Host country characteristics as determinants for sovereign wealth funds’ location choice

(1)

1

Master Thesis

Host country characteristics as determinants for

sovereign wealth funds’ location choice

Jason Steendijk (11175370)

University of Amsterdam, Faculty of Economics and Business

MSc. in Business Administration - International Management track

Supervisor: Dr. V.G. Scalera

Date of submission: 26 January 2017 – Final Version

(2)

Statement of originality

This document is written by Student Jason Steendijk who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

List of figures

Figure 1: Conceptual framework ... 20

List of tables

Table 1. Overview of hypotheses ... 24

Table 2. SWFs and number of investments ... 27

Table 3. Number of SWFs investing in each country ... 28

Table 4. Definition and source of variables ... 33

Table 5. Descriptive statistics ... 36

Table 6. Correlation matrix ... 38

Table 7. Logit regression results ... 42

Table 8. Overview of hypotheses evaluation ... 45

(5)

Abstract

Sovereign wealth funds (SWFs) have lately become major actors in the global financial market with total assets worth USD 7.4 trillion, resulting in gained political attention. Opposition and accusations of threatening national security towards SWFs grew ever since their investments in large Western financial firms during the most recent economic crisis. Nevertheless, little is known about SWFs’ strategies, behaviours and intentions. Accordingly, to reveal more about SWF strategies for FDI, recent studies suggest further research on SWF location choice and host country characteristics’ influence. This thesis therefore focused on finding out whether and how SWFs’ location choices are influenced by host country characteristics, contributing to current literature by revealing more about SWFs’ intents. As in previous SWF and multinational enterprise (MNE) literature, the following host country characteristics are researched: GDP growth, tax rate, unemployment rate, bilateral trade agreements (BTAs), market size and corruption rate. These characteristics have been selected by using the location pillar of the OLI paradigm, which covers demand conditions and government induced policies. Data about global SWF acquisitions was mainly gathered through online databases. The final dataset covers the period from 2005 to 2013, consisting of 28 SWFs who have invested in 61 countries, thus 1708 observations. After quantification of the data, statistical tests including a logit regression analysis allowed for interpretation of the data. The results showed a positive and statistically significant influence of taxation, unemployment, BTAs, market size and corruption on SWF location choice.

Keywords: Sovereign wealth funds, location choice, host country characteristics

(6)

1. Introduction

As a result of their rapid growth in numbers and size throughout the last few decades, sovereign wealth funds (SWFs), which are government-owned investment vehicles (Bertoni & Lugo, 2014), have become an increasingly important class of investors in the global capital market (Kotter & Lel, 2011; Ngoc, 2015). According to Megginson and Fotak (2014), this is demonstrated by more than 25 countries, which have launched new SWFs since the beginning of 2008. SWFs have such an important role in the field of global capital flows as they have access to assets with an estimated worth of USD 7.4 trillion (Sovereign Wealth Fund Institute, 2016; Ngoc, 2015). Thus, SWFs are important actors in today’s global financial stability (Ngoc, 2015). Recently, this has resulted in gained attention from the media, general public and politicians (Bortolotti, Fotak & Megginson, 2015). Alhasel (2015, p. 2) states that ‘’this increased attention is mainly due to the large investments these SWFs have made into some of the largest firms of the Western world, specifically financial firms’’. For instance, non-Western SWFs investing in Western financial firms like Citicorp and Merrill Lynch during the recent economic crisis (Alhasel, 2015).

Furthermore, SWFs have been accused of threatening the national security of the countries that they invest in and of stealing intellectual property by investing in strategic industries (Alhasel, 2015). Resulting in growing political opposition towards SWFs, which was exemplified by German Chancellor Angela Merkel and by members of the U.S. Congress (Megginson & Fotak, 2014). They expressed their concerns when Russian SWFs bought pipelines and energy infrastructure in Europe back in 2007 (Megginson & Fotak, 2014), giving those Russian SWFs a certain degree of control over European oil and energy supply. Consequently, the impact of SWFs has surged a lot of discussions among scholars and practitioners (Megginson & Fotak, 2014). Nevertheless, little is known and understood about the strategies, objectives and behaviour of SWFs (Kotter & Lel, 2011). Accordingly, recent

(7)

academic studies suggest that SWFs and their location choice for foreign direct investment (FDI) are important domains for future research as it could reveal more about SWF strategies (Murtinu & Scalera, 2016; Shi, Hoskisson & Zhang, 2016; Ngoc, 2015).

There have been numerous studies that have investigated the FDI location choices of MNEs. However, it is questionable whether the results of these studies are applicable to SWFs’ investments due to an important difference between MNEs and SWFs. Unlike SWFs (Alhasel, 2015; Bortolotti et al., 2015; Ngoc, 2015; Fatemi, Fooladi & Kayhani, 2011), MNEs are not necessarily government-owned (Shi et al., 2016; Rudy, Miller & Wang, 2016). Furthermore, little research is done on the location choice by SWFs (Murtinu & Scalera, 2016). Previous studies did address matters such as determinants of SWF investments’ stock prices (Murtinu & Scalera, 2015) and the effect of geopolitical factors on location choice by SOEs (Shi et al., 2016).

In addition, the current literature on SWFs is mainly focussed on the impact of SWFs’ investments on the market value and performance of listed firms (Murtinu & Scalera, 2015; Fernandes, 2014) and on geographical distribution of SWF activities worldwide (Fatemi et al., 2011; Chhaochharia & Laeven, 2009; Bernstein, Lerner & Schoar, 2013). Yet, none of them specifically elaborated upon the influence of host country specific characteristics on SWFs’ location choices. Thus, as suggested by Murtinu and Scalera (2016) and Quer, Claver and Rienda (2012), it is worthwhile to relate SWF location choice with host country characteristics as this is lacking in current academic research. Especially since country specifics could play an important role in SWF location choice (Murtinu & Scalera, 2016; Quer et al., 2012).

Hence, the main purpose of this thesis is to research whether and how the following host country characteristics influence the location choice of SWFs: (i) population total (market

(8)

corruption rate and (vi) BTAs. These host country characteristics have been selected by using the location pillar of Dunning’s (2000) OLI paradigm. Amongst others, this pillar consists of matters like demand conditions, government induced policies, and regulations (Dunning, 2000). On top of that, these host country characteristics have been considered often in SWF and MNE literature (Murtinu & Scalera, 2016; Zhang & Markusen, 1999; Buettner & Ruf, 2007; Billington, 1999; Cheng & Kwan, 2000; Quer et al., 2012; Chowdhury & Mavrotas, 2006; Strat, Davidescu & Paul, 2015; Mathur & Singh, 2013; Neumayer & Spess, 2005; Büthe & Milner, 2008; Murtinu & Scalera, 2015; Luo, Brennan, Liu & Luo, 2008).

Following Murtinu & Scalera’s (2016) approach, data about SWF investments and their location choice was collected from Bureau van Dijk’s Zephyr, which is a comprehensive database that contains detailed information about worldwide merger and acquisition deals (Bureau van Dijk, 2016, ‘’Overview’’, para. 1). Data about host country characteristics was mainly gathered from the World Bank website through the World Bank Open Data, which is a database that holds data about development in countries around the globe (The World Bank, 2016, ‘’Data’’, para 1.). Data about BTAs was gathered through the World Trade Organization’s RTA and PTA databases with regional- and preferential trade agreements (World Trade Organization, 2016, ‘’Regional trade agreements and preferential trade agreements’’, para 1.). Moreover, following Murtinu and Scalera’s (2016) approach, statistical analysis was done with a logit regression, which reveals the influence of the selected host country characteristics on location choice by SWFs.

1.1 Expected contributions

First of all, authors of recent academic studies argue that SWF location choice is an important field for future research (Murtinu & Scalera, 2016; Shi, et al., 2016). In terms of expected practical contributions, revealing the influence of host country characteristics on SWF

(9)

location choice could provide valuable insights on the extent to which kind of locations are likely to be targeted by SWFs. Furthermore, target countries’ governments and/or firms can use the outcome of this study to predict location choice by SWFs, as the location will tell something about the likelihood of an investment. In turn, this will enable target countries’ governments and/or firms to respond to these investments when required.

In terms of theoretical contributions, this study will contribute to existing theory on geographical distribution of global SWF activities, the effect of geopolitical factors on location choice, investment vehicle strategies and the impact of SWFs’ investments on market value (Fatemi, et al., 2011; Chhaochharia & Laeven, 2009; Bernstein et al., 2013; Shi et al., 2016; Murtinu & Scalera, 2016; Murtinu & Scalera, 2015; Fernandes, 2014), as it will extend the current literature by testing the influence of host country characteristics on location choice. This thesis could therefore serve as input for future academic research on the influence of other host country characteristics on SWFs’ location choices that are outside the scope of this thesis.

1.2 Thesis structure

The first chapter of this thesis introduces the aim and purpose. The second chapter provides an overview and a critical analysis of the existing literature on SWFs, location choice and country characteristics. Continuing with the gap identified from the literature, followed by the conceptual framework and an overview of the hypotheses. The third chapter discusses the methods used. Covering matters such as research strategy, sampling and procedures. Subsequently, the fourth chapter includes an extensive overview and analysis of the results. Finally, the fifth and sixth chapter discuss the results and its conclusions, implications, limitations and directions for further research.

(10)

2. Literature review

This section provides an overview and critical evaluation of the existing literature on SWFs, location choice and country characteristics. Leading to the gap that is identified in the literature, which will be discussed further in section 2.3.

2.1 Sovereign wealth funds

SWFs have grown rapidly in the last few decades (Ngoc, 2015). Ever since, SWFs play an important role in the field of global capital flows as they have access to assets with an estimated worth 7.4 trillion USD (Sovereign Wealth Fund Institute, 2016; Bernstein et al., 2013). Lately, this resulted in increased attention of the media, general public and politicians (Bortolotti et al., 2015). Nevertheless, this domain is lacking academic research (Ngoc, 2015). According to Alhasel (2015), the increased attention is caused by the magnitude of the investments that SWFs have made into some of the largest financial firms of the Western World, such as Citicorp and Merrill Lynch.

2.1.1 Defining SWFs

Bertoni and Lugo (2014, p. 21) define sovereign wealth funds as ‘’government-owned investment vehicles that manage portfolios including foreign financial assets’’. Murtinu and Scalera (2016, p. 3) state that SWFs are funded through ‘’the surpluses given by export commodities, balance of payment surpluses, foreign currency operations, proceeds of privatizations, and/or fiscal surpluses’’. In addition, it is argued that SWFs are either identified as commodity funds, which are financed by the exports of natural resources, or as non-commodity funds, which invest in foreign exchange reserves (Murtinu & Scalera, 2016). Allen and Caruana (2008, p. 5) identify five main SWF types: ‘’(i) stabilization funds, where the objective is to insulate the budget and the economy against commodity (usually oil) price swings; (ii) saving funds for future generations, which aim to convert non-renewable assets

(11)

into a more diversified portfolio of assets; (iii) reserve investment corporations, whose assets are often still counted as reserve assets, and are established to increase the return on reserves; (iv) development funds, which typically held fund socio-economic projects or promote industrial policies that might raise a country’s potential output growth; and (v) contingent

pension reserve funds, which provide (from other sources than individual pension

contributions) for contingent unspecified pension liabilities on the government’s balance sheet’’.

Furthermore, Aizenman and Glick (2008, p. 2) state that SWFs ‘’have been established for various purposes, including stabilization of fiscal revenues, management of inter-generational savings, and sterilization of the effects of balance of payments inflows on domestic inflation’’.

2.1.2 Historical background

Alhasel (2015) states that SWFs have already existed for some time but that the level of visibility has reached an all-time high due to their large investments during the economic crisis and gained media attention. It is argued that this rise of SWFs represents an important structural change to the global economy, since new SWFs will mostly emanate from non-Western countries (Dixon & Monk, 2012).

Kuwait was the first state to establish an actual SWF back in 1953: the Kuwait Investment Authority (Alhasel, 2015). Followed by funds established by the governments of Abu Dhabi and Singapore in the 1970’s (Chhaochharia and Laeven, 2009). The number of SWFs continued to grow rapidly throughout the years to about 74 SWFs (Sovereign Wealth Fund Institute, 2016), which have an estimated worth of more than USD 5 trillion (Bernstein et al., 2013). According to the Sovereign Wealth Fund Institute (2016), the estimated worth is even more today: USD 7.4 trillion, of which USD 4.2 trillion (57%) is oil and gas related.

(12)

2.1.3 SWF strategies

In terms of governance, Aguilera, Capapé and Santiso (2016) propose a framework that describes the strategic governance types of SWFs. The framework states that there are two types of investment motivations: financial and strategic (Aguilera et al., 2016). It is also argued that SWF ownership is either public or private (Aguilera et al., 2016). Furthermore, the four types of strategic SWF governance specified by Aguilera et al. (2016) are shareholder activism (financial investment motivation and public ownership), in-house capabilities (financial investment motivation and private ownership), legitimacy and decoupling (strategic investment motivation and public ownership) and long-term learning (strategic investment motivation and private ownership).

Regarding political influence, Bernstein et al. (2013) stipulate that SWFs that involve politicians tend to invest in industries that show major decreases in price-to-earnings ratios. It is argued that this is done in order to maximize the return on investment (ROI) (Bernstein et al., 2013). Moreover, SWFs with high amounts of political involvement are more likely to invest domestically (Bernstein et al., 2013). Thus, national political pressures force SWFs to invest in local underperforming industries (Bernstein et al., 2013). Knill, Lee and Mauck (2011) also argue that political relations influence where the SWF invests in. However, political influence does not determine how much is invested (Knill et al., 2011). Furthermore, the degree to which the SWF has strong political relations determines the extent to which it invests in private equity (Johan, Knill & Mauck, 2013). Bortolotti et al. (2015) claim that highly politicized SWFs receive lower returns on their stock market investments, which supports the notion that political influence negatively influences firm value and performance.

(13)

2.1.4 Investments vehicles

When looking at SWF investments, researchers find that the relation with politicians and/or political parties influences the likelihood that SWFs will invest in private equity and in sectors with high earnings levels (Johan et al., 2013; Bernstein et al., 2013). SWFs only tend to invest in public equity when the investment is done outside their home country (Johan et al., 2013). Furthermore, SWFs are more likely to invest in less cultural distant countries, while bilateral trade agreements increase the investment values (Megginson, You & Han, 2013; Neumayer & Spess, 2005; Büthe & Milner, 2008). Other country-level determinants that have a positive effect on the probability of SWF investments are strong economic performance and a high degree of openness to trade (Megginson et al., 2013).

In order to do cross-border investments, SWFs could decide to use an investment vehicle (Murtinu & Scalera, 2016). This is done to indirectly enter a foreign country and to avoid and/or reduce the hostility from the host country (Murtinu & Scalera, 2016). Such market reactions are affected by SWF-specific characteristics (Bortolotti et al., 2015). An example of a SWF factor determining market reactions is the degree of transparency (Ngoc, 2015). In addition, Murtinu and Scalera (2016, p. 4) identify three distinct investment vehicle types: ‘’(i) non SWF majority-owned financial vehicles, such as private equity funds, venture capital funds, investment banks, asset management companies, commercial banks, investment management companies, financial branches of big corporations, real estate investment trusts, and investment advisory firms; (ii) non SWF majority-owned corporate vehicles, in the form of non-financial corporations, or companies controlled by public agencies not controlled by SWF or by the government of the country in which the SWF originates; and (iii) other SWF

investment vehicles, including SWF majority-owned financial and non-financial

(14)

Additionally, Murtinu and Scalera (2016) investigated when SWFs would use an investment vehicle in cross-border acquisitions. Their research showed that it is more likely that investment vehicles are used in targeting strategic industries as well as in the presence of fund opacity, fund politicization and majority ownership choices (Murtinu & Scalera, 2016).

2.2 Location choice

According to Murtinu and Scalera (2016, p. 14), location choice by MNEs is ‘’a strategic decision, which is widely studied in international business (IB)’’. Location choice is an important strategic decision within IB as it involves assessing the projected revenues and whether these outweigh the projected costs, while also taking a firm’s foreign experience into account (Benito & Gripsrud, 1992). Nevertheless, little research is done on location choice by SWFs but there is some information on location preferences. For instance, Chhaochharia and Laeven (2009) find that Norwegian SWFs prefer to invest at home and only abroad when investing in large and liquid stock markets. In addition, compared to other SWFs, those from Qatar and Dubai invest a far larger share in the UK than in the US (Chhaochharia & Laeven, 2009). It is argued that this preference for the UK relates to weaker disclosure rules for large shareholdings compared to those in the US (Chhaochharia & Leaven, 2009; Doidge, Karolyi & Stulz, 2007).

In one of the earliest studies on FDI location choice, Hymer (1976) argues that FDI location choice relates to the notion of profit maximization by having an advantage over local firms and selling this advantage to imperfect markets. Subsequently, Johanson & Vahlne (1977) set up a preliminary model for location choice. Rugman, Verbeke and Nguyten (2011, p. 757) called this model the ‘’Uppsala model for international expansion’’. The model states that firms should initially look for foreign markets that have similar country-specific advantages

(15)

(CSAs) (Johanson & Vahlne, 1977). In addition, Rugman et al. (2011, p. 757) refer to the Uppsala model as the ‘’mirror image’’ of Hymer’s (1976) analysis due to their differences. Rugman et al. (2011) extended the Uppsala model by stating that whenever firms decide to locate their business abroad, potential benefits of exploiting firm-specific advantages (FSAs) should be weighed against the costs of operating in unfamiliar markets. In addition, it is stated that firms will have to deal with the liability of foreigners (LOF) when extending their businesses to foreign countries (Rugman et al., 2011). This means that firms have to deal with the impact of cultural, economic, institutional and geographic distance (Rugman et al., 2011). Moreover, location choice decisions are mainly focused on managerial processes whereas country-specific characteristics are often neglected (Rugman et al., 2011).

Rugman et al. (2011) also mention that the Uppsala model builds further upon Dunning’s (2000) eclectic/OLI paradigm. The eclectic/OLI paradigm states that FDI undertaken by MNEs is determined by three interdependent variables: ownership advantages, location advantages and internationalization advantages (Dunning, 2000). It is furthermore argued that the key factors in MNE location choice are endowment effects, agglomeration effects and policy-induced effects (Dunning, 2000).

On top of that, Dunning (2009) states that MNEs have traditional and modern motives for FDI location choice. Strategies belonging to the traditional motives are resource seeking, market seeking and efficiency seeking while strategic asset seeking is claimed to be a modern motive (Dunning, 2009). Benito and Gripsrud (1992) add that a choice for a certain location choice strategy relies on a firm’s experience in doing business abroad as it may decrease the costs and uncertainty of operating in an unknown market. Demirbag and Glaister (2010) support this claim but also state that the knowledge infrastructure, level of wages and political situation in the host country are important factors in location choice decisions.

(16)

Although there have been many studies done that investigated the FDI location choices of MNEs, it is questionable whether their results also hold for SWFs, which is mainly due to an important difference between the two: SWFs are government-owned whereas MNEs are not necessarily government-owned (Alhasel, 2015; Bortolotti et al., 2015; Ngoc, 2015; Fatemi, et al., 2011; Shi et al., 2016; Rudy et al., 2016).

2.3 Gap in the literature

Due to their growing importance, SWFs as well as their impact have recently gained political attention (Bortolotti et al., 2015). However, the domain is lacking academic research (Ngoc, 2015). As described in the previous sections, there are many studies that discuss MNEs’ FDI location choices and host country determinants. Yet, little is known about SWFs’ location choice, as it is a largely neglected research topic (Murtinu & Scalera, 2016).

Murtinu and Scalera (2015) conducted a study on the effect of target industry and location on SWFs and stock prices. However, this study was mainly focused on examining the determinants of SWF investments’ stock prices (Murtinu & Scalera, 2015). Shi et al. (2016) investigated the effect of geopolitical factors on SOEs’ location choice, but this study did not address the influence of country specific characteristics. Furthermore, other existing literature on SFWs has mainly focused on the impact of SWFs’ investments on the market value and performance of listed companies (Murtinu & Scalera, 2015; Fernandes, 2014) and on global geographical distribution of SWF activities (Fatemi et al., 2011; Chhaochharia & Laeven, 2009; Bernstein et al., 2013).

Thus, as Murtinu and Scalera (2016) suggest that it would be useful to relate location choice to country characteristics, this thesis will research the effect of host country characteristics on SWF location choice. Resulting in the main research question:

(17)

2.4 Country characteristics

Many studies in IB have examined the effect of host location determinants on FDI by MNEs. These characteristics can also be framed in the context of SWFs’ investments, as SWFs will have to deal with the same host country characteristics as MNEs. Three main topics have been distinguished in order to classify the different country specifics: economic & regulatory, geo- & demographic and cultural.

2.4.1 Economic & regulatory

According to Cheng and Kwan (2000), governments can positively influence the chance that a firm invests in the country by having preferential policies towards certain firms. Luo et al. (2008) agree by stating that important determinants for FDI are policy incentives but also industrial agglomeration. In addition, Buettner and Ruf (2007) found that host country specifics such as taxing incentives and labour costs are important country characteristics for FDI location choice. Billington (1999) and Strat et al. (2015) agree with most academic research on this topic by stating that tax and interest rates are important determinants for FDI location choice, while also adding unemployment rate. Gaba, Pan & Ungson (2002) argue that host country risk conditions such as an inadequate legal, political and regulatory frameworks might prevent firms from entering foreign markets. Furthermore, GDP is stated to be another important factor in FDI (Zhang & Markusen, 1999; Chowdhury & Mavrotas, 2006).

2.4.2 Geo- & demographic

Cheng and Kwan (2000) found that the larger the market and the better the (knowledge) infrastructure increased the likelihood of incoming FDI. Buettner and Ruf (2007) and Billington (1999) add that the market size, in other words the population of a country, plays

(18)

an important role as well. It may be assumed that the larger the market, the more potential customers one may have.

Goerzen, Asmussen and Nielsen (2013) argue that the presence of global cities, which are characterized by high levels of centrality, influence on the world economy and interconnectedness, are important determinants for MNEs to invest in those particular countries. Typical examples of global cities are London, Paris, Frankfurt, New York, Chicago and Tokyo (Goerzen et al., 2013).

2.4.3 Cultural

Kinoshita and Campos (2003) and Megginson et al. (2013) add that institutions and openness to trade are important factors for FDI location choice. Furthermore, the level of corruption is argued to shape the institutional and political landscape of a country (Murtinu & Scalera, 2016; Mathur & Singh, 2013). Some studies also take cultural distance into account. For instance, Du, Lu and Tao (2012) and Megginson et al. (2013), who state that countries that are culturally distant from the SWF home location are unlikely to be invested in. Additionally, Chhaochharia and Laeven (2009) argue that culture and physical distance play a role in global capital flows. Moreover, Castellani, Jimenez and Zanfei (2013) argue that cultural and social differences as well as institutional proximity are important factors in international FDI (in the case of R&D projects).

Hofstede (1994) stipulates that MNEs operate in different countries and lines of business and thus, that a good understanding of people and their cultural background is therefore necessary. Hence, Hofstede (1994) distinguishes five dimensions that can illustrate cultural differences between countries: power distance, individualism vs. collectivism, masculinity vs. femininity, uncertainty avoidance and long-term vs. short-term orientation.

(19)

2.4.4 Selecting host country characteristics

As described in 2.4.1 to 2.4.3, there are many host country characteristics to be taken into account. However, the sake of narrowing down requires selecting a number of characteristics. As this thesis aims to say something about SWFs’ location choices, Dunning’s OLI paradigm (Dunning, 2000) will be used in order to select a number of host country characteristics. Dunning’s OLI paradigm has three dimensions: the ownership pillar, location pillar and internationalization pillar (Dunning, 2000). Accordingly, this thesis will follow the location pillar, which holds matters such as factors of production, demand conditions, government induced policies, institutional differences and regulations (Dunning, 2000).

More specifically, this thesis focuses on government-induced policies, factors influencing demand and regulations. Hence, the following host country characteristics have been selected: (i) population total (market size), (ii) GDP growth, (iii) tax rates, (iv) unemployment rate, (v) corruption rate and (vi) bilateral trade agreements.

Taxes are governmentally determined regulations (Buettner & Ruf, 2007), while bilateral trade agreements are treaties between governments to stimulate trade amongst countries (Megginson et al., 2013; Neumayer & Spess, 2005; Büthe & Milner, 2008). Furthermore, corruption is caused by a lack factors such as government control and institutional strength (Murtinu & Scalera, 2016; Mathur & Singh, 2013), while market size, unemployment rate and GDP influence demand (Buettner & Ruf, 2007; Billington, 1999; Goerzen et al., 2013; Zhang & Markusen, 1999; Chowdhury & Mavrotas, 2006; Strat et al., 2015), Moreover, these host country characteristics have been used frequently in prior SWF and MNE literature (Murtinu & Scalera, 2016; Zhang & Markusen, 1999; Buettner & Ruf, 2007; Billington, 1999; Cheng & Kwan, 2000; Chowdhury & Mavrotas, 2006; Strat et al., 2015; Mathur & Singh, 2013; Neumayer & Spess, 2005; Büthe & Milner, 2008; Luo et al., 2008).

(20)

2.5 Conceptual framework

As stated in 2.4.4, the location pillar of Dunning’s (2000) OLI paradigm was used to gain knowledge on SWFs’ location choices in the context of FDI. The location pillar includes several factors, of which this thesis will focus on government-induced policies, factors influencing demand and regulations (Dunning, 2000). Based on that, the host country characteristics that will be used are market size, GDP growth, tax rates, unemployment rate, corruption rate and bilateral trade agreements. Thus, the main purpose is to investigate whether and how these host country characteristics influence SWF location choice. This is illustrated in figure 1: the conceptual framework.

Figure 1: Conceptual framework Location choice Tax rate Unemployment rate Market size Corruption rate Bilateral trade agreements GDP growth

(21)

2.6 Hypotheses

This thesis contains six independent variables. Thus, there are six hypotheses defined, each of which is briefly discussed below. Furthermore, table 1 presents an overview of all hypotheses in this thesis.

2.6.1 Economic & regulatory

Zhang and Markusen (1999) state that the greater the GDP (growth) of a host country is, the higher FDI inflows become. In addition, Chowdhury and Mavrotas (2006) also argue for a positive relation between high degrees of GDP and incoming investments. In that sense, GDP growth is an indicator of how well a country’s economy is doing. Therefore, when linking this to SWFs, Fatemi et al. (2011) argue that economic growth in countries is strongly related to the amount of incoming investments from SWFs. On top of that, Fatemi et al. (2011) also state that most SWFs adopt a strategy in which they solely invest in countries that are expected to sustain stable economic growth. Aguilera et al. (2016) agree and add to this that SWFs tend to only invest abroad in economically growing markets in order to obtain long-term economic returns. Since GDP growth is the main indicator of how well an economy is doing, as it stands for the growth in economic value of a country’s produced products and services (D’Alisa, Demaria & Kallis, 2015), this results in the first hypothesis:

H1: GDP growth has a positive relationship with the SWF’s location choice

Several studies argue that favourable taxing incentives and tax rates in general have a positive impact on a country’s attractiveness in terms of inward investments (Buettner & Ruf, 2007; Luo et al., 2008; Billington, 1999). In the case of SWFs, host country tax rates and favourable tax incentives are often linked to the likelihood of incoming investments from those SWFs. For instance, Fleischer (2008) suggests that granting SWFs with favourable

(22)

commenced or completed investing in the host country. Furthermore, governments that make tax exemptions for SWFs as well as lower taxes in general, attract more investments (Fernandez & Eschweiler, 2008). Thus, low tax rates will enable SWFs to increase their returns. Hence, the lower or the more favourable a country’s taxation is, the likelier is it that the SWF invests in the target country, as one of SWFs’ main goals is to maximize return on investment (Bernstein et al., 2013). Therefore, the second hypothesis is as follows:

H2: Tax rate has a positive relationship with the SWF’s location choice

According to Billington (1999), unemployment rate is an important determinant for FDI. Strat et al., (2015) add to this claim by arguing that for some countries unemployment rates could be of negative influence in terms of attractiveness for investments due to dropping demand for products and services. Governments should therefore persistently aim to lower unemployment (Strat et al., 2015). On the contrary, the consequences of unemployment could be much different for SWFs. Some scholars state that unemployment may have a positive effect on inward investments as it fosters labour availability (Billington, 1999; Rowthorn, 1999). Moreover, SWFs may benefit from excessive labour availability by having more power in terms of wage bargaining (Rowthorn, 1999), which in the end, could lower the SWFs labour costs and in turn increase the country’s attractiveness for investments (Billington, 1999). Consequently, this results in the third hypothesis:

H3: Unemployment rate has a positive relationship with the SWF’s location choice Neumayer and Spess (2005) state that bilateral trade agreements provide certain legally binding standards. In addition, bilateral trade agreements prevent the investor from being exposed to political risks and should thus increase investment inflows (Neumayer & Spess, 2005). Megginson et al. (2013) and Büthe and Milner (2008) agree and add that bilateral trade agreements increase investment values as it reassures investors. According to Alhasel

(23)

(2015), this also applies to SWFs, as bilateral trade agreements address the risk of a foreign investment, while already containing predefined solutions to any potential dispute. Hence, the presence of a bilateral trade agreement between the SWF’s home country and the target country reduces the risk for SWF investments, facilitates the entry of SWFs in the host country and mitigates fear surrounding SWF investments. This is all due to the on going political relation between the two countries (Murtinu & Scalera, 2016). Therefore, this results in the fourth hypothesis:

H4: Bilateral trade agreements have a positive relationship with the SWF’s location choice 2.6.2 Geo- & demographic

Cheng and Kwan (2000) argue that the larger the market is, the larger the likelihood is for incoming FDI. Buettner and Ruf (2007), Billington (1999) and Goerzen et al. (2013) support this claim by arguing that large markets hold lots of potential in terms of return on investment and are thus likelier to be invested in. Nevertheless, the question is whether this claim is also applicable to SWFs. Alhasel (2015) argues that SWFs invest in particular markets in order to gain a strategic advantage or political power in that county or region. However, according to Bernstein et al. (2013), the SWF’s main goal for investing in certain countries remains to maximize the returns on their investment. In doing so, large markets are widely acknowledged as markets with high potential in terms generating returns, and are thus a key factor in increasing a country’s attractiveness for (SWF) investments (Quer et al., 2012; Mascarenhas, 1992; Yu, 1990; Rudy et al., 2016). Thus, the larger the market is, the higher the chance for SWFs is to maximize their returns (Rudy et al., 2016). This leads to the fifth hypothesis:

(24)

2.6.3 Cultural

Mathur and Singh (2013) suggest that countries that score poorly on the Corruption Perception (CP) Index, receive low FDI. Therefore, corruption plays an important role in an investor’s decision (Mathur & Singh, 2013). Murtinu and Scalera (2016) support this claim by stipulating that corruption shapes the institutional and political landscape of a country. As SWFs aim to gain political or strategic power in certain regions or industries and to maximize the returns on their investments (Bernstein et al., 2013; Alhasel, 2015), it is absolutely crucial to invest in a politically stable country with strong institutions and governmental control in place (Murtinu & Scalera, 2016; Marthur & Singh, 2013). Corruption is likely to be high in countries where this is lacking (Aizenman & Glick, 2008), as it puts the SWF’s investment at risk because host country governments and politicians might use it to their advantage (Kotter & Lel, 2011). Consequently, this leads to the sixth hypothesis:

H6: Corruption rate has a negative relationship with the SWF’s location choice

Table 1. Overview of hypotheses

H1: GDP growth has a positive relationship with the SWF’s location choice H2: Tax rate has a positive relationship with the SWF’s location choice

H3: Unemployment rate has a positive relationship with the SWF’s location choice

H4: Bilateral trade agreements have a positive relationship with the SWF’s location choice H5: Market size has a positive relationship with the SWF’s location choice

H6: Corruption rate has a negative relationship with the SWF’s location choice

(25)

3. Methodology

In order to ensure validity, this thesis uses multiple cases of SWF investments, making sure that the outcome is about what it states to be about (Saunders & Lewis, 2012). Furthermore, this thesis assures reliability, as similar data collection methods will produce consistent findings on other occasions (Saunders & Lewis, 2012), mainly due to the data being gathered from databases.

Moreover, this thesis has a largely deductive approach as it starts with theory that is followed by data analysis (Saunders & Lewis, 2012). The study is also partly inductive, since it will make use of data concerning specific SWF deals with the aim to add something to broader existing SWF theory (Saunders & Lewis, 2012).

3.1 Research strategy

This thesis is a combination of an exploratory and a descriptive study as its purpose is to answer whether and how host country characteristics influence location choice by SWFs. Answering this question will put the topic in a new light as existing literature paid little attention to the role of host country characteristics in FDI choices by SWFs (Saunders & Lewis, 2012; Murtinu & Scalera, 2016; Quer et al., 2012; Ngoc, 2015).

On top of that, by answering the main research question, this thesis will also provide an accurate representation of the role that host country characteristics have played in FDI decisions by SWF (Saunders & Lewis, 2012). This study will mainly use secondary data from databases and Internet sources to gather information about previous SWF investments.

3.2 Data & sample

The size of the data set is comparable with datasets of other SWF studies (Bernstein et al., 2013; Bortolotti et al., 2015; Murtinu & Scalera, 2016; Dewenter, Han & Malatesta, 2010;

(26)

Knill et al., 2011; Kotter & Lel, 2011; Quer et al., 2012). Following Murtinu and Scalera’s (2016) approach, data about SWF investments and their location choice was gathered through Bureau van Dijk’s Zephyr, which is a comprehensive database containing detailed information about worldwide mergers and acquisition deals (Bureau van Dijk, 2016, ‘’Overview’’, para. 1). Before being able to gather information about the SWF investments, Murtinu and Scalera’s (2016) approach was followed in order to identify the actual SWFs, which was done by using the list of SWFs reported by the Sovereign Wealth Fund Institute (2016) and Truman (2009).

Country specific information was mainly gathered from the website of the World Bank through the World Bank Open Data, which is a database that offers free and open access to data about development in countries around the globe (The World Bank, 2016, ‘’Data’’, para 1.). To gather data about the presence of bilateral trade agreements between the home country of the parent SWF and the host country, the World Trade Organization’s RTA and PTA databases with regional- and preferential trade agreements were used (World Trade Organization, 2016, ‘’Regional trade agreements and preferential trade agreements’’, para 1.). Though, there are some limitations to the use of information from existing databases as they initially may have been gathered for other purposes (Bono & McNamara, 2011).

Previous studies gathered data on SWF investments from a certain period (Quer et al., 2012; Castellani et al., 2013; Murtinu & Scalera, 2015; Murtinu & Scalera, 2016). Thus, the SWF dataset in this study covers the period from 2005 to 2013. Hence, this thesis is a longitudinal study as it studies the location choices of SWFs’ investments over a certain period of time (Saunders & Lewis, 2012). The final sample of this thesis consists of 28 SWFs who have invested in 61 countries, meaning that there were 1708 combinations possible (28*61=1708). Table 2 provides an overview of the 28 SWF parent acquirers and the number of completed investments. Table 3 provides an overview of the number of SWFs investing in each country.

(27)

Table 2. SWFs and number of investments

SWF parent acquirer Number of investments

Abu Dhabi Investment Authority Alaska Permanent Fund Corporation

Alberta Investment Management Corporation Caisse de dépôt et placement du Québec

California Public Employees Retirement System Canada Pension Plan Investment Board

China Investment Corporation China National Social Security Fund DIFC Investments LLC

Dubai International Capital LLC

Gosudarstvennyi Neftyanoi Fond Azerbaijana

The Government of Singapore Investment Corporation Pte Ltd Government Pension Fund – Global

Guardians of New Zealand Superannuation Fund International Petroleum Investment Company Investment Corporation of Dubai

Istithmar PJSC

Khazanah Nasional Bhd Korea Investment Corporation Kuwait Investment Authority Lybian Investment Authority

Mubadala Development Company Pjsc The National Pensions Reserve Fund Qatar Investment Authority

Saudi Arabian Monetary Agency State General Reserve Fund Stichting Pensioenfonds ABP Temasek Holdings Pte Ltd Total 1 1 1 10 4 12 12 1 2 9 1 22 8 2 24 7 8 11 7 4 3 12 1 8 3 4 13 24 215

(28)

Table 3. Number of SWFs investing in each country Country SWFs Country SWFs Algeria Austria Australia Bahrain Belgium Bermuda Brazil Bulgaria Canada Cayman Islands Chile China Denmark Egypt Finland France Germany Greece Hong Kong Hungary India Indonesia Ireland Italy Japan Jordan Kenya Kuwait Libya Luxembourg Malaysia 1 1 7 2 4 4 8 1 9 5 1 7 2 1 1 10 6 1 3 2 7 2 2 7 2 2 1 1 1 2 4 Maldives Mexico Morocco Netherlands New Zealand Norway Oman Pakistan

Papua New Guinea Philippines Portugal Qatar Russia Singapore Saudi Arabia South Africa South-Korea Spain Sri Lanka Sweden Switzerland Taiwan Thailand Turkey

United Arab Emirates United Kingdom United States Uzbekistan Vietnam Virgin Islands Total 1 3 1 5 1 1 3 1 1 1 2 1 4 8 2 4 7 7 1 3 3 2 2 2 8 15 16 1 1 1 215

(29)

3.3 Variables & measurement

This section provides a description of the variables that were used for this thesis and will also explain where these variables were gathered and how they were measured. As previously stated, the data that was used for this thesis covers the period from 2005 to 2013.

3.3.1 Dependent variable

This thesis’ dependent variable is the SWF’s location choice. In order to measure location choice, this thesis follows the same approach as Quer et al. (2012) and Castellani et al. (2013). In doing so, location choice is measured with a dummy that is assigned 1 when the SWF actually completed the investment and 0 if done otherwise, which could be incomplete investments, or assumptions of possible investments (Quer et al., 2012). Data about the SWFs and their location choice was gathered from Bureau van Dijk’s (2016) Zephyr database. 3.3.2 Independent variables

As mentioned in paragraph 2.4.4, there are several country specifics that are considered most often in SWF research. The country characteristics that are used as independent variables, and thus cause a change in the dependent variable (Saunders & Lewis, 2012), in this thesis are: (i) population total (market size), (ii) GDP growth, (iii) tax rates, (iv) unemployment rate, (v) corruption rate and (vi) bilateral trade agreements.

First, the market size/population total is a numerical variable measuring the number of residents (measured in millions) in the host country. This measure is used in order to investigate if the population total of host countries plays a role in SWF location choice. In order to have a normal distribution, an ln-measure was applied to this variable. This data was gathered from World Bank (2016) Open Data Database. GDP growth is a numerical variable, which was collected through the World Bank (2016) Open Data Database, measuring the

(30)

country (D’Alisa et al., 2015). This variable should provide insights into the motives of SWFs investing in growing economies. The tax rate is also a numerical variable gathered through the World Bank (2016) Open Data Database, measuring percentages of tax on commercial profits. This measure will reveal whether taxation has an influence on location choice by SWFs.

Moreover, unemployment rate is measured as a percentage of the total workforce, retrieved from the World Bank (2016) Open Data Database. This variable will examine if the assumed positive effect of unemployment in the host country on SWFs’ location choice is indeed correct. The corruption rate is a numerical value ranging from approximately -2.5 to 2.5, where -2.5 is very corrupt and 2.5 not at all corrupt, unveiling the actual effect of host country corruption on SWF location choice. This estimate reveals the degree to which public power is used for private gain, giving the country’s score in units of a normal distribution (World Bank, 2016). This data was retrieved from the Worldwide Governance Indicators Database of the World Bank (2016). Bilateral Trade Agreements (BTA) is a dummy variable equalling 1 when there is a preferential trade agreement (PTA) or regional trade agreement (RTA) between the home country of the SWF parent acquirer and the host country and equalling 0 when there is no trade agreement. This measure examines whether it is correct that BTAs have a positive influence on SWF location choice. This data was gathered through the World Trade Organization’s RTA and PTA databases (2016).

3.3.3 Control variables

This thesis controls for the SWF’s home country, the use of an investment vehicle by the SWF, the presence of politicians in the SWF’s management, SWF size and SWF industry, as they may influence location choice strategies (Murtinu & Scalera, 2016).

(31)

The SWF size is the firm’s size measured in billions of assets (USD), which was collected from Truman’s (2009) SWF-article. The SWF home country variables are dummy variables that equal 1 when the particular country is the SWF’s country of origin and 0 if the country is not the SWF’s home country. This includes a total of 17 SWF home countries: the United Arab Emirates, the United States of America, Canada, China, Azerbaijan, Singapore, Norway, New Zealand, Malaysia, South-Korea, Kuwait, Libya, Ireland, Qatar, Saudi Arabia, Oman and the Netherlands. This data was gathered from Bureau van Dijk’s (2016) Zephyr Database. Following the approach of Murtinu and Scalera (2016) and Bernstein et al. (2013), the presence of politicians in the SWF’s management is measured with a dummy equalling 1 when there are politicians present in the SWF’s management and 0 if not. This variable relates to the politicization of the SWFs, for which data was gathered by consulting the report of J.P. Morgan (Fernandez & Eschweiler, 2008). According to Bernstein et al. (2013) and Bortolotti et al. (2015), the active presence of politicians in the SWF’s management might lead to investment strategies that do not maximize shareholder value, by aiming at strategic advantages in particular industries, which in turn raise opposition in the target countries (Murtinu & Scalera, 2016).

Furthermore, the SWF source of funds are five dummy variables equalling 1 when the SWF’s main source of funds is either natural resources, employee contributions, foreign exchange reserves, fiscal surpluses or government enterprises. More specifically, 1 represents either natural resources, which are funds gained from access to or ownership of raw materials such as oil as the SWF’s main source of funds (World Trade Organization, 2010), employee contributions, funds gained from employee retirement plans (Choi, Laibson, Madrian & Metrick, 2002), foreign exchange reserves, funds obtained from gold holdings or convertible foreign currencies (Polterovich & Popov, 2003), fiscal surpluses, funds from a budget surplus

(32)

funds that are obtained from government-supported business activities (Harris & Wiens, 1980). This data was also gained from Truman’s (2009) SWF-article.

The use of an investment vehicle is a control variable too, as this thesis controls for the role of investment vehicles in location choice decisions. More specifically, SWFs could opt for an investment vehicle when investing abroad, which is done to enter the foreign market indirectly (e.g. by means of a foreign affiliate or acquiring an established local firm), and thus avoiding the host country’s opposition and hostility (Murtinu & Scalera, 2016). The investment vehicle variable is measured with a dummy equalling 1 when the SWF used an investment vehicle and 0 if not. Following Murtinu and Scalera’s (2016) approach, this data was gathered through Bureau van Dijk’s (2016) Zephyr Database.

Table 4 presents an overview of all the variables that are used for this thesis, as well as their definitions and source.

(33)

Table 4. Definition and source of variables

Variable Definition Source

Dependent variables Location choice Independent variables Market size GDP growth Tax rate Unemployment rate Corruption rate

Bilateral trade agreements

Control variables SWF home country Politicians SWF size SWF source of funds Investment vehicle

A dummy equalling 1 when the SWF completed the investment Number of residents (millions) in the host country

Current % of growth in value of all output in the host country Current % of tax on commercial profits in the host country Current % of unemployment of the host country’s total

workforce

Value ranging from -2.5 to 2.5, where -2.5 is a very corrupt host country and 2.5 not corrupt A dummy equalling 1 when there is a BTA between the SWF’s home country and the host country

Dummy variables equalling 1 when the country is the SWF’s home country (see 3.3 for all home countries)

A dummy equalling 1 when there are politicians in the SWF’s management

SWF size in billions of assets (USD)

Dummy variables equalling 1 when the SWF’s main source of funds is natural resources, employee contributions, foreign exchange reserves, fiscal

surpluses or government enterprises

A dummy equalling 1 when the SWF used an investment vehicle

Bureau van Dijk (2016)

World Bank (2016) World Bank (2016) World Bank (2016) World Bank (2016)

World Bank (2016)

World Trade Organization (2016)

Fernandez and Eschweiler (2008)

Truman (2009) Truman (2009)

(34)

3.4 Procedure & model specification

In order to answer the research question, this thesis will examine the hypotheses stated in section 2.6 by using a research strategy that is able to test hypotheses (Saunders & Lewis, 2012). This testing involves using the IBM SPSS Statistical program. This program enables one to do statistical analysis through various types of tests (Saunders & Lewis, 2012).

First of all, raw data has been collected from several sources, which are described extensively in paragraph 3.3. Subsequently, this data has been quantified and put in SPSS. Second, before being able to run the statistical test for the hypotheses, the independent variables were checked for excessive correlation (above 0.7) and a normal distribution with descriptive statistics (Saunders & Lewis, 2012). The quantification of the data, to e.g. dummies, allowed for statistical interpretations through SPSS. Then, different statistical tests were performed in SPSS to examine the relations between the independent variables and dependent variable. Lastly, a regression analysis was performed in order to test the conceptual model as a whole. By following Quer et al.’s (2012) and Murtinu and Scalera’s (2016) approach, this thesis will use a logit regression analysis. According to Powers and Xie (1999), a logit regression analysis is appropriate due to the fact that the dependent variable is a dummy variable which has only two possible alternatives, namely yes (=1) or no (=0). This is very common in IB and more specifically, in prior location choice studies, such as Quer et al. (2012). Hence, as suggested by Powers and Xie (1999), the logit regression model is specified as follows: Logit (p location choice) = log (p location choice/1 - p location choice) = β0 + β1 Market size1 + β2 GDP growth1 + β3 Tax rate1 + β4 Unemployment rate1 + β5 Corruption rate1 + β6 Bilateral trade agreements1 + β7 SWF home country1 + β8 Politicians1 + β9 SWF size1 + β10 SWF source of funds1 + β11 Investment vehicle1

(35)

4. Results

This section consists of a description of the results, which will commence with a normality check and the descriptive statistics of the dependent, independent and control variables that were used for this thesis, followed by the correlations of these variables and the regression analysis. In addition, the SWF home country and source of funds dummies are excluded from the descriptive statistics and correlation tables to preserve an orderly overview, but they will be taken into account in the regression analysis.

4.1 Normality check

Before running the descriptive statistics, the independent variables in the dataset that were used for this thesis were checked for a normal distribution. Normal distribution was checked by looking at Skewness and Kurtosis, which are measures that give indications for a normal distribution (Groeneveld & Meeden, 1984). The values of these measures should be close to 0, while values between -2 and +2 are considered acceptable to prove for a normal distribution (George & Mallery, 2010; Field, 2009).

Research showed that market size, tax rate, corruption rate and BTA all have a normal distribution, as their Skewness and Kurtosis values were between -2 and +2, and within the range of the lower and upper confidence interval of 95%. Furthermore, it was revealed that GDP growth and unemployment rate are not normally distributed as their Kurtosis was above 2. GDP growth’s Skewness is -1.298 and its Kurtosis is 4.888, while unemployment rate’s Skewness is 1.806 and its Kurtosis is 3.477. According to van Dalen and de Leede (2014), this can be caused by a number of outliers. Therefore, an outlier analysis was conducted to check for the presence of data entry or processing errors (van Dalen & de Leede, 2014), which was not the case. Subsequently, the outliers were excluded from GDP growth’s and

(36)

4.2 Descriptive statistics

As stated in table 5, 1596 out of the initial 1708 observations are determined to be valid (listwise). Meaning that 112 observations were ruled out due to missing values. In this final sample of 1596 SWF investments, 13% of them were actually realized, meaning that location choice was completed. In regards to market size, the smallest belong to the Cayman Islands, Bermuda and the Virgin Islands with about 100 thousand residents. China has the largest market size with a population of 1.37 billion. GDP growth ranges from -10.2 to 8.5, with an average GDP growth of about 2.7%. The average tax rate is estimated to be almost 39%, while the average unemployment rate lies at 7.7%.

Furthermore, when looking at the World Trade Organization’s (2016) World Governance Indicators of corruption, the average corruption rate is estimated at 0.53. From the 1596 final SWF investments, 38% could have benefitted from a bilateral trade agreement between the SWF home country and target country, regardless of whether the investment was realized or not. In addition to that, the presence of politicians in the managing bodies of the SWF is confirmed in 43% of the cases. The smallest SWF is Gosudarstvennyi Neftyanoi Fond Azerbaijana, with 2 billion (USD) worth of assets, while the largest SWF in terms of assets is the Abu Dhabi Investment Authority with 875 billion (USD). Lastly, only 6% of the investments were done by means of an investment vehicle.

Table 5. Descriptive statistics

Variable N Min Max Mean S.D.

1. Location choice 2. Market size 3. GDP growth 4. Tax 5. Unemployment 6. Corruption 7. BTA 8. Politicians 9. SWF size 10. Vehicle 1596 1596 1596 1596 1596 1596 1596 1596 1596 1596 0 -0.9 -10.2 11.3 0.3 -1.6 0 0 2 0 1 7.2 8.5 72.7 26.3 2.3 1 1 875 1 0.130 3.146 2.723 38.684 7.726 0.530 0.380 0.430 131.50 0.060 0.333 1.656 2.988 14.984 5.457 1.102 0.487 0.495 181.252 0.228

(37)

4.3 Correlations

Table 6 presents an overview of the Pearson correlation coefficients between the dependent, independent and control variables of this thesis. Location choice has a significant and a tendency towards a positive relation with market size (r = 0.147, p = 0.01) and corruption rate (r = 0.099, p = 0.01). Location choice also has a significant and positive relation with investment vehicle (r = 0.633, p = 0.01). Besides market size having a significant relation with location choice, it also has a significant and tendency towards a positive relation with GDP growth (r = 0.155, p = 0.01), bilateral trade agreements (r = 0.113, p = 0.01) and investment vehicle (r = 0.065, p = 0.01). Furthermore, market size is also positively and significantly related to taxation (r = 0.558, p = 0.01) and has a tendency towards a negative and significant relation with corruption rate (r = -0.376, p = 0.01). Next to having a significant relation with market size, GDP growth has a significant and a tendency towards a negative relation with unemployment rate (r = -0.266, p = 0.01) and corruption rate (r = 0.069, p = 0.01).

As previously suggested, tax rate is significantly related to market size. On top of that, tax rate is also significantly and positively related to unemployment rate (r = 0.183, p = 0.01) and corruption rate (r = 0.208, p = 0.01). In addition to unemployment rate having a significant relation with location choice, market size, GDP growth and tax rate, it is also significantly and slightly negatively related with corruption rate (r = -0.221, p = 0.01) and bilateral trade agreements (r = -0.091, p = 0.01). Moreover, corruption rate is also significantly and slightly negatively related with bilateral trade agreements (r = -0.175, p = 0.01). The variable bilateral trade agreement is significantly and somewhat negatively associated with the presence of politicians in the SWF’s management (r = -0.135, p = 0.01). Lastly, the presence of politicians is significantly and negatively associated with SWF size (r = -0.381, p = 0.01).

(38)

Table 6. Correlation matrix Variable 1 2 3 4 5 6 7 8 9 10 1. Location choice 2. Market size 3. GDP growth 4. Tax 5. Unemployment 6. Corruption 7. BTA 8. Politicians 9. SWF size 10. Vehicle 1 0.147* -0.041 0.064 -0.021 0.099* 0.043 0.011 -0.001 0.633* 1 0.155* 0.558* -0.046 -0.376* 0.113* 0.000 0.000 0.065* 1 -0.009 -0.266* -0.069* 0.038 0.000 0.000 -0.040 1 0.183* 0.208* -0.062 0.000 0.000 0.050 1 -0.221* -0.091* 0.000 0.000 -0.017 1 -0.175* 0.000 0.000 0.054 1 -0.135* 0.038 0.018 1 -0.381* 0.046 1 -0.024 1 * Significant at P-value < 0.01 (two-tailed test)

(39)

4.4 Regression analysis

As done by Quer et al. (2012), several logit regression models were carried out for this thesis’ regression analysis. The first model solely covers the control variables, after which each model introduces an independent variable. The independent variables all refer to the SWFs’ target countries. Table 7 presents an overview of the results the regression analyses.

In the first model of the logit regression, the control variables (SWF home country, presence of politicians, SWF size, SWF source of funds and the use of an investment vehicle) were entered. This model solely including control variables is significant with a Chi-Square of 557.616 and p = 0.000. It turned out that the overall predictive power of model 1 is 92.8%. The first independent variable, GDP growth, was added in the second model of the logit regression analysis. Model 2 controls for SWF home country, politicians, SWF size, SWF source of funds and investment vehicles. This model is significant as the Chi-Square is estimated at 554.076 and the p-value at 0.000. Model 2 has a Nagelkerke R2 of 0.528, which is slightly higher than the first model. Moreover, the predictive power of model 2 is also estimated at 92.8%. GDP growth’s coefficient is set at p = 0.284 and is thus not significant (β = -0.033). Hence, hypothesis 1 is not supported.

The third model includes tax rate as independent variable, controlling for SWF home country, politicians, SWF size, SWF source of funds and investment vehicles. This model is significant with a Chi-Square of 528.023 and p = 0.000. Just like the second model, model 3 has a Nagelkerke R2 of 0.528, while its predictive power is also estimated at 92.8%. Tax rate is statistically significant as its p-value is 0.090 (β = -0.012), supporting the second hypothesis.

(40)

model is significant with a Chi-Square of 553.514 and p-value of 0.000. Resulting in a Nagelkerke R2 of 0.527, which is slightly lower than the third model. The predictive power of this model is yet again set at 92.8%. Unemployment rate is statistically insignificant as its p-value is 0.463 (β = 0.001), not supporting the third hypothesis yet.

The fifth model includes BTAs in the analysis, controlling for SWF home country, politicians, SWF size, SWF source of funds and investment vehicle. Model 5 is significant with a Chi-Square of 558.511 and p = 0.000. This model accounts for a Nagelkerke R2 of 0.525, which is lower than the fourth model. The overall predictive power of model 5 is also 92.8%. Furthermore, BTA is statistically insignificant, since the p-value is estimated at 0.342 (β = 0.198). Therefore, the fourth hypothesis is not yet supported.

The sixth regression model includes market size in the analysis, controlling for SWF home country, politicians, SWF size, SWF source of funds and investment vehicle. The results of the Chi-Square show that the model is significant at 565.862 and p = 0.000. The Nagelkerke R2_{of this model is estimated at 0.537 and the model’s overall predictive power is 92.8%.} Market size is highly significant as the p-value is 0.000 (β = 0.001), supporting the fifth hypothesis.

The seventh regression model adds corruption rate to the analysis, controlling for SWF home country, politicians, SWF size, SWF source of funds and investment vehicle. The model is significant with a Chi-Square of 571.401 and p = 0.000. This model has a Nagelkerke R2_of 0.536), slightly lower than model 6. The overall predictive power of model 7 is 92.8%. Moreover, corruption rate is highly significant with a p-value of 0.000 (β = 0.344). However, the results do not support the sixth hypothesis.

The eight and last regression model includes all independent variables and control variables. First of all, the Chi-Square of this model indicates that it is significant at 622.674 and p =

(41)

0.000. Furthermore, the model has a Nagelkerke R2 of 0.606 and has an overall predictive power of 93.2%, which are both higher than all preceding models. In this model, five out of six independent variables are statistically significant: tax rate (p = 0.013, β = 0.024), unemployment rate (p = 0.020, β = 0.050), BTA (p = 0.090, β = 0.421), market size (p = 0.000, β = 0.838) and corruption rate (p = 0.000, β = 0.943). The last model supports each hypothesis, except for hypotheses 1 and 6.

As stated in table 7, not every model has the same amount of observations, which is due to missing values for some of the variables. All regression models were therefore also conducted with 1596 (the lowest) observations, and it turned out that these results were in line with the ones in table 7.

Host country characteristics as determinants for sovereign wealth funds’ location choice

Master Thesis