a portfolio invested in Africa, outside Africa, and in risk-free assets, respectively.

R1 and R2, the random returns of the two risky assets, respectively. Let also R1 and R2 be the

expected returns of the two risky assets, and Rf the return of the risk-free asset. Finally, let σ1

and σ2 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 X1, 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 (2001) and Grinblatt and Keloharju (2001), who found

that investors prefer to invest in familiar investment opportunities as opposed to unfamiliar

investments, as well as Guiso et et al. (2009), 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(Rp, X1) = −e−(φRp+(1−φ)θ1X1) (1)

where Rp = θ1R1 + θ2R2 + (1− θ1 − θ2)Rf . This utility function is increasing and concave

in Rp, and X1. The Jensen inequality allows us to determine the expected negative exponential

utility given by:

E(U(Rp, X1)) = −E(e−[φRp+(1−φ)θ1X1]

)= −e−E[φRp+(1−φ)θ1X1] (2)

with E(Rp) = Rp = θ1(R1 −Rf ) + θ2(R2 −Rf ) +Rf and E(X1) = X1 and V(X1) = σ2X . If

there is no correlation between the random return of portfolio, Rp, and the networking parameter

X1, then the log-normalization directly implies that:

E(U(Rp, X1)) = −e−{φθ1(R1−Rf )+φθ2(R2−Rf )+φRf+(1−φ)θ1X1− 12 [φ2θ21σ2

1+φ2θ22σ

22+(1−φ)2θ21σ2

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

maxθ1,θ2

φθ1(R1−Rf )+φθ2(R2−Rf )+φRf+(1−φ)θ1X1−1

2

[φ2θ21σ

21 + φ2θ22σ

22 + (1− φ)2θ21σ

2X

](4)

The first-order conditions directly imply that:

θ∗1 =

φ(R1−Rf )+(1−φ)X1

φ2σ21+(1−φ)2σ2

X

θ∗2 =φ(R2−Rf )

φ2σ22

θ∗3 = 1− θ∗1 − θ∗2

(5)

The percentages of portfolio θ∗1, θ∗2, and θ∗3 depend on the standard deviations σ1, σ2 and

σX ; as well as the expected returns R1, R2, and Rf and the networking parameter measured by

X1.

Result The percentage of investments in Africa, θ∗1, depends negatively on risk in Africa, σ1

and risk of network, σX , but positively on excess return in Africa (R1−Rf ) and the networking

parameter, X1.

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.

2 Empirical investigation

In terms of econometric specifications, we estimate the following equation:5

log(1 + θ1ij) = α1 + α2(R1ij −Rf ) + α3σ21ij + α4Xj + α5Yi + εij (6)

where i and j denote SWF and African country, respectively. θ∗i,j is the share of investments of

SWF i in African country j, (R1ij − Rf ) is the excess return defined as the difference between

the expected returns of African country j and the risk-free asset. σ1ij is the risk of investment

in African country j. Xj and Yi are country j and SWF i characteristics, respectively. εij is the

error term. α1 is the intercept and, α2, α3,α4 and α5 are constant coefficients. If the rate of

return on investment in Africa has a positive and significant effect on SWF investments, then

α2 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 α3 should

be negative. The coefficients of the networking variables, α4 and α5, are also expected to have

5See the Appendix for the derivation of equation (6) from the first equation of the system given by (5).

12

a positive and significant effect on SWF investments. The introduction of country and SWF

characteristics captured by Xj and Yi 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 (2009) 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.

(2012) find that political relations are important in why SWFs invest. However, for Avendano

and Santiso (2009), 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 26 SWFs in Africa over the

period 1985-2013.

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 1985-2013 taken from the new

Penn-World Table 8.0 of Feenstra et al. (2013). The United States is used as a proxy in terms

of investment in a risk-free asset, Rf . The expression Rj , 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 1985-2013. Our networking variables Xj and Yi are proxied by religion and

political fund, respectively. Specifically, Bortolotti et al. (2015) use the question 11 of Truman

(2008): “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

6We also use the average per-capita GDP growth rate to test the robustness of our results.7The database of political SWFs are from Bortolotti et al. (2015).8The database are publicly available at http://www.pewforum.org/2015/04/02/religious-projection-table/2010/number/all/9The data for financial development depth and bank efficiency come from the Global Financial Development

Database (GFDD), publicly available at http://econ.worldbank.org/.10The World Development Indicators are publicly available at http://www.worldbank.org/.

13

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. (2008) 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.12 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 1.11. The

maximum is 3.93, and the minimum is 0. In terms of transparency, the average is 7.25, with a

maximum of 10, and a minimum of 1. The average of SWF capitalization in $ billion is 103.95.

However, we observe a high dispersion between SWFs in terms of capitalization, confirmed by

the value of the standard deviation, which is equal to 184.04. The average of the rate of return

is -0.81, while the average of the risk is 0.22. The average of the dummy variable political fund

is 0.61. 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 -0.49. The maximum level

corruption is 0.5, while the minimum is equal to -1.12. The maximum of the variable property

rights takes a value equal to 3, while the minimum is 2. The averages of private credit is 30.03

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 108.63 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.

11Their dataset is publicly available at http://mba.tuck.dartmouth.edu/pages/faculty/rafael.laporta/publications.htmland https://www.prsgroup.com/about-us/our-two-methodologies/icrg.

12The database of SWF transparency measured by the Linaburg-Maduell index and capitalization are publiclyavailable at http://www.swfinstitute.org/

13The list of members are publicly available at http://www.oic-oci.org/oicv2/states/

14

Table 4: Summary statistics

Variables Obs Mean Std. Dev. Min. Max.

Logarithm of one plus the investment share 44 0.834 1.114 0 3.931Return 44 -0.813 1.051 -3.209 2.156Risk 44 0.224 0.064 0.064 0.277Political fund 44 0.613 0.492 0 1Christians (%) 44 28.478 30.592 2.672 91.946Muslims (%) 44 32.467 36.251 0.02 96.54

Transparency 32 7.25 2.552 1 10Capitalization (in $ billion) 44 103.954 184.041 1.400 893Private credit 44 30.025 34.916 4.397 108.63School 44 4.724 3.678 0.573 12.636Inflation 44 62.039 85.724 4.532 573.967Budget balance 39 -1.300 1.009 -4.769 -0.106Trade 44 66.469 17.466 25.858 119.502Corruption 44 -0.486 0.479 -1.12 0.5

Source: Authors’ calculations.

2.1 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 in-

vestment 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. (2012), but contradict those of Avendano and Santiso (2009), 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 (2009). Our results suggest that long-

term SWF investments in Africa depend more on rate of return. They also imply that political

15

SWFs tend to invest more in Africa. Using the Linaburg-Maduell transparency index for SWF

transparency as control in column (2), we find our coefficient of interest, α2, 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, α2, remains positive and significant at the 5 percent

level, while α3 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 statisti-

cally 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 (2011).

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

Table 5: SWFs and long-term investments in Africa

(1) (2) (3) (4) (5) (6) (7) (8)

Return 0.284∗∗ 0.438∗∗∗ 0.285∗∗ 0.276∗∗ 0.375∗∗ 0.300∗∗ 0.296∗∗ 0.448∗∗∗

(0.036) (0.010) (0.041) (0.048) (0.034) (0.047) (0.022) (0.002)Risk -3.346 -9.918∗∗∗ -3.350 -2.997 -2.126 -3.131 -3.525 -2.018

(0.156) (0.007) (0.161) (0.260) (0.418) (0.204) (0.143) (0.344)Political SWF 1.450∗∗∗ 0.962∗ 1.459∗∗∗ 1.536∗∗∗ 1.498∗∗∗ 1.452∗∗∗ 1.443∗∗∗ 1.444∗∗∗

(0.000) (0.080) (0.001) (0.002) (0.000) (0.000) (0.001) (0.001)Christians 0.011∗ 0.011∗∗ 0.011∗ 0.013∗ 0.011∗ 0.011∗ 0.010 0.009

(0.070) (0.025) (0.074) (0.094) (0.085) (0.072) (0.110) (0.174)Muslims -0.011∗ -0.001 -0.011∗ 0.000 -0.010∗ -0.010∗ -0.011∗∗ -0.010

(0.060) (0.783) (0.073) (0.986) (0.080) (0.079) (0.050) (0.102)SWF transparency -0.191

(0.158)SWF capitalization 0.000

(0.903)Dum. Islam SWF -0.877

(0.673)Private credit 0.005

(0.402)Overhead costs -0.039 -0.325∗∗

(0.756) (0.031)Cost to income ratio 0.022∗∗ 0.053∗∗∗

(0.042) (0.002)Cons. 0.966 4.143∗∗ 0.959 0.757 0.535 1.147 -0.309 -0.592

(0.189) (0.021) (0.197) (0.441) (0.509) (0.237) (0.706) (0.497)

Number of observations 44 32 44 44 44 44 44 44R2 0.388 0.540 0.388 0.395 0.402 0.390 0.411 0.453

Note that (***, ** and * ) indicate significance at the 1%, 5% and 10% levels, respectively. p-values are in parenthesis. Thedependent variable is the logarithm of the share of investments in Africa by an SWF. Column (1) regresses the logarithm of theinvestment 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 (2) 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 overheadcosts and cost to income ratio.

17

frontier, and transparency measured by corruption as controls. In column (2), 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 sig-

nificant. 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 posi-

tively 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 fron-

tier (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

Table 6: SWFs and long-term investments in Africa

(1) (2) (3) (4) (5) (6) (7)

Return 0.440∗∗ 0.290∗∗ 0.441∗ 0.309∗∗ 0.620∗∗∗ 0.291∗∗ 1.416∗∗∗

(0.013) (0.037) (0.086) (0.036) (0.005) (0.029) (0.000)Risk -1.225 -2.879 -6.113∗ -3.976 -0.295 -3.269 -15.137

(0.578) (0.252) (0.053) (0.178) (0.890) (0.221) (0.505)Political fund 1.541∗∗∗ 1.451∗∗∗ 1.420∗∗∗ 1.454∗∗∗ 1.558∗∗∗ 1.452∗∗∗ 1.424∗∗∗

(0.000) (0.000) (0.001) (0.001) (0.000) (0.000) (0.002)Christians 0.011∗ 0.012∗ 0.012 0.010 0.012∗ 0.011∗ 0.007

(0.074) (0.073) (0.152) (0.107) (0.069) (0.076) (0.370)Muslims -0.009 -0.011∗ -0.009∗ -0.011∗ -0.009∗ -0.011∗ -0.006

(0.102) (0.065) (0.092) (0.059) (0.093) (0.066) (0.311)School 0.098∗ 1.078

(0.082) (0.362)Inflation -0.000 0.097

(0.345) (0.325)Budget balance 0.244∗ -2.020

(0.100) (0.186)Trade 0.004 -0.033

(0.639) (0.467)Prox. technology frontier 0.698∗∗ 1.408

(0.026) (0.291)Corruption 0.033 -5.861

(0.931) (0.280)Cons. 0.014 0.893 2.031∗ 0.879 2.157∗∗ 0.969 -5.712

(0.166) (0.244) (0.056) (0.241) (0.028) (0.191) (0.586)

Number of observations 44 44 39 44 44 44 44R2 0.437 0.391 0.430 0.391 0.451 0.388 0.557

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 humancapital measured by school enrollment. Column (2) regresses the logarithm of the investment share in Africa of an SWFon rate of return, risk, political fund, religion and controls for inflation. Columns (3) controls for macroeconomic policiescaptured 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 controlvariable.

19

vestment share of SWFs on rate of return, risk, and networking indicators using cross-country

data over the period 1985-2013 for 26 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.

20

References

[1] Avendano, R. and Santiso, J. (2009) “Are Sovereign Wealth Funds’ investments politically

biased? A comparison with Mutual Funds”, OECD Development Working Paper, No. 283.

[2] Bartolotti, B., Fotak, V. and Megginson, W. (2015) “The sovereign wealth fund discount:

Evidence from Public Equity Investments”, The Review of Financial Studies, forthcoming.

[3] Bartolotti, B., Fotak, V., Megginson, W. and Miracky, W. (2009) “Sovereign Wealth Fund

Investment Patterns and Performance”, Institutions and Markets Working Paper Series,

Note di Lavoro.

[4] Bernstein, S., Lerner, J. and Schoar, A. (2013) “The Investment Strategies of Sovereign

Wealth Funds”, Journal of Economic Perspectives, 27, 219-238.

[5] Chhaochharia, V. and Laeven, L. (2009) “Sovereign Wealth Funds: Their Investment

Strategies and Performance”, CEPR Discussion Papers, 6959.

[6] Dixon, A.D. and Monk, A.H.B. (2011) “What Role for Sovereign Wealth Funds in Africa’s

Development”, Center for Global Development Working Paper.

[7] Feenstra, R.C., Inklaar, R. and Timmer, M.P. (2013) “The Next Generation of the Penn

World Table”.

[8] International Monetary Fund (2008) “Sovereign Wealth Funds-A Work Agenda”

[9] Grinblatt, M. and Keloharju, M. (2011) “How Distance, Language and Culture Influence

Stockholdings and Trades”, Journal of Finance, 56, 1053-1073.

[10] Guiso, L., Sapienza, P. and Zingales, L. (2009) “Cultural Biases in Economic Exchange”,

Quaterly Journal of Economics, 124, 1095-1131.

[11] Huberman, G. (2001) “Familiarity Breeds Investment”, The Review of Financial Studies,

14, 659-680.

[12] Knill, A.M., Lee, B.S and Mauck, N. (2012) “Bilateral Political Relations and Sovereign

Wealth Fund Investment”, Journal of Corporate Finance, 18, 108-123.

[13] Laporta, R., Lopez-de-Silanes, F., Shleifer, A. and Vishny, R. (1998) “Law and Finance”,

Journal of Political Economy, 106, 1113-1155.

[14] Laporta, R., Lopez-de-Silanes, F., Shleifer, A. and Vishny, R. (2008) “The Economic Con-

sequences of Legal Origins”, Journal of Economic Literature, 46, 285-332.

[15] Preqin Report. (2011) “The 2011 Preqin Sovereign Wealth Fund Review”

[16] Triki, T. and Faye, I. (2011) “Africa’s Quest for Development: Can Sovereign Wealth Funds

Help?”, International Finance Review, 12.

21

[17] Truman, E.M. (2008) “A Blueprint for Sovereign Wealth Fund Best Practices”, Peterson

Institute for International Economics Policy Brief.

[18] Turkisch, E. (2011) “Sovereign Wealth Funds as Investors in Africa: Opportunities and

Barriers”, OECD Development Center Working Paper, No.303

[19] Zoellick, R.B (2010) “The End of the Third World”, Speech at the Woodrow Wilson Inter-

national Center for scholars.

22

Appendix.

Deriving a linear function of the optimal portfolio allocation to

asset 1.

θ1 =φ(R1 −Rf ) + (1− φ)X

φσ21 + (1− φ)2σ2X

For brevity we dropped the overline on the rate of return in the next sections. Hence,

1 + θ1 =φ(R1 −Rf ) + (1− φ)X + φσ21 + (1− φ)2σ2X

φσ21 + (1− φ)2σ2X

Use a log transformation it is equal to

log(1+θ1) = log(φ(R1 −Rf ) + (1− φ)X + φσ21 + (1− φ)2σ2X

)− log

(φσ21 + (1− φ)2σ2X

)(7)

i.e.

log(1 + θ1) = g1((R1 −Rf ), X, σ21, σ2X) + g2(σ

21, σ

2X) (8)

where

g1((R1 −Rf ), X, σ21, σ2X) = log

(φ(R1 −Rf ) + (1− φ)X + φσ21 + (1− φ)2σ2X

)(9)

and

g2(σ21, σ

2X) = − log

(φσ21 + (1− φ)2σ2X

)Let us linearize equation 8

First order approximation A first order appromixation of g1((R1−Rf ), X, σ21, σ2X) around

(R01 −R0

f );X0;σ021;σ02X is given by

23

g1((R1 −Rf ), X, σ21, σ2X) '

g1((R01 −R0

f );X0;σ021 ;σ02X ))

+(R1 −Rf ) ∗∂g1((R1 −Rf ), X, σ21, σ

2X)

∂((R1 −Rf ))|[(R0

1−R0f );X

0;σ021 ;σ02

X ]

+X ∗∂g1((R1 −Rf ), X, σ21, σ

2X)

∂(X)|[(R0

1−R0f );X

0;σ021 ;σ02

X ]

+σ21 ∗∂g1((R1 −Rf ), X, σ21, σ

2X)

∂(σ21)|[(R0

1−R0f );X

0;σ021 ;σ02

X ]

+σ2X ∗∂g1((R1 −Rf ), X, σ21, σ

2X)

∂(σ2X)|[(R0

1−R0f );X

0;σ021 ;σ02

X ]

g1((R1 −Rf ), X, σ21, σ2X) ' γ0 + γ1(R1 −Rf ) + γ2X + γ3σ

21 + γ4σ

2X

γ0 = g1((R01 −R0

f );X01 ;σ021 ;σ02X )) = log

(φ(R0

1 −R0f ) + (1− φ)X0

1 + φσ021 + (1− φ)2σ02X)

γ1 =∂g1((R1 −Rf ), X, σ21, σ

2X)

∂((R1 −Rf ))|[(R0

1−R0f );X

0;σ021 ;σ02

X ]

∂g1((R1 −Rf ), X1, σ21, σ

2X)

∂((R1 −Rf ))=

∂(log(φ(R1 −Rf ) + (1− φ)X1 + φσ21 + (1− φ)2σ2X

)∂((R1 −Rf ))

=φ

φ(R1 −Rf ) + (1− φ)X + φσ21 + (1− φ)2σ2X

γ1 =φ

φ(R01 −R0

f ) + (1− φ)X0 + φσ021 + (1− φ)2σ02X

γ2 =∂g1((R1 −Rf ), X, σ21, σ

2X)

∂(X)|[(R0

1−R0f );X

0;σ021 ;σ02

X ]

∂g1((R1 −Rf ), X, σ21, σ2X)

∂(X)=

∂(log(φ(R1 −Rf ) + (1− φ)X + φσ21 + (1− φ)2σ2X

)∂(X)

=1− φ

φ(R1 −Rf ) + (1− φ)X + φσ21 + (1− φ)2σ2X

γ2 =1− φ

φ(R01 −R0

f ) + (1− φ)X0 + φσ021 + (1− φ)2σ02X

24

γ3 =∂g1((R1 −Rf ), X, σ21, σ

2X)

∂(σ21)|[(R0

1−R0f );X

0;σ021 ;σ02

X ]

∂g1((R1 −Rf ), X, σ21, σ2X)

∂(σ21)=

∂(log(φ(R1 −Rf ) + (1− φ)X + φσ21 + (1− φ)2σ2X

)∂(σ21)

=φ

φ(R1 −Rf ) + (1− φ)X + φσ21 + (1− φ)2σ2X

γ3 =φ

φ(R01 −R0

f ) + (1− φ)X0 + φσ021 + (1− φ)2σ02X

γ4 =∂g1((R1 −Rf ), X, σ21, σ

2X)

∂(σ2X)|[(R0

1−R0f );X

0;σ021 ;σ02

X ]

∂g1((R1 −Rf ), X, σ21, σ2X)

∂(σ2X)=

∂(log(φ(R1 −Rf ) + (1− φ)X + φσ21 + (1− φ)2σ2X

)∂(σ2X)

=(1− φ)2

φ(R1 −Rf ) + (1− φ)X + φσ21 + (1− φ)2σ2X

γ4 =(1− φ)2

φ(R01 −R0

f ) + (1− φ)X0 + φσ021 + (1− φ)2σ02X

Hence,

g1((R1 −Rf ), X1, σ21, σ

2X) '

log(φ(R0

1 −R0f ) + (1− φ)X0

1 + φσ021 + (1− φ)2σ02X)

+φ(R1 −Rf )

φ(R01 −R0

f ) + (1− φ)X0 + φσ021 + (1− φ)2σ02X

+(1− φ)X1

φ(R1 −Rf ) + (1− φ)X + φσ21 + (1− φ)2σ2X

+φσ21

φ(R1 −Rf ) + (1− φ)X + φσ21 + (1− φ)2σ2X

+(1− φ)2σ2X

φ(R01 −R0

f ) + (1− φ)X0 + φσ021 + (1− φ)2σ02X

A first order appromixation of g2(σ21, σ

2X) around (σ021;σ02X) is given by

25

g2(σ21, σ

2X) '

g2((σ021 ;σ02X ))

+σ21 ∗∂g1(σ

21, σ

2X)

∂(σ21)|(σ021 ;σ02X )

+σ2X ∗∂g1(σ

21, σ

2X)

∂(σ2X)|σ021 ;σ02X

δ1 =∂g2(σ

21, σ

2X)

∂(σ21)|(σ02

1 ;σ02X )

∂g2(σ21, σ

2X)

∂(σ21)=

∂(− log(φσ21 + (1− φ)2σ2X

)∂(σ21)

=−φ

φσ21 + (1− φ)2σ2X

=−φ

φσ021 + (1− φ)2σ02X

δ2 =∂g2(σ

21, σ

2X)

∂(σ2X)|(σ02

1 ;σ02X )

∂g2(σ21, σ

2X)

∂(σ2X)=

∂(− log(φσ21 + (1− φ)2σ2X

)∂(σ2X)

=−(1− φ)2

φσ21 + (1− φ)2σ2X

=−(1− φ)2

φσ021 + (1− φ)2σ02X

δ0 = g2(σ021 ;σ02X ) = − log

(φσ021 + (1− φ)2σ02X

)Hence,

g2(σ21, σ

2X) '

− log(φσ021 + (1− φ)2σ02X

)− φσ21φσ021 + (1− φ)2σ02X

−(1− φ)2σ2X

φσ021 + (1− φ)2σ02X

It follows that:

26

log(1 + θ1) ' log(φ(R0

1 −R0f ) + (1− φ)X0

1 + φσ021 + (1− φ)2σ02X)− log

(φσ021 + (1− φ)2σ02X

)+

φ(R1 −Rf )

φ(R01 −R0

f ) + (1− φ)X0 + φσ021 + (1− φ)2σ02X

+(1− φ)X

φ(R1 −Rf ) + (1− φ)X + φσ21 + (1− φ)2σ2X

+

(1

φ(R1 −Rf ) + (1− φ)X + φσ21 + (1− φ)2σ2X− 1

φσ021 + (1− φ)2σ02X

)φσ21

+

(1

φ(R01 −R0

f ) + (1− φ)X0 + φσ021 + (1− φ)2σ02X− 1

φσ021 + (1− φ)2σ02X

)(1− φ)2σ2X

log(1 + θ1) ' (γ0 + δ0) + γ1(R1 −Rf ) + γ2X + (γ3 + δ1)σ21 + (γ4 + δ2)σ

2X

log(1 + θ1) = α1 + α2(R1ij −Rf ) + α3σ21ij + α4Xj + α5σ

2Xj + εij

Let us assume that i is the individual index of a SWF and j is the country we will then have:

log(1 + θ1ij) = α1 + α2(R1ij −Rf ) + α3σ21ij + α4Xj + α5σ

2Xj + εij (10)

where i and j denote SWF and African country, respectively. θ1ij is the share of investments

of SWF i in African country j, (R1ij−Rf ) is the excess return defined as the difference between

the expected returns of African country j and the risk-free asset. σ1ij is the risk of investment in

African country j of SWF i. Xj is country j characteristics. εij is the error term. Since we use

religion as networking variable captured by Xj and that this variable is constant then σ2Xj = 0,

it follows directly that:

log(1 + θ1ij) = α1 + α2(R1ij −Rf ) + α3σ21ij + α4Xj + α5Yi + εij (11)

We add Yi in order to take into account the characteristics of SWF i such as capitalization,

transparency and political index.

27