The effect of Foreign Direct Investment on stock market development within Sub-Saharan Africa

(1)

The effect of Foreign Direct Investment

on

stock market development within

Sub-Saharan Africa

Abstract

This research examines if a long-run linkage between Foreign Direct Investments and stock market development exists within Sub-Saharan Africa. Leading to this research was a paper earlier published in 2009 by Adam and Tweneboah who argued that Foreign Direct Investment could affect stock market development. Their research suggests it does for Ghana. Subsequently, a Johansen Cointegration test was being used for both Zimbabwe as well as Kenya, followed by a Vector Error Correction Model to see if a long-run linkage does exists for both countries. Market capitalization as a % of GDP was hereby being used as a proxy for stock market development. Eventually, a long-run linkage between Foreign Direct Investment and stock market development was being found in both cases. However, it is important to note that the direction of causality is ambiguous within the linkage. Further research is necessary if either Foreign Direct Investment affects stock market development or vice versa, or to reconfirm that the direction of causality is indeed bivariate.

Dave (J.J.E) van der Molen 29-06-2015

10012699 Supervisor: prof. J.E. Ligterink

(2)

1

Statement of originality

This document is written by student J.J.E van der Molen who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is 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)

2

In 2012 the Financial Times published an article about Foreign Direct Investment, FDI, in Sub-Saharan Africa, which was based on a report being published earlier by EY (Financial Times, 2012). According to this article FDI activity in Sub-Saharan Africa, SSA, increased with 27% in 2011, totaling to an amount of $82bn. Also, it was predicted that this amount would grow to $150bn in 2015. Although the predictions have not been reached, the article shows the interest within the region, which grew substantially during last decade. It is important to note, however, that SSA FDI still compromises to a small portion of total FDI activity worldwide and subsequently there is still a lot to gain (Asiedu, 2002).

Several factors are attracting more investment into SSA, namely the improvement of infrastructure, stability in macroeconomic factors and easing of legislation (EY, 2014). According to Asiedu (2006) these are, amongst others, considered as the biggest constraints for companies to invest in foreign countries. Attracting FDI is important for developing countries since it provides several benefits. Asiedu (2002) discusses that FDI has several spillover effects which, amongst others, increases employers productivity, entrepreneurship activity and resource allocation. Especially the latter is important, due to the millennium goals of reducing poverty rates by half which were set to be reached by SSA in 2015 (Asiedu, 2006; Dupasquier&Osakwe, 2006).

However, there are some difficulties with reaching the millennium goals due to a shortage of funds to meet the investment needs, which are subsequently caused by low levels of savings (Gwenhamo, 2011). So did the Economist (2015) published an article saying that there is great need for capital by local entrepreneurs within SSA. Gwenhamo (2011) acknowledges that FDI inflows are therefore considered important for overcoming the domestic resource gap. Here, FDI is assumed to reinforce domestic capital and thereby stimulating the productivity of domestic investments (Borensztein, Gregorio&Lee, 1998). However, for resource allocation to be optimal a developed stock market will be beneficial, considering

stock markets intermediate funds towards investment

projects (Adam&Tweneboah, 2009) – which is acknowledge by

Caporale, Howells & Soliman (2004). Stock market development can thereby increase liquidity and minimizes efforts from companies to raise capital. Considering the importance of a developed stock market and SSA increasing its efforts to attract FDI leads to the question if FDI can have an effect on stock market development; is there a long-run linkage between both?

In order to investigate this linkage between FDI and stock market development there are several other linkages to consider, which will compromise to a chain of linkages. So does Alfaro, Chanda & Kalemli-Ozcan (2004) research provides empirical results for a linkage between FDI activity and economic growth. In addition, Levine and Zervos (1996) focuse on the linkage between economic

(5)

4 growth and stock market development, for which they find evidence supporting a positive linkage. Furthermore, Adam and Tweneboah (2009) acknowledge previous investigated linkages and goes a step further by indicating that FDI activity can have an impact on stock market development. By doing so, they take economic growth out of the overall linkage. In short, in order to investigate the linkage between FDI activity and stock market development, the following has to be covered and discussed according to previous research:

1. Effect of FDI on the economy (and vice versa)

2. The effect of the economy on stock market development (and vice versa)

Taking economic growth out of linkage will lead to the following one:

3. Effect of FDI on stock market development (and vice versa)

Although the third linkage is the focus point of this research, the other linkages will have to be discussed before coming to the effect of FDI on stock market development. The first two linkages set the foundation and cannot be simply neglected. The intention of this research is to find a direct link between FDI and stock market development, but both are also caused by other factors – which are indicated by linkages one and two. All this causes it to be a difficult one to investigate, since possible found results can also be caused by for example economic growth – although it is taken out of the linkage.

Empirical analysis will be done to investigate the linkage between FDI net inflows and market capitalization as a % of GDP. A Johansen Cointegration test will be used to check for cointegration factors, which will be followed by a vector error correction model to see if a long-run linkage exists. By doing so, this research contributes to earlier findings by Adam and Tweneboah and extends their research to more countries. It thereby also acknowledges earlier research about the previous two linkages mentioned.

The remainder of the research will be structured as follows. In the second section, the theoretical framework, both concepts FDI and stock market development will be explained extensively. Following the theoretical framework, all previous mentioned linkages will be covered throughout the conceptual framework – which will set the foundation for the remainder of the research. Subsequently the methodology will be covered, followed by the results and discussion. Finally, a conclusion will be derived to sum up all findings and implications.

One thing to bear in mind is that goal of the research is to find a long-run linkage and not a relationship. Since the time-series of two variables will be compared, the concept of a linkage will be used to interpretate the results.

(6)

5

2. Theoretical framework

In this paragraph concepts like Foreign Direct Investment, FDI, and stock market development will be covered. It is important to first define both concepts before moving on to the conceptual framework. While covering FDI, the factors attracting FDI will also be covered. Also, in what way FDI could possibly effect stock market development will be briefly discussed.

2.1 Foreign Direct Investment

According to Alfaro, Chanda & Kalemli-Ozcan (2004) the United Nations Conference on Trade and Development defines FDI as an investment involving a long-term relationship and reflecting a lasting interest and control of a resident entity in one economy in an enterprise resident in an economy other than that of the foreign direct investor. Simply said, FDI is a capital inflow provided by a foreign investor. The world bank has a more specific definition for FDI, namely: ‘Foreign direct investment

are the net inflows of investment to acquire a lasting management interest (10 percent or more of voting stock) in an enterprise operating in an economy other than that of the investor. It is the sum of equity capital, reinvestment of earnings, other long-term capital, and short-term capital as shown in the balance of payments.’According to Asiedu (2002) FDI is not only the money being brought into the country by foreign companies looking for expanding opportunities, but also managerial skills, employment and technology. Also, in her research she mentions two different types of FDI, namely market-seeking and non-market seeking. She explains market-seeking FDI as having the objective of serving the domestic market. Goods are being produced in the host country and sold in the local market. As a consequence demand is determined on a local scale, making market-seeking FDI less likely in small economies. Non-market seeking FDI refers to the fact that goods are being produced in the host country, but sold abroad. Here demand is not being influenced by local demands but more by the ease with which firms can export their products, making it a better explanation for FDI activities in poor countries with small economies. According to Asiedu FDI in SSA is mainly non-market seeking, due to the focus of FDI in natural resources which can be found in SSA.

Lately, FDI has been increasing rapidly in developing countries. Alfaro, Chanda & Kalemli-Ozcan (2004) mentions that in 1998, FDI accounted for more than half of all private capital flows to developing countries. As in their belief this change in FDI inflows has to do with a shift among policymakers, in developing countries, in an attempt to attract more FDI. According to Alfaro et al. this has been especially the case since the 1980s debt crisis and the turmoil in emerging countries during the 90’s. Nair-Reichert and Weinhold (2001) acknowledges what Alfaro et al. are saying in a way by showing that since late 1980, FDI in developing countries has increased by 17 percent each year. In 1970 for example, total FDI to developing countries accounted for $3.85 billion and in 2013

(7)

6 for $778.3 billion (UNCTADstat, 2015). Looking at the net FDI to developing countries in 2013, it accounted for more than 50% of the total FDI and net FDI to Africa accounted for 7%.

Attracting FDI is important for developing countries since it provides positive effects. Among the possible effects Alfaro, Chanda & Kalemli-Ozcan (2004) discusses productivity gains, technology transfer, the introduction of new processes, managerial skills, and employee training. Due to these possible positive effects there has been increasing competition among developing countries to attract FDI. Asiedu (2006), who concentrates on Sub-Saharan Africa, addresses the question if FDI is dependent on natural resources and market size, or that government policy is also an important factor in attracting FDI. She emphasizes that it is especially important for Sub-Saharan Africa, SSA, to attract FDI, since it will help in achieving its Millennium Development Goal of reducing poverty rates by half in 2015. Also, it is mentioned in the New Partnership for Africa’s development (NEPAD) declaration that in order to achieve this goal the region needs to fill an annual resource gap of US$64 billion, which is about 12 percent of GDP (Asiedu, 2006). Here Asiedu mentions the fact that due to low income levels, domestic savings and accessibility to international capital markets, this additional financial resource has to largely come from FDI. She backs up her findings when comparing net official loans being used and FDI capital. In 1980-1984 FDI contributed to 7% of total foreign investment and in 1995-1999 it grew to a share of 27%. Asiedu conclusions are acknowledged by Dupasquier & Osakwe (2006), who also emphasizes the importance of FDI to achieve its Millennium Development Goal. Furthermore, according to Asiedu (2002) during 1980-1989 and 1990-1998 FDI in SSA grew by 59%. This in comparison with an increase of 5200% for Europe and Central Asia, 942% for East-Asia and Pacific for example and 672% for all developing countries shows the struggles of SSA with attracting FDI. In the millennium development goals report 2014 it shows that SSA did not managed to reach their goal yet of reducing poverty rates by half, showing again the importance of attracting more FDI (United Nations, 2014). SSA managed to reduce their poverty rate by 8% in 2010, from 56% to 48%, while the poverty rate worldwide reduced by half, from 36% to 18%.

In her research Asiedu (2006) tries to find factors which deter FDI for the region of SSA. Several surveys she discusses shows that companies find corruption, weak infrastructure, taxes and regulation, macro and political instability amongst the biggest constraints when it comes down to FDI. Her findings suggest that FDI can also be attracted by small countries, lacking natural resources, by improving their institutions and policy environment. Also, regional economic cooperation may enhance FDI, since it will increase their market size. It shows that attracting FDI is not only affected by exogenous factors. Dupasquier & Osakwe (2006) also thinks the above factors will have a negative influence on FDI activity within SSA, but also includes increased competition among FDI, increased volatility for countries with more natural resources, external debt and poor marketing as important factors. After their empirical analysis they find the following factors having a significant influence:

(8)

7 economic growth, inflation, openness, international reserves and natural resource availability. The results from Asiedu (2006), however, were obtained by distributing surveys to several international companies concerning questions about SSA only. In an earlier research performed by Asiedu (2002) she discusses that infrastructure and openness to trade both have a smaller marginal benefit in comparison with non-SSA countries. Dupasquier & Osakwe (2006) support her findings partial by stating that political rights and infrastructures are inconsequential for SSA countries in comparison with other developing countries. Both research’s show there is a difference between the factors influencing FDI for SSA in comparison with other developing countries, something Asiedu also addresses in her research in 2006. The starting point for her research in 2006 was the fact that several investor surveys indicated that factors attracting FDI to Africa are different from the factors that drive FDI into other regions. Also, her empirical research earlier in 2002 indicates there is a difference indeed. Asiedu uses a dummy variable into her regression analysis to account for the fact if a country is from continental Africa or not. She found a substantial difference in the explanatory factor of the variables, indicating a difference among influence for African countries. The results of her research shows that on average FDI/GDP for a country in SSA is about 1.3% less than that of a comparable country outside the region. However, Asiedu(2002) does not include real wages, trade policies and tax legislation into here research, because according to her they are not readily available.

Stapper (2009), however, did concentrated on how tax policies can affect FDI for SSA. He used eight case studies to investigate the relationship between corporate tax policies and FDI flows. According to his findings there is no significant correlation between corporate taxes and FDI. However, several attributes of tax policies have a correlation with FDI activity. According to Stapper the amount of double taxation treaties between countries, to avoid any double taxation, do have a positive effect on FDI as a percentage of GDP. Furthermore, the existence of Export-Processing Zones has a negative effect on FDI activity. Export-Processing Zones refers to special areas within a country where governments try to stimulate foreign activity, it is a specific example of a Free Trade Zone. According to Dupasquier & Osakwe (2006) there are several benefits to be obtained from FDI activity, they mention the following: employment generation and growth, supplementing domestic savings, integration into the global economy, raising skills of local manpower, transfer of modern technologies, enhanced efficiency and other spillover effects. As mentioned earlier, Asiedu (2002) adds that managerial skills, employment and technology will contribute to the positive effects of FDI. These benefits will affect the growth of the economy according to the authors. Alfaro, Chanda & Kalemli-Ozcan (2004) go one step further and think that the interaction between financial markets and FDI spillovers is an important one for exploiting this economic growth. According to them a developed financial system is necessary to fully exploit the FDI inflows. Adding to that, the lack of

(9)

8 financial markets also can constrain potential entrepreneurs. In conclusion, developed financial markets can be the link between foreign activity and the domestic market according to Alfaro, Chanda & Kalemli-Ozcan. Furthermore, Adam and Tweneboah (2009) mentions that FDI to developing economies in West Africa increased from $1.9 billion in 1995 to about $15.8billion in 2006. The market capitalization of emerging market countries almost tripled from about $2 trillion to about $5 trillion over the same period, indicating a certain impact of FDI according to them. In the conceptual framework, which will be covered next, the linkage between FDI and stock market development will be extensively discussed.

2.2 Stock market development

Now that FDI has been extensively covered it is time to look at stock market development. According to Garcia and Liu (1999) stock market development is a multidimensional concept. According to them it is usually measured by stock market size, liquidity, volatility, concentration, integration with world capital market and the regulations and supervision in the market. Demirgiic-Kunt and Levine (1996) agrees with Garcia and Liu (1999) that stock market development is multidimensional, with not one measurement to cover and measure the whole concept. However, they also mention that, despite not having one measurement to be comprehensive, analysts frequently use market capitalization as a percentage of GDP as a proxy for stock market development. According to Demirgiic-Kunt and Levine the assumption behind market capitalization is that market size is positively correlated with the ability to mobilize capital and diversify risk. Furthermore, Billmeier and Massa (2009) find that developed legal systems are crucial for the development of capital markets. They argue that the existence of transparency and regulations increases investor confidence. They also mention the macroeconomic variables previous mentioned by Garcia and Liu (1999) as factors influencing stock market development. Also, according to Billmeier and Massa it is important to note that resource rich countries rely less on government efforts to improve institutions, where they attract FDI due to their resources. However, in the long-run institutional reforms becomes more important for those countries as well.

Besides defining a measurement for stock market development Demirgiic-Kunt and Levine mingles themselves into the discussion if stock markets and debt markets are substitutes or complements from each other. Several research suggests that the development of stock markets is correlated with the development of financial intermediaries, with stock markets and bank acting as complements from each other when it comes down to capital sources. Following previous research, Demirgiic-Kunt and Levine include a proxy of financial intermediaries development into their

(10)

9 regression analysis. They find that if domestic credit to the private sector divided by GDP increases by one percentage point, market capitalization increases by 0.527 percentage point. Thus, financial intermediary development promotes stock market development according to Demirgiic-Kunt and Levine. Here, financial intermediary development is a combination of the development of banks, nonbank financial corporations, insurance companies and private pension funds.

Yartey and Adjasi (2007) concentrates purely on Sub-Saharan Africa when it comes down to stock market development. They mention the fact that only five stock market exchanges existed in 1989 and three in North-Africa. This number grew to 23 active stock exchanges, from which 21 are a member from the African Securities commission (Wikipedia, 2015). However, Yartey and Adjasi stresses that not on every stock exchange actively trading occurs. According to them trading only occurs in a few stocks on several stock exchanges, which contribute to a large fraction of the market capitalization. Also, there are some problems with information and disclosure shortcomings and the supervision and regulatory authorities are often far from adequate. They emphasize the liquidity problems some stock exchanges faces. This liquidity problem has to do with the size of the stock exchange and the low daily trading volume. For investors this imposes a risk factor to take into account. Important to note is that according to Demirgiic-Kunt and Levine (1996) Nigeria and Zimbabwe does, amongst other countries outside the African continent, had the worst stock market development, looking at the period 1986-1993.

Before exploring the linkage between FDI and stock market development it is important to first see which benefits can arise from FDI. According to Dupasquier & Osakwe (2006) there are several benefits to be obtained, they mention the following: employment generation and growth, supplementing domestic savings, integration into the global economy, raising skills of local manpower, transfer of modern technologies, enhanced efficiency and other spillover effects. As mentioned earlier, Asiedu (2002) adds that managerial skills, employment and technology will contribute to the positive effects of FDI. These benefits will affect the growth of the economy according to the auteurs. Alfaro, Chanda & Kalemli-Ozcan (2004) go one step further and think that the interaction between financial markets and FDI spillovers is an important one for exploiting this economic growth. According to them a developed financial system is necessary to fully exploit the FDI inflows. Adding to that, the lack of financial markets also can constrain potential entrepreneurs. In conclusion, developed financial markets can be the link between foreign activity and the domestic market according to Alfaro, Chanda & Kalemli-Ozcan. Furthermore, Adam and Tweneboah (2009) mentions that FDI to developing economies in West Africa increased from $1.9 billion in 1995 to about $15.8billion in 2006. The market capitalization of emerging market countries almost tripled from about $2 trillion to about $5 trillion over the same period, indicating a certain impact of FDI according to them.

(11)

10

3. Conceptual framework

Now that FDI and stock market development both as concepts are covered, this paragraph is meant to derive where the linkage between FDI and stock market development comes from. Eventually this linkage will be the starting point for the statistical research and so it is important to show its origins and theoretical implications. According to Adam and Tweneboah (2009) there has been considerable research on determinants for the development of the financial sector. Previous research shows that financial markets tend to develop as the economy grows and financial reform progresses. Furthermore, Adam and Tweneboah mention that a large number of empirical studies suggests that FDI is an important source of capital, which complements domestic sources of capital and boosts overall economic growth in host countries. Following the suggestions and conclusions from previous research, Adam and Tweneboah suggest that there are three linkages one can observe, namely:

1. FDI stimulates economic growth

2. Economic growth stimulates stock market development 3. FDI stimulates stock market development

The last linkage is a combination of the previous two and is the focus point of this research, here Adam and Tweneboah take out economic growth to see if there is a direct linkage between FDI and stock market development. Following their procedure, this paragraph will first cover the first two linkages before coming to the linkage between FDI and stock market development.

3.1 FDI and economic growth

Alfaro, Chanda & Kalemli-Ozcan (2004) focusses on the link between FDI and economic growth. The role of financial markets is incorporated within this link by them to see how it affects growth. They use a cross-sectional regression analysis, where they incorporate financial markets as an intermediary and a control variable to see how financial markets affects the relation between FDI and economic growth. According to them one way FDI can affect the economy is by technology that foreign companies bring with them, which will improve overall productivity. FDI inflows can modernize the local market and by doing so improve the economy. Nair-Reichert and Weinhold (2001) agrees that technology transfer, diffusion, and spillover effects contributes to economic growth. Furthermore, Alfaro, Chanda & Kalemli-Ozcan concludes that the link between FDI and growth is a causal one, proven by empirical evidence, but again the role of financial markets seems to be important within this link. They checked for causality by looking at endogeneity of FDI and financial market indicators. Hermes and Lensink (2010) agrees with Alfaro, Chanda & Kalemli-Ozcan that economic growth is dependent on the development of the financial system. In addition,

(12)

11 Borensztein, Gregorio & Lee (1998) uses a cross-country regression analysis as well to investigate the linkage. They use a unseemingly unrelated regression technique (SUR) with panel data. Their research suggests FDI has a positive impact on economic growth through technology and knowledge transfer, and have a larger impact than domestic investments. However, one constrain they mention is human capacity. According to Borensztein, Gregorio & Lee the effectiveness of the transfer of technology and knowledge is dependent on human capacity, seen as absorptive capacity. Here their research deviates from Alfaro, Chanda & Kalemli-Ozcan by stating that a country should have a particular threshold value of human capital to take advantage of FDI effects. Also, Borensztein et al. shows an important limitation to its research by saying that FDI is not all of what multinational corporations invest in a developing country, they can also raise debt or equity within the developing country. In their research they check if either FDI has a crowd-out effect of domestic investments. They conclude by saying that FDI is more productive than domestic investment and boosts economic growth. However, the effect of FDI is depending on the amount of human capital the host country has. Furthermore, Balasubramanyam, Salisu & Sapsford (1999) acknowledges the importance of human capital to exploit FDI. They include the size of the domestic market and the competitive climate in relation to local producers as important factors as well. Their conclusion is that FDI is more productive in countries that have pursued export promotion rather than import substitution policies. Both Alfaro et al. as Borensztein et al. refer to spillover effects contributing to economic growth. According to Crespo and Fontoura (2007) there are several ways in which spillovers can be achieved, namely: demonstration/imitation, labor mobility, exports, competition, and backward and forward linkages - being a supplier or customer - with domestic firms. However, labor mobility and competition, for example, can also be negative when multinationals attract the best employees from domestic firms and gain certain market shares from domestic firms. Due to this it is hard to conclude if the overall spillovers will be either positive or negative. Adding to the statement of Borensztein et al. that FDI spillover are dependent on absorptive capacity, Crespo and Fontoura also thinks that regional effect, domestic firm characteristics, FDI characteristics and technological cap are determinative. However, empirical evidence does not provide conclusions for these factors, only that the absorptive capacity of domestic firms seems to be an important factor to capture the FDI spillover benefits.

Nair-Reichert and Weinhold (2001) indicate in their research that not only FDI can have a positive effect on economic growth, but economic growth can attract FDI as well. They construct a model to test for causality, where they take into account the heterogeneity of the relationship between FDI and stock market development for different countries – according to Nair-Reichert and Weinhold this is lacking in previous research. They propose to use a mixed fixed and random coefficient approach to allow for heterogeneity and long-run coefficients in the causal relationship

(13)

12 between FDI and growth, first indicated by Hsiao (1989). They emphasize that there are several potential problems existing with the traditional techniques for testing causality, like for example that the coefficients found for all explanatory variables in the causal relationship are assumed to be the same for all datasets. Nair-Reichert and Weinhold suggests to look for a causality distribution, indicating that the coefficients of the explanatory variables should not have to be the same in every situation. Goal of their causality distribution is to give more information about dynamic heterogeneity in the causal relationship between FDI and economic growth. In the end, Nair-Reichert and Weinhold results suggests that the relationship is heterogeneous and also causal – which runs from FDI to economic growth.

In conclusion, Asiedu (2006) covers much of the previous research papers mentioned. She states that it is important to note that increased FDI does not necessarily imply higher economic growth. Where some studies found a positive relationship between both, others conclude that FDI enhances growth only under certain conditions and some say that the linkage is not robust at all. Here, however, the assumption about a positive linkage will be followed.

3.2 Economic growth and stock market development

Cheung and Lilian (1998) uses a Johansen Cointegration test to investigate the relationship between macroeconomic variables and stock returns. They emphasize little research is done for the long-run co-movement between both variables, which is why they use a Johansen Cointegration test. If a cointegration equation is found it indicates a long-run equilibrium relationship. As a next step they use an error-correction model to look for short-term dynamics. Eventually they find that real returns on stock indexes are generally related to deviations from the empirical long-run relationship and to changes in macro variables. In addition, Atje and Jovanovic (1993) find a significant correlation between economic growth and the value of stock market trading relative to GDP for forty countries over the period 1980-88.

However, most research so far focused on the effect of stock market development on economic growth. Caporale, Howells & Soliman (2004) argue about stock market development increasing resource allocation and liquidity, which in turn increases economic growth. They follow the VAR-procedures to test this, where they distinguish financial development into stock market development and bank development. They argue that stock markets can both affect financial development as economic growth, making it an omitted variable bias according to them. Thus, they carry out a series of trivariate causality tests to check if financial development causing economic growth is dependent upon an omitted variable, being stock market in their research. Here they indicate that previous research about the same topic using ordinary least squares regressions suffers

(14)

13 from a so called simultaneity bias. Their findings suggest that stock market development in particular affects economic growth, while also being a causal relationship. Arestis, Demetriades & Luintel (2001) also split their focus between the banking sector and stock market sector. They variables are stated as logarithms, which then are being used to perform a VAR-test. According to Arestis, Demetriades & Luintel research suggests that with one hundred data points, the maximum likelihood approach of Johansen (1988) performs in general better than a range of other estimators of long-run relationships (cointegrating vectors). Arestis et al. state that stock market development is important for the liquidity of the market, which makes trading in stocks less riskier. It also provides easy access to the capital market for companies. So overall it improves the allocation of capital, which - according to them - is an important channel for economic growth. However, increasing liquidity can also negatively affect the saving rates and corporate control (since stocks can be traded quickly), which negatively affects economic growth. By using exogeneity tests they check if a causal relationship exists. Their results suggests that both stock market as well as bank development causes economic growth, with causal directions being different for several countries. Overall, they conclude that bank-based financial systems may be more able to promote long-term growth than capital-market-bank-based ones, which does not matches Caporale et al. conclusion.

Garcia and Liu (1999) do not agree with Arestis et al. that increasing liquidity will decrease saving rates. According to them, by using economies of scale and expertise, financial intermediaries and markets are able to provide savers with a relatively higher yield, and therefore stimulate savings. Furthermore, by reducing information and transactions costs, the financial intermediaries and markets creates the link between savings and investments. According to Garcia and Liu it overall improves the allocation of resources.

Levine and Ross (1996) and Levine and Ross (1998) both find a positive linkage between stock market development and long-run growth. Levine and Ross (1998) uses a cross-country regressions analysis, where they use four indicators of growth into their regression model. For their growth indicators they use output growth, stock growth, productivity growth or savings averaged over a time period of 1976-1993. Subsequently they use four stock market variables, namely turnover, value traded, capitalization or volatility, while they control their banking development variable. So they run 16 basis regression equations to see if a linkage between stock market development and economic growth exists. Levine and Ross mentions in their study that stock markets enhances growth by increasing liquidity and decreasing risk. They do not find support that increasing liquidity or market size decreases saving rates and so hinder economic growth. Just as with Caporale, Howells & Soliman they also find that market liquidity is related to economic growth more significantly than market size. However, they do not find a causal relationship, where Caporale et al. did found one.

(15)

14 market development and economic growth for seven SSA countries for the period 1980-2004. They use an autoregressive distributed lag (ARDL) bounds test to check for cointegration. Subsequently, they perform a causality test based on a vector error correction model to look for possible directions of a causal relationship. Their results are ambiguous, where the results from some countries do show a causal relationship running from stock market development to economic growth, this does not apply for some other countries. Overall, their empirical results shows how hard it is to find a causality relationship and that research is bound to make only suggestions. Furthermore, they add to previous research that diffuse ownership may negatively affect corporate governance and invariably the performance of listed firms, thereby impeding the growth of the stock markets.

Although the direction of causality between economic growth and stock market development is ambiguous, Adam and Tweneboah (2009) will be followed in their assumption that economic growth enhances stock market development in the remainder of this research.

3.3 FDI and stock market development

As mentioned before Adam and Tweneboah take out economic growth to see if there is a direct linkage between FDI and stock market development. They use the multivariate Johansen Cointegration technique to look for cointegration equations, which indicate a long-run relationship. Subsequently, a generalized impulse response function (GIRF) from vector error correction model (VECM) is being used to investigate the linkages between FDI and stock market development in Ghana further. They find a long-run relationship between both variables, running from FDI net inflows to stock market development. To see the short-run dynamics of the found long-run relationship, Adam and Tweneboah uses an impulse response function and variance decomposition from the error correction model. The results of both techniques suggests that although a long-run relationship is being found, FDI net inflows does not have the power to predict stock market development. Initially they found two reasons to believe a linkage between both variables could exist. Firstly, in their study they refer to Errunza (1983) who found that foreign capital inflows have a long term impact on stock market development and increase investor participation. Secondly, they mention Yartey (2008) who argues that foreign investment is associated with institutional and regulatory reform, adequate disclosure and listing requirements and fair trading practices, which inspire greater confidence in domestic markets. Subsequently this increases the investor’s participation, which leads to more capital flows. Adam and Tweneboah here emphasizes that the linkage between FDI and stock market development is quite a new one within the field of research and so there is still a lot to be gained.

(16)

15 research, amongst others, into the linkage between FDI and stock market development. They argue that FDI is a complement of stock market development, both domestic as internationally, and found a positive correlation between both. Here they make a distinguishment between domestic and international activity, since according to Claessens et al. international capital markets have a higher liquidity. To make such a distinguishment they both look at domestic market capitalization and international market capitalization as the dependent variable in their regression model. To quote Claessens et al.: ‘FDI is also positively associated with measures of the internationalization of stock

markets, such as market capitalization of international firms, value traded abroad, and capital raised abroad.’ According to them this implies that foreign investment might take place through

international markets and that it helps companies to move abroad. It is important to note, however, that FDI is one of the variables being used in their regression model. They also looked at factors like GDP per capita, inflation, legislation and financial liberalization. Due to this it deviates from the research performed by Adam and Tweneboah, who solely concentrates on the effect of FDI. The remainder of this research will be focused on the effect of FDI on stock market development.

(17)

16

4. Methodology

In this section the methodology, which will be used to explore the linkage between FDI and stock market development, will be covered. This will set the foundation for the data analysis following. Following Adam and Tweneboah (2009), focus of this research is to find a long-run linkage. By looking at the long-run linkage between FDI and market capitalization as a % of GDP it also sheds light on the nature of their short-run dynamics (Cheung&Ng, 1998). According to Cheung and Ng (1998) if the variables are cointegrated it implies that they tend to move together in the long-run. The Johansen Cointegration test can be used to check if there are any cointegration equations and this technique is widely adopted by researchers like Cheung&Ng. Subsequently, an error-correction model can be used to identify short-run dynamics and the effect of long-run restrictions. According to Granger (1983): ‘cointegration is the statistical equivalence of the economic theoretic notion of

stable long-run relationship’. According to Arestis, Demetriades & Luintel (2001) with a dataset of

around one hundred data points a Johansen Cointegration test is best suited to find a long-run relationship. Following previous research a Johansen Cointegration test in incorporated within this research. The Johansen Cointegration test takes for example two independent linear variables, X and Y, and checks if there exists any linear co-movement between them. A restriction, however, with the Johansen Cointegration test is that all variables should be stationary and if they are not this could be achieved by differencing all variables by the same amount which should lead to them being stationarity afterwards (Johansen, 1991). If all variables are differenced by the same order, for example one time, they are called to be all integrated of order one and can be used to perform a Johansen Cointegation test. Differencing the variables one time refers to the fact that the current value of the variable is subtracted by the value one time period ago, the difference between both values is thereafter being checked for stationarity.

Stationarity refers to the fact that from this variable certain statistical measures, like the mean and variance, will be constant over time (Hendry & Juselius, 2001). So while neither X or Y lies around a constant value over time, some combination of them does. As mentioned above, stationarity is necessary to find a long-run equilibrium relation between both variables. It helps in predicting future values of both variables. However, goal of this research is to see if a linkage between FDI and market capitalization as a % of GDP exists and not so much forecasting their values. Since several problems with causality and the strength of the found coefficient of the explanatory variable are present, it is not wise to use the found equation as a predictor – more of these drawbacks are covered in the next paragraph. Following stationarity, the corresponding variables are assumed to have no stochastic or deterministic trend (Eviews, 2013).

(18)

17 cointegration. In this example a drunk lady and a puppy are being used as two random walks, variable X and Y. When someone asks outside a bar where the puppy is, the most likely answer will be: ‘Well, he was here ten minutes ago’. This emphasizes a key trait of random walks, namely the prediction of their location is best guessed by looking at their latest location. Also, the longer someone waits, the further they can be from the latest location – which is another key trait of random walks. Now the assumption that the dog belongs to the drunk lady is being included. What happens next is that the drunk owner calls for her dog outside of the bar and the dog answers back by barking. The two follow each other noise and by doing so they are trying to close the gap between them. Here stationarity refers to the fact that if one person finds the lady, the dog should not be far away. If this is true, than the distance between the two paths are stationary and the walks of the lady and the dog are said to be cointegrated of order zero.

The reason why the notion of cointegration is important here, is because it deals with non-stationary variables. Also, it incorporates both short-term dynamics, deviations from equilibrium, and long-run expectations, corrections to equilibrium. When using a Johansen Cointegration test all non-stationary variables are made non-stationary, by differencing the variables, before proceeding with the test. Standard regression analysis fails when dealing with non-stationary variables, leading to suggestions of certain relationships when they might not exist. Granger (1974) refers to this phenomenon as spurious regressions. Here he mentions that there can exist an autocorrelation between residual errors which can show a relationship as a result, while that does not have to be the case. Although the results can show a high R-squared, fraction of the results explained by the independent variables, Granger (1974) shows that should not necessarily be regarded as evidence for a significant relationship. It leads to misleading results for time series with trends. He suggests taking the first differences of the variables, making them stationary, as a possible solution.

An alternative for the Johansen Cointegration test is the Engle-Granger Cointegration test (Engle&Granger, 1987; Granger, 1981). The Engle-Granger test is based on a two-step approach. First step is to establish that each series is integrated of the same order. Subsequently an ordinary least squares regression is being used to find the missing factors within the equation. Second, an augmented Dickey-Fuller test is used to test if the errors are integrated of order zero (Engle&Granger, 1987; Azire, 2006). However, from the practical point of view there are quite a few critical notes on the Engle-Granger test. Most important one is that the Engle-Granger test cannot be used for linkages with more than two variables and has several problems acknowledged by amongst other Azire (2006). Azire only mentions the fact that the Engle-Granger test has some drawbacks, but Masih (1998) mentions one specific drawback, namely that not enough is known to assert that it is robust to various departures from normality. This makes statistical inference problematical and weakens results derived from the test. Also he, and Ghatak & Utkulu (2010), acknowledges that the

(19)

18 Engle-Granger test does not account for the possibility of multiple cointegration relationships and hence all the possible dynamic interactions that could exist between two or more time series. Masih emphasizes that, although multiple interactions applies to multivariate models, the Johansen test still is being used extensively in various bivariate studies in comparison to the residual based single equation Engle-Granger OLS approach. Azire agrees with Masih about the fact that the Johansen test is more reliable to test for cointegration and therefore is being used more often. Finally, Cheung and Ng (1998) argues that the Johansen Cointegration test is more efficient that the two-step Engle-Granger approach.

Because a Johansen Cointegration test requires that the variables should be integrated of the same order, Adam and Tweneboah (2009) test for unit roots by using the Augmented Dickey-Fuller (ADF) and Philips-Perron (PP) approach - where Love (2005) only uses the ADF test. Adam and Tweneboah uses the unit roots test to see if their variables, FDI net inflows, market capitalization as a % of GDP and the exchange rate with the $, are stationary. Subsequently they use the results to see if a long-run relationship between all variables exist, with market capitalization as a % of GDP as the depending variable, by using a Johansen Cointegration test and an error-correction model. Both the ADF as well as the PP test check of which order all variables are integrated. According to Jong et al. (1992) the ADF test is the most reliable to test for stationarity, or how they call it ‘most well-behaved’. Subsequently, a test for the appropriate lag length should be executed. Adam and Tweneboah uses the Akaike information criterion and the Schwartz Bayesian criterion to select the appropriate lag length, which will be used for the Johansen Cointegration test. The lag length shows how many terms back the autoregressive process checks for serial correlation. Thus, choosing the appropriate lag length is important. According to Woolridge the determination of lag length is a trade-off between the curse of dimensionality and reduced models, which are not appropriate to indicate the dynamic adjustment (Eviews, 2013). If the lag length is too short, autocorrelation of the error terms could lead to apparently significant and inefficient estimators. Therefore, one would receive wrong results. Including more lags causes the loss of more data, and will change the information criteria (Eviews, 2013). Lumsdaine and Ng (1999) adds that selecting a lag length in an autoregression that is lower than the true order will establish a misspecified model that often result in serial correlation. Also, even with inclusion of a small lag length interval many parameters still have to be estimated and increasing the amount of parameters causes the degrees of freedom to decrease (Eviews, 2013). Love (2005) suggests the appropriate lag length can be also be derived by using the Engle-Granger (1987) method by starting with fewer lags and then go ahead and test afterwards for added lags. The criteria is that if non-autocorrelated residuals is achieved, the appropriate amount of lags has been selected.

(20)

19 the trace test and the maximum eigenvalue test (Eviews, 2013). The null hypothesis for both tests is that the number of cointegrating vectors is less than or equal to r, where r is 0, 1 or 2. In each case, the null hypothesis is tested against a general alternative. Only difference between both tests is that the alternative hypothesis for the maximum eigenvalue is explicit, here the null hypothesis of for example no-cointegration is tested against the alternative of 1 cointegration vector, and subsequently against 2 cointegration vectors etc (Azire, 2006).

Following the Johansen Cointegration test a VAR-test is the next step in determining if there exists a long-run linkage. Preconditions to perform a VAR-test are that all variables are stationary, so they do not follow a certain trend, and the lag length is known (Eviews, 2013). According to Hendry and Juselius (2001) in a VAR, each variable is explained by its own lagged values, and the lagged values of all other variables in the system. For example:

(1) Yi = C + α1*Yi-1 + α2*Xi-1 +

ɛ

i

Here the variable Yt-1 depends on a certain constant, C, and has an autocorrelation of a factor α1. Furthermore it depends on an one year lagged variable of Xt, with a factor α2. In this example Y depends on both itself as on X. Granger (1987) rewrites above equation, by subtracting Yi-1 both on the left as on the right side of the equation, as:

(2) ΔYt = α2(Yt−1−βXt−1) + Ct

Here (3) Yt−1−βXt−1 is the extent of disequilibrium of the long-run relationship and (4) α2 is the speed of adjustment, showing that the cointegrated variables adjust to math their equilibrium (Tsay, 2010). Although cointegration indicates the presence or absence of Granger causality, it does not indicate the direction of causality between variables (Masih, 1998). This direction of the Granger (or temporal) causality can be detected through the vector error correction model (VECM) derived from the long-run cointegrating vectors (Masih, 1999). In the presence of cointegration vectors, a normal causality test would not suffice, leading to the reason to incorporate an error correction model (Ghatak & Utkulu, 2010). However, an assumption about the direction of causality is needed to perform a VECM within Eviews, so a Granger test of Causality should be performed first. This could lead to a serious flaw regarding causality for the remainder of the research, which will be discussed further in the next section.

A VECM can been seen as a VAR model designed for use with nonstationary time series that are known to be cointegrated (Lada&Wójcik, 2007). To visualize a VECM, take the dog and his drunk owner as an example again. The moment the lady is wandering around the bar to find her dog and starts following the barking of her dog is an error correction, since she will adjust her pattern of

(21)

20 walking around to the barking she is hearing. When finding cointegration vectors from the Johansen Cointegration test and a long-run equation from the VAR model, both must be combined to find an error-corrected equation. Johansen's (1991) approach is to estimate the VECM by maximum likelihood, under various assumptions about the trend or intercept parameters and the number r of cointegrating vectors, and then conduct likelihood ratio tests. The resulting long-run equation from the VECM will be the following:

(5) Yi = C +

α

*Xi +

ɛ

i

Here Y is the depending variable, which is stock market development. Following Adam and Tweneboah (2009) market capitalization as a % of GDP will be used as a proxy. C is a constant factor and X is FDI activity, following again Adam and Tweneboah FDI net inflows will be used as a proxy of FDI activity. Finally, α is the correlation between X and Y. In addition, the result of the VECM will distinguish between short-run dynamics and a long-run linkage. It is important to note, however, that the causality of the linkage is ambiguous.

Adam and Tweneboah (2009) uses a variation decomposing technique, followed by an impulse response function to find a long-run relationship – this technique is also adopted by Masih (2009). These techniques are based on and developed by Sims (1980) vector autoregressive procedure. However, these techniques will not be used here, instead the previous mentioned VECM will be used to see if there exists a long-run linkage between FDI and stock market development.

(22)

21

5. Results and Discussion

In the section the results from the previous mentioned models will be discussed. The software program Eviews will be used to execute all steps mentioned above, to see if there is a long-run linkage between FDI and stock market development. Two countries were chosen to check for the linkage between both variables, namely Kenya and Zimbabwe. Part of the reason for choosing these countries was because of the available data, which stretches until 1988 within the database of the World bank. As mentioned before, when performing a Johansen Cointegration test it is important to have as much data as possible. Also, both countries do differ from each other in certain ways. So does Kenya has a stock exchange since 1954 and is known as a driving economy of continental Africa since years, whereas Zimbabwe established a stock exchange in 1993 open for foreign investment and dealt with hyperinflation. Using one well known and one upcoming stock exchange will be interesting for this research and the interpretation of the found results.

5.1 Kenya

The first country to check for a long-run linkage is Kenya. The stock exchange in Kenya is called the Nairobi Securities Exchange (NSE) and was established in 1954 (Acca, 2014). The NSE is under control by the Capital Markets Authority (CMA) of Kenya and is a member of the East African Securities Exchanges Association. Following Adam and Tweneboah (2009) the variable X is the amount of FDI net inflows and variable Y is market capitalization as a % of GDP, which is also being used as a proxy for stock market development by Caporale, Howells & Soliman (2004).

Table 1 shows the amount of FDI net inflows and market capitalization as a % of GDP for Kenya on an annual basis for 1988-2012 (Worldbank, 2015). Here, the same amount of years is being used as with the research of Enisan and Olufisayo (2007). The next step is to convert the annual data to quarterly data, to increase the amount of datasets necessary to perform a Johansen Cointegration test, which is done by adjusting the frequency for the data within Eviews. Adam and Tweneboah (2009) follow the suggestion made by Khan (1976), however, converting by Eviews is being widely used by other researchers and subsequently here as well. Hereby, Cheung and Ng (1998), Caporale, Howells & Soliman (2004) and Arestis, Demetriades & Luintel (2001)are being followed as well. Figure 1 shows the unit roots test for Kenya, which tests for stationarity – both Adam and Tweneboah (2009) as Masih (1998) are being followed for this procedure. The results of figure 1 show that both variables are stationary after differencing them once, at a significance level of 0.05 – in other words; they are integrated of order one. According to Jong (1992) the Augemented-Dickey Fuller (ADF) test is most reliable, or as they call it most well-behaved, when testing for stationarity, which was also used in this research. Now that the requirement for stationarity if fulfilled and all

(23)

22 variables are integrated of the same order, the next step is to identify the appropriate lag length. Previous paragraph already discussed why determining the appropriate lag length is important. Figure 2 shows that according to the Schwarz information criterion (SC) the appropriate lag length is two, meaning that the autoregressive process test for serial correlation two terms back. However, the Akaike information criteria (AIC) indicates a lag length of 6 is appropriate. Adam and Tweneboah (2009) also uses a ADF test and focuses on both AIC as well as SC. They, however, uses the AIC as leading indicator for the maximum lag length, whereas here the SC is being used as a leading indicator. Subsequently figure 3 shows the VAR-test, when assuming a lag length of two. When testing for a lag length of six, the AIC and SC are slightly bigger – where it is better to minimize both (Eviews, 2013). But most importantly, when using a lag length of six no cointegration was found as a result of a Johansen Cointegration test – indicating no long-run equilibrium relationship. Also, looking at the quarterly data being used it seems unlikely that today’s data is dependent on its own lagged variables as far back as six periods – adding up to 1.5 years back. For these reasons a lag length of two is being used for the remainder of the tests.

Looking at the VAR outcome in figure 3, it shows that FDI net inflows is dependent of its own lagged variables but also the lagged variables of stock market development Also, stock market development is dependent of the lagged variables of itself and the lagged variables of FDI net inflows. It shows that both FDI and stock market development are depending on each other and the causal direction seems unknown here. Following the outcome of the test, the following equations can be derived:

(1) Xt = -3037201 + 1.517619*Xt-1 - 0.711201*Xt-2 - 430964.4*Yt-1 + 1586803*Yt-2

(2) Yt = 0.819921 + 1.702816*Yt-1 - 0.741614*Yt-2 + 7.08E-10*Xt-1 -9.93E-11*Xt-2

Within this equation X represents FDI net inflows and Y stock market development. Because a lag interval running from one to two is assumed, both equations show dependency on the one and two year lagged variables. The first equation shows that the amount of net FDI inflows on time t (Xt)is dependent on a negative constant factor, namely -3037201, and is positively affected by both the amount of net FDI inflows from previous year (Xt-1) and the amount of market capitalization as a % of GDP of two years ago (Yt-2). Furthermore, it is negatively affected by both the amount of FDI net inflows from two years ago (Xt-2) and the amount of market capitalization as a % of GDP of one year ago (Yt-1). The second equation shows that the amount of market capitalization as a % of GDP on time t (Yt) is dependent on a positive constant factor, namely 0.819921, and is positively affected by both the amount of market capitalization as a % of GDP and the amount of net FDI inflows from previous year. Furthermore, it is negatively affected by both the amount of market capitalization as a % of GDP and the amount of net FDI inflows from two years ago. Since market capitalization as a % of

(24)

23 GDP on time t is an relative number, the effect of FDI net inflows is small, since otherwise the amount of FDI net inflows will have a large effect on market capitalization as a % of GDP – leading to misleading results.

Following the VAR-test, figure 4 shows if the VAR-test is stationary. Since all dots are within the inner circle of one, the conclusion can be made that the VAR-model is stationarity after differencing all variables once – in other words; all variables integrated of order one (Granger, 1987). Note that, however, all dots are close to the border of the circle, meaning that the conclusion of all variables being stationary of order one is quite fragile. The dots represents all lagged variables being used in the VAR-model (two lagged variables for X and two for Y). Figure 5 shows the results of the Granger Causality test. This test is to see in which direction the causal linkage runs. Granger (1969) describes the concept of causality as following: ‘X is said to be Granger-Cause Y, if Y can be predicted

with greater accuracy by past values of X rather than not using such past values and all other relevation information remaining the same (Love, 2005)’.The results shows that the null hypothesis

of Y not affecting X is rejected, meaning the direction goes from variable Y to variable X. So following the Granger Causality test market capitalization as a % of GDP has an effect on FDI net inflows, it helps in attracting more FDI activity. These results are a first indication, necessary to perform a VECM. Several authors has criticized the Granger Causality test, which will be covered during the discussion. The fact that due to an omitted variable bias, economic growth in this case, a causal relationship is being found cannot be ruled out as well. Since no alternatives are known to test for causality the results will be used to perform a VECM, although it is not ideal.

Next in figure 6, the results of the Johansen Cointegraton test is shown. Both Adam and Tweneboah (2009) and Cheung and Ng (1998) are being followed in this procedure. The results indicate that there is one cointegration equation, meaning that there exists a long-run equilibrium relationship between both variables (Masih, 1998). The Johansen Cointegration test incorporates both the trace test as the maximum eigenvalue test, the difference between both has been explained in previous section. Hence, both give the same result of one cointegration vector. The final step is to incorporate the error terms and derive a final long-run equation between Y and X. This is done by using a Vector Error Correction Model (VECM), which figure 7 shows the results for. Again both Adam and Tweneboah (2009) and Cheung and Ng (1998) are being followed when executing a VECM. The results show the long-run equation as well as the short-term adjustments when the variables deviate from the equilibrium, which is part of the stationarity characteristics. Following figure 7, the long-run equation is as follows:

(25)

24 Within the found equation - 0.206763 refers to the error-correction amount (Lada&Wójcik, 2007). Taking the small amount of - 0.206763 as zero, compared to the large amounts in the equation the effect of this small number is neglectable, this can be rewritten as follows (Eviews, 2013):

(4) Xt-1= – 29276362 + 6565037*Yt-1 +

ɛ

t-1

For interpretation purposes equation (4) can be rewritten to:

(5) ΔXt = – 29276362 + 6565037*ΔYt +

ɛ

t

Equation (5) shows that FDI net inflows is depending on market capitalization. This is a result of an assumption behind the VECM procedure which has to do with the fact that the independent and dependent variable has to be clarified before executing the model. Following the results of the previous Granger Causality test it is assumed that X is the independent variable and Y the depending, so the causal relationship runs from X to Y. This assumption can be seen directly within equation (5). Furthermore, it shows that with every increase of market capitalization of 1% FDI net inflows increases by $6.565.037,-. The constant factor here has a negative impact on the amount of FDI and makes sure an adjustment downwards will be made of $29.276.362 at every point of time. Finally,

ɛ

t refers to the error-correction part of the equation, which incorporates random shocks to the model. Ghatak and Utkulu (2010) refer in their research to Jones and Joulfaian (1991) who suggest the interpretation that the changes of the lagged independent variables describe the short-run causal impact, while the error-correction term captures the long-run effect - this is also acknowledged by Love (2005).

5.2 Zimbabwe

The second country which will be used to investigate the linkage between FDI and stock market development is Zimbabwe. The stock exchange of Zimbabwe is called the Zimbabwe Stock exchange (ZSE) and was established in 1993 (Acca, 2014). The ZSE is under control by the Securities Commission (SeC) of Zimbabwe and is a member of the Southern African Development Committee Stock Exchanges. Important to note is that in 1989 a new investment code was adopted to increase the proportion of after-tax profits that multinational companies could repatriate from 50 to 100%, adopted to increase FDI activity (Gwenhamo, 2011). Again the variable X will be used to denote the amount of FDI net inflows and the variable Y the market capitalization as a % of GDP.

Table 2 shows the amount of FDI net inflows and market capitalization as a % of GDP for Zimbabwe on an annual basis for 1988-2012. Again the data was converted to quarterly by using the

(26)

25 techniques used by Eviews. Subsequently figure 1 shows the unit roots test for Zimbabwe, where again an Augmented Dickey-Fuller test was being used. Both variables are integrated of order one at a significance level of 0.05, where no intercept or trend were assumed while performing the test. Figure 2 shows the lag length test, which is necessary for performing a Johansen Cointegration test. Similar to the data analysis for Kenya, the Schwarz information criterion is being followed, resulting in a maximum lag length of two.

Figure 3 shows the outcome from the VAR-test. The results show that FDI net inflows is dependent on its own lagged variables , but also on the lagged variables of stock market development – the same applies to stock market development. Subsequently, the following equations can be derived from the VAR-test:

(1) Xt = 4609932 + 1.682989*Xt-1 - 0.733812*Xt-2 + 5481.213*Yt-1 + 6157.722*Yt-2

(2) Yt = 8.458916 + 3.27E-08*Xt-1 -2.61E-08*Xt-2 + 1.656117*Yt-1 - 0.786907*Yt-2

Within this equation X represents FDI net inflows and Y stock market development. Because a lag length interval running from one to two is assumed, just as the case is with Kenya, the variables Xt and Yt show dependency on the one and two year lagged variables. Looking at both equations stipulated above; equation (1) shows the dependency of foreign direct investment at time t. It shows that Xt has a constant factor of 4609932 and is positively affected by its own one year lagged variable (Xt-1) as well as both lagged variables of Y (Yt-1 and Yt-2). Also, the variable Xt is negatively affected by its own two year lagged variable (Xt-2) – although this effect is less stronger than the positively effect of its own one year lagged variable. Furthermore, equation (2) shows the dependency of market capitalization as a % of GDP at time t (Yt). The resulting equation shows that Yt has a positive constant factor of 8.458916 and is positively affected by its own one year lagged variable (Yt-1) and both lagged variables of X (Xt-1 and Xt-2). It also shows that Yt is negatively affected by its own two year lagged variable (Yt-2) – although this effect is also less stronger than the positively effect of its own one year lagged variable.

Comparing the results to the previous found results of Kenya the higher constant factors draws the attention, especially since the constant factor for Xt for Kenya is negative. Looking at the tables 1 and 2, where all the annual data from both Zimbabwe and Kenya is stipulated, shows that FDI net inflows and market capitalization as a % of GDP typically are both higher for Zimbabwe – resulting in higher factors within the equation. There is also a difference in affecting factors, where for example with Kenya Xt is negatively affected by its own two year lagged variable and the one year lagged variable from Yt. With Zimbabwe Xt is only negatively affected by, previously mentioned, its own two year lagged variable. Any other conclusions, besides the observation, cannot be derived from both equations.

The effect of Foreign Direct Investment on stock market development within Sub-Saharan Africa