International Journal of Commerce and Finance, Vol. 5, Issue 2, 2019, 167-186 167 RELATIONSHIP BETWEEN STOCK MARKETS IN AFRICA: A CASE OF FIVE SELECTED COUNTRIES Abdifitah Mohamed Jama DEER Istanbul Commerce University,Turkey Ayben KOY Istanbul Commerce University, Turkey Abstract This article aims to analyze the relationship between the stock markets in Africa (Egypt, Kenya, Morocco, Nigeria and South Africa). The sample used in the study is beginning from 2009 to 2018 in a weekly data range. The main findings in the study are: (1) price indices of Casablanca stock exchange are not influenced by other stock markets in the long run (2) Egyptian stock market can be used to predict the Kenyan stock market but not Morocco, South Africa, or Nigeria, (3) South African stock market can be used to pred ict the Egyptian, Nigerian, and Kenya stock markets, and (4) Johannesburg stock exchange plays a vital role in effecting the stock prices of other African countries. Keywords: Stock Market, Causality 1. Introduction One of the results of globalization is a free flow of capital speculations, dominatingly from developed markets to developing nations such as Asia, Africa, Eastern Europe and South America (Adjei, 2015, p.7). A large number of such emerging and frontier economies have come a long way with democratic governance, accountability and upgrades in their financial administration, which has led to strengthened institutions and administrative frameworks. As of 2012, the emerging economies had contributed 38% of worldwide GDP which could ascend to 63% by 2050 (Adjei, 2015, p.8). Statistics show that the African continent has yet to play a unique role in global economic growth, though the potential exists. Nonetheless, some of the world top growing economies are in Sub-Saharan Africa (IMF, 2015). Between 1995 and 2013, the economy of Sub-Saharan Africa grew at 4.5% per annum on average (World Bank, 2015). Commonly, many African face issues that limit speculation, such as obsolete business laws and guidelines, poor infrastructure, challenges in getting to money related capital, resolute and complex duty approach, and frequent conflicts that have resulted to regional instability. The formal financial markets in Africa have developed from the last two decades (Ntim et al., 2011). The considerable growth in formal financial markets was brought by the establishment of several capital and money markets in Africa. From the mid-1990s, African nations sold the vast majority of their stakes of public organizations to provoke the development of the private sector (Ly, 2011, p.6). There has been an increased number of African nations setting up stock markets from 18 in 2002 to 29 active stock markets in 2018 (Coetzer, 2018). This mirrors a picture of readiness to grow and eventually overcome any gap between Africa and other nations across the globe. The aim behind the creation of stock trades in Africa was to mobilize national resource and attract foreign investors (Mahama, 2013, p.6). In general, African stock markets can't flaunt an exceptionally outstanding share of the world market capitalization yet (Coetzer, 2018). There has however been at least one African stock trade on the world top ten best performing stock markets. This is viewed as an impression of the endeavors on the headway of stock trades in the African district. It is an inspiration for both neighborhood and remote financial specialists to think about putting resources into African stocks. The rise of African stock markets has raised worries about the mix of African stocks as money related center point with the influence of pulling in universal financial specialists to Africa (Mahama, 2013, p.7). In te rn a ti o n a l Jo u rn a l o f C o m m e rc e a n d F in a n c e In te rn a ti o n a l Jo u rn a l o f C o m m e rc e a n d F in a n c e In te rn a ti o n a l Jo u rn a l o f C o m m e rc e a n d F in a n c e Abdifitah Mohamed Jama DEER & Ayben KOY 168 In this study, the stock market for five countries in the last ten (10) years are analyzed in a weekly data range. The five countries (Egypt, Kenya, Morocco, Nigeria and South Africa) position themselves as the giants in the African region. They also command the highest stock market in the entire African continent. Stock markets in Africa represent a quickly developing portion of the world economy. They offer conceivably exceptional returns. In the last decade, some of the best performers in the stock market have hailed from this region. Experimentally, it has been shown that albeit individual developing stock markets have been volatile, their risk- balanced return in groups has been higher than their counterparts in developed markets. 2. Literature Review 2.1 History of Development of African Stock Markets African Stock Markets have indicated an aggregate indication of agile development and improvement after some time. In 1987, the stock market had only eight stock exchanges, and by 2018 it had a total of 29 stock markets from a total of 38 countries (Coetzer, 2018). The oldest stock exchange is the Egyptian stock exchange which was established in 1883 (ASEA, 2012). It is thus correct to say that African stock exchanges have been in the market for quite some time. African stock markets differ significantly in institutional and market infrastructural attributes. Ntim (2012) offers a five-tier characterization of African stock markets. The first level consists exclusively of South Africa – the most well built up, the biggest, and one of the early established financial exchanges in Africa. The second level comprises of several medium-size markets among them Egypt, Tunisia, Nigeria, Morocco, Kenya, and Zimbabwe (Ntim, 2012). The third level is comprised of new and small, however quickly developing markets, consisting of Ghana, Cote d'Ivoire, Namibia, Mauritius and Botswana (Ntim, 2012). The fourth level comprises of exceptionally new and small markets, including Libya, Sudan, Uganda, Tanzania, Mozambique, Malawi, Zambia and Swaziland. The fifth tier includes seven markets, to be specific; Algeria, Cameroon, Gabon, Cape Verde, Rwanda, Angola, and Sierra Leone, which either regardless of having been in presence for moderately longer time are not widely known or are not formally known in light of the fact that they are essentially excessively youthful (Ntim, 2012). Among the five stock markets under this study, the oldest market in Egyptian stock exchange founded in 1883 followed by Johannesburg stock exchanged founded in 1887. Casablanca stock exchange was established in 1929 while the Nairobi stock exchange was started in 1954. The youngest among the five is Nigerian stock exchange which was launched in 1998. All the stock markets are located in their respective countries capital cities (Coetzer, 2018). 2.1.1. Egyptian Stock Exchange (Egypt) The Egyptian Stock Exchange (EGX) is the oldest stock exchange in Africa. Though in Africa, the market is in most cases counted as a Middle East stock market. The number of firms trading with the EGX has increased tremendously over the years (African Securities Exchanges Association (ASEA), 2018). The EGX trades in stocks, funds and known related items issued by specific global monetary organizations. The EGX has task instrument where there are potential outcomes of intraday and online exchanging availabilities, which shows how well the stock trade has developed throughout the years. The EGX has kept its performance consistently throughout the years. In 2014 the EGX was positioned second by the MSCI lists in 2013 and number one among developing business sector peers in two years spreading over 2012-2014 (ASEA Yearbook, 2014). In the year ending 2012, it was voted the most lucrative securities exchange by garnering 49.56%. This may be credited to the European predicament, preparing for expansion into developing markets. The main index under the EGX is the EGX 30 which includes 30 listed companies. 2.1.2. Casablanca Stock Exchange (Morocco) The Casablanca Stock Exchange (CSEX) was initially known as was known as the "Office de Compensation des Valeurs Mobilières." It is Africa's third biggest Bourse after Johannesburg Stock Exchange and Nigerian Stock Exchange (ASEA, 2018). In January 1997, Morocco enacted a law making further enhancements to securities exchange organization. In 2000 the name of the stock market was further changed from Société de la Bourse des Valeurs de Casablanca (SBVC) to Casablanca Stock Exchange. In January 2007, the CSE upgraded its visual character with a desire to help its adjustment in size. Relationship Between Stock Markets in Africa: A Case of Five Selected Countries 169 Formation of a Casablanca Stock Exchange follow-up council made by the Board chiefs for the redoing of the rules of the organization was done in December 2008. The CSE officially received corporate governance with Board of Directors and General Management in April 2009. In 2015 CSE recorded a value turnover of US$ 20.4 billion with 130 exchanges. The Stock trade which recorded a total market capitalization of US$ 45,763,178,295 in 2015, and had a sum of 75 listed firms. The CSX has two indices; Morocco All share and FTSE CSE Morocco 15 Index. 2.1.3. Johannesburg Stock Exchange (South Africa) Johannesburg Stock Exchange (JSE) is one of the most established markets in Africa. JSE is currently positioned nineteenth biggest stock trade globally and the greatest in Africa as far as market capitalization (ASEA, 2018). The JSE is a member of World Federation of trades after joining 1963. The JSE has more than 395 listed firms (Johannesburg Stock Exchange, 2018). JSE has an automatic trade service. In 2010, total trade raised to $438.08 billion from $374.00 billion in 2009. In 2015, its sum market capitalization was 756.8 billion USD. It recorded a turnover of 392.6 billion USD with a total of 61,894,253 exchanges. The JSE index is JSE/FTSE all share index with approximately 99% of the market capitalization. This study will utilize the FTSE/JSE Africa All-Share as a stock market representing South Africa. 2.1.4. Nairobi Securities Exchange (Kenya) Nairobi Securities Exchange (NSE) was set up in 1954 (ASEA, 2018). NSE has throughout the years experienced numerous changes to turn into the most exceptional stock trade in the East Africa region and a standout amongst the most lucrative markets worldwide (Ventures Africa, 2013). The NSE has developed into a full securities service exchange dealing with derivatives, debt, clearing, and repayment of equities and other related instruments (Nairobi Securities Exchange, 2018). It is one of only a few stock exchanges with live exchanging through computerized trading services. In 2015, with 63 listed firms, NSE recorded a market capitalization of US$20 billion. Its turnover was US$2,045, and it had 406,634 exchanges amid a similar period. It is a price weight index, and the individuals are chosen dependent on weighted market performance for a year: 40% for Market Capitalization, 30% for Shares Traded, 20% for Number of deals 20%, and 10% for Turnover (ASEA, 2018). The NSE has one major index which is the NSE-20 started in 1994. The index return of the NSE-20 is fundamentally founded on capital increase/decrease of the 20 most prominent securities recorded The study will use the NSE 20-Share Index (NSE-20) to determine the stock market of Kenya as its members make up 80% of the stock trade in the country. 2.1.5. Nigerian Stock Exchange (Nigeria) Nigerian Stock Exchange (NGSE) was built up in 1960. It was officially started in 1961 with a total of 19 listed firms. The total number of listed firms has risen to 184 companies by 2018 (Nigeria Stock Exchange, 2018). The managing body is the Securities and Exchange Commission (SEC) while its certification comes from Investments and Securities Act (ISA). The listed firms hail from 11 key sectors among them financial services, construction, agriculture, and consumer goods. In 2015, it recorded a total market capitalization of US$49,456,969,735, and it had 184 registered organizations. Around the same time, it registered a turnover of US$3,931,503,298 and had 917,946 exchanges. The NGSE has two main indices; the NGSE 30 Index and the NGSE All-Share Index (ASI) (Nigeria Stock Exchange, 2016). In the year 2012, the NSE all offer Index shut the year with its most astounding execution since 2008 with a 35.45% addition. All out exchange an incentive in the year-end 2013 was terrific and has been same in earlier years and volume of trade developed from 89 billion of every 2013 to 105 billion out of 2013, connoting the potential displayed by the market. Additionally, market capitalization for the years 2010, 2011, 2012 and 2013 has been $53.40, $43.06, $57.77 and $80.69 billion, of course clarifying the development in the market and its chances accessible to financial specialists (ASEA, Yearbook, 2014). Abdifitah Mohamed Jama DEER & Ayben KOY 170 2.2. Empirical Review In the US and Norway, Næs, Skjeltorp and Ødegaard (2011) set out to demonstrate that financial exchange liquidity is a fitting driving marker of the real economy. Concentrating on the year 1947 – 2008, Næs et al. (2011) picked various liquidity estimates dependent on their requirement for a sensibly prolonged time series. To quantify the condition of the real economy, the indicators utilized are unemployment, real consumption, real GDP and real investment. The study uses the VAR methodology. Næs et al. (2011) finds a significant relationship between liquidity and the real economy (30). Moreover, it is discovered that there is a connection between time difference in market liquidity and the adjustments in interest in the stock market. A Chile study by Brandao-Marques (2016) examines the liquidity of the stock market in the Chilean stock exchange. Time series is used to test the cyclical behavior and liquidity (6). A total of 23 emerging markets are included for the year ranging from 2003 to 2014 using panel regression. (Brandao-Marques, 2016: 8). It is discovered that market liquidity improved with a better assurance of minority investors. Brandao-Marques (2016) noted the likelihood that the positive connection among liquidity and financial specialist assurance is a minor impression of the cross-country disparity of the significance of institutional speculators, similar to insurance agencies, annuity reserves and shared assets (12). In Taiwan, Hoang, (2017) examines the connection between liquidity and stock returns. The study covers the period from 2007 to 2014. The study utilize the CAPM Three Moment model and the Fama – French model. Results of the investigation establish a positive connection between illiquidity proportion and stock returns. In Poland, Lischewski and Voronkova (2012, p.13) explore the impact of value, size and liquidity on stock return in the most exceptional stock trades in the country. Results in this market are steady to those in developed markets as far as market, estimate, book – to – market yet could barely completely clarify whole equity premium notwithstanding when incorporating liquidity factor in the model. Liquidity factor, however, assumes a job in diminishing event of the measurably huge hazard balanced abundance return, has no proof as an estimated factor in this market (Lischewski & Voronkova, 2012, p.21). A study conducted in Latin America examine the relationship among four (4) Latin American nations to be specific Argentina, Brazil, Chile and Mexico while utilizing the U.S market as linkage (Diamandis & Drakos, 2010, p.384). The dynamic relationship between stock and foreign trade markets are analyzed by utilizing a cointegration approach. Moreover the markets show their connections with the United States market (Diamandis & Drakos, 2010, p.386). In China, Jayasuriya (2011, p. 420) analyze the inter-linkages between the Chinese stock and the emerging markets in its neighborhoods. Vector autoregressive (VAR) model is employed in this study. Monthly data covering the year 1993 to 2008 was used for the countries including the Philippines, Thailand, Indonesia, and China. Results indicated that China stock market had a controlling role in the return behavior of the other markets throughout the years. This is evidence that stuns originating from China are incredibly felt in different markets (Jayasuriya, 2011, p.422). The markets were found not to be interrelated, but instead, there is a dimension of connection among China and different markets when foreign investor return is represented. Focusing on Asian markets, Ali, Butt, and Rehman (2011, p.398) explored co-developments among developing and developed stock markets. The stock markets in the study included Pakistan, China, Taiwan, Malaysia, Indonesia, Singapore, Japan, UK, and the USA. Monthly data for the period ranging from July 1998 to June 2008 was used. Ali et al. (2011, p.401) employed the Gregory and Hansen cointegration test, the Johansen cointegration test, Engle- Granger cointegration test, and Granger causality test. The outcomes from the investigation demonstrated that there is no cointegration connection between the Pakistani securities exchange and the Malaysia, Taiwan, Singapore, UK and USA markets. Pakistani stock market was observed to be cointegrated with China, Indonesia, India, and Japan stock markets (p.402) Using African stocks markets, Alagidede, Panagiotidis, and Zhang (2010) analyzed the level of integration between African markets, emerging markets and developed markets. African countries were represented by Kenya, Egypt, South Africa and Nigeria. The emerging markets selected were India, Brazil and Mexico while the UK, Japan and USA represented the developed markets. Bivariate cointegration tests were run (Alagidede et al., 2010: 4). The outcomes demonstrated the presence of a long-run connection among Egypt and Japan, Kenya and Japan, and South Africa and Brazil. Conversely, African markets had no cointegration amongst each other. Breitung test found a cointegration between only Egypt and Brazil, and Egypt and South Africa (Alagidede et al., 2010: 9). Relationship Between Stock Markets in Africa: A Case of Five Selected Countries 171 The frontier economies (Tunisia, Ghana, Kenya, Ivory Coast, Mauritius and Botswana) and IFC Global category (Egypt, Morocco, Nigeria and South Africa) are analyzed by Agyei-Ampomah (2011) beginning from 1998 to 2007 in a monthly data range. Results have got low dimensions of the connection between's African stock markets. Besides, no proof of integration is found between them and the worldwide stock markets except for South Africa (Agyei- Ampomah, 2011: 12). Ncube and Mingiri (2015) compare Johannesburg stock exchange to other African countries. Using a monthly data beginning from 2000 to 2008, Johansen and Julius cointegration methods are used. Ncube and Mingiri (2015) notes a division of the different value markets. The consequences of whether world market exercises influence the African equity markets additionally proved positive. In South Africa, Vacu (2013) analyze the long-run relationship between the stock market and the development of the economy. Quarterly data is used covering the years 1990 to 2010. All share index, turnover ratio, market capitalization and GDP are analyzed by Johansen cointegration test, Vector Error Correction Model (VECM) and Granger causality tests. Results indicate a long run relationship existing between the factors. Focusing on three African countries, Osamwonyi and Kasimu (2013) examine the causal and direction of the relationship for the countries Ghana, Kenya, and Nigeria. In the period beginning from 1989 to 2009, Granger test for causality is employed. Outcomes indicate that the stock market for Nigeria and Ghana have got no causal connection to economic development. In Kenya, a unidirectional and bidirectional causal relationship is found. The unidirectional causality that moved from the Stock market to GDP is found in market capitalization and number of listed securities. The bidirectional causality is found in the ratio of stock turnover and GDP (Osamwonyi & Kasimu, 2013, p.88). Mohanasundaram and Parthasarathy (2015) examined the presence of securities exchange relationship and cointegration between India, South Africa, and the USA. The study employed the Johansen-Juselius multivariate cointegration approach. A month to month estimations of the market major indices (for example the Indian National Stock Exchange CNX NIFTY 50, the JSE Africa All Share file and the NASDAQ Composite) are utilized. The study covers the period from April 2004 to March 2014. Correlation test revealed high degrees of connection between the markets (481). Granger causality test is also run, and results show unidirectional causality relationship between the NIFTY 50 and JSE All Share index (483). 3. Empirical Results and Analysis 3.1. Descriptive Statistics Business performance is a multi-dimensional concept (Hofer, 1983; Lenz, 1981) as it depends on a large number of different decisions, actions and measures. That’s why it is used in several research areas. For this reason, parallel to the study discussed in the literature, the following sub-dimensions will be analyzed briefly in this section. 3.1.1 Financial Performance The mean, median, maximum and minimum data of the respective stock market prices of the selected countries are as shown in Table 2. As shown in the table, Nigeria (318.3987) has got the least standard deviation followed by Kenya (690.6346). The standard deviation for Morocco, Egypt and South Africa are 1318.455, 3581.288 and 8920.007 respectively. Egypt, Kenya Morocco and Nigeria have got a positive skewness indicating that their data are skewed toward the right. South Africa shows a negative skewness indicating that the data tends towards the left. The kurtosis of Egypt is close to the recommended value of 3 (at 2.732502). The value of kurtosis for other countries is less than the recommended value, 1.947046 for Kenya, 1.755484 for morocco, 1.772826 for Nigeria and 1.676452 for South Africa. The probability value of less than 5% (0.00 for all the five countries) confirms that the data is not normally distributed. The results imply that the data under study is not normally distributed. Abdifitah Mohamed Jama DEER & Ayben KOY 172 Table 1 Descriptive Data 3.2. Stationarity Tests Stationarity tests are used to check the presence or absence of unit root. If a model has unit root it is said to be non- stationary while absence of unit root indicates a stationary model. In this study, both Augmented Dickey-Fuller (ADF) test and Phillips-Perron (PP) test were applied. 3.2.1 Unit Root Test at Level It is recommended that models should be non-stationary at level. The following hypothesis are applied; Null hypothesis: variable is not stationary at level Alternative hypothesis: variable is stationary The null hypothesis is rejected if p value is less than 5% hence the model is said to be non stationary at level. Table 2 Unit Root Test at Level Augmented dickey-fuller test Phillips-Perron Test Variable With intercept With intercept and trend With intercept and trend With intercept With intercept and trend With intercept and trend CSE -1.781499 -1.780278 -0.005967 -1.796627 -1.786004 0.149022 ESX -0.834036 -1.796313 0.788502 -0.769235 -1.819126 0.790037 JSE -1.801935 -2.568563 1.491941 -1.546017 -2.136619 1.435523 NGSE -2.867576 -2.001596 -0.119393 -2.037062 -1.817870 0.082386 NSE -2.160294 -2.219030 -0.110967 -1.105160 -1.082913 -0.533515 The probability value for all the variables is more than 5% thus we cannot reject the null hypothesis. Therefore, the conclusion is that the variables are not stationary at level. 3.2.2 Unit Root Test after First Difference It is expected that after first difference, the model should become stationary. The following hypotheses were therefore tested; Null hypothesis: variable is not stationary Alternative hypothesis: variable is stationary Relationship Between Stock Markets in Africa: A Case of Five Selected Countries 173 Table 3 Unit Root Test after First Difference Augmented dickey-fuller test Phillips-Perron Test Variable With intercept With intercept and trend With intercept and trend With intercept With intercept and trend With intercept and trend CSE -11.51419* -11.50245* -11.52352* -21.69550* -21.67780* -21.71274* ESX -20.63693* -20.62013* -20.59118* -20.61512* -20.59824* -20.58874* JSE -12.44174* -12.54008* -12.23853* -24.76864* -25.33432* -23.99369* NGSE -6.051426* -6.080612* -6.048666* -20.00283* -20.01353* -20.01330* NSE -7.273237* -18.90213* -7.284279* -18.85886* -18.90213* -18.87249* The * sign denotes a probability value of less than 0.05 or 5%. This therefore means the we reject the null hypothesis and conclude that the variables are stationary at first difference. 3.3. VAR lag order VAR lag order for every two coutries stock indices are determined. To get the optimal lag length for each relationship, the order that has got the highest number of recommendation is taken. According to the results of Final Prediction Error (FPE), Akaike Information Criteria (AIC), Schwarz Criterion (SC), and Hannan Quinn (HQ) the optimal number of lags are as following: CSE & ESX: 3; CSE & JSE: 1; CSE & NGSE: 1; CSE & NSE: 3; ESX & JSE: 2; ESX & NGSE: 2; ESX & NSE:2, JSE & NGSE: 3; JSE & NSE:3 and NGSE & NSE: 3. 3.4. VAR Analysis Following determining the lags for all index duals, VAR test is used. The test is applied to the first difference since the variables are stationary at first difference. Tables 7-16 are the VAR Analysis results for the times series duals. Table 4 VAR Analysis between CSE and JSE Coefficient Std. Error t-Statistic Prob. C(1) 0.058066 0.043687 1.329134 0.1844 C(2) 0.008736 0.008722 1.001585 0.3170 C(3) 1.668440 7.104298 0.234849 0.8144 R-squared 0.005331 Mean dependent var 2.290212 Adjusted R-squared 0.001483 S.D. dependent var 161.7794 S.E. of regression 161.6594 Akaike info criterion 13.01461 Sum squared resid 13511153 Schwarz criterion 13.03915 Log likelihood -3380.799 Hannan-Quinn criter. 13.02423 F-statistic 1.385505 Durbin-Watson stat 2.013289 Prob(F-statistic) 0.251125 Coefficient Std. Error t-Statistic Prob. C(4) -0.235098 0.219581 -1.070667 0.2848 C(5) -0.023638 0.043838 -0.539210 0.5900 C(6) 54.19977 35.70780 1.517869 0.1297 R-squared 0.002773 Mean dependent var 52.27754 Adjusted R-squared -0.001085 S.D. dependent var 812.0962 S.E. of regression 812.5365 Akaike info criterion 16.24395 Sum squared resid 3.41E+08 Schwarz criterion 16.26849 Log likelihood -4220.427 Hannan-Quinn criter. 16.25357 F-statistic 0.718806 Durbin-Watson stat 2.009816 Prob(F-statistic) 0.487820 Abdifitah Mohamed Jama DEER & Ayben KOY 174 Table 5 VAR Analysis between CSE and ESX Coefficient Std. Error t-Statistic Prob. C(1) 0.025052 0.044773 0.559531 0.5760 C(2) 0.072052 0.044698 1.611957 0.1076 C(3) 0.065623 0.044939 1.460262 0.1448 C(4) 0.025265 0.024301 1.039704 0.2990 C(5) 0.010785 0.024216 0.445340 0.6563 C(6) 0.017542 0.024136 0.726781 0.4677 C(7) -0.360219 7.073295 -0.050927 0.9594 R-squared 0.018018 Mean dependent var 0.588630 Adjusted R-squared 0.006328 S.D. dependent var 159.6266 S.E. of regression 159.1207 Akaike info criterion 12.99081 Sum squared resid 12760982 Schwarz criterion 13.04884 Log likelihood -3312.151 Hannan-Quinn criter. 13.01356 F-statistic 1.541269 Durbin-Watson stat 1.992900 Prob(F-statistic) 0.162635 Coefficient Std. Error t-Statistic Prob. C(8) 0.101411 0.082699 1.226265 0.2207 C(9) 0.192504 0.082561 2.331640 0.0201 C(10) 0.075770 0.083007 0.912815 0.3618 C(11) 0.074382 0.044885 1.657166 0.0981 C(12) -0.050733 0.044730 -1.134208 0.2572 C(13) 0.012937 0.044581 0.290183 0.7718 C(14) 17.60973 13.06497 1.347858 0.1783 R-squared 0.025276 Mean dependent var 18.37526 Adjusted R-squared 0.013672 S.D. dependent var 295.9395 S.E. of regression 293.9094 Akaike info criterion 14.21802 Sum squared resid 43536913 Schwarz criterion 14.27606 Log likelihood -3625.705 Hannan-Quinn criter. 14.24078 F-statistic 2.178238 Durbin-Watson stat 1.990813 Prob(F-statistic) 0.043774 Table 6 VAR Analysis between CSE and NGSE Coefficient Std. Error t-Statistic Prob. C(1) 0.066001 0.046192 1.428846 0.1537 C(2) 0.006760 0.176524 0.038292 0.9695 C(3) 1.325417 7.305795 0.181420 0.8561 R-squared 0.004362 Mean dependent var 1.440596 Adjusted R-squared 0.000098 S.D. dependent var 158.3176 S.E. of regression 158.3098 Akaike info criterion 12.97335 Sum squared resid 11703956 Schwarz criterion 12.99985 Log likelihood -3045.737 Hannan-Quinn criter. 12.98378 F-statistic 1.023036 Durbin-Watson stat 1.999807 Prob(F-statistic) 0.360306 Coefficient Std. Error t-Statistic Prob. C(4) 0.001543 0.012075 0.127792 0.8984 C(5) 0.078820 0.046147 1.708026 0.0883 C(6) 1.030684 1.909873 0.539661 0.5897 R-squared 0.006264 Mean dependent var 1.129234 Adjusted R-squared 0.002008 S.D. dependent var 41.42680 S.E. of regression 41.38519 Akaike info criterion 10.29009 Sum squared resid 799846.7 Schwarz criterion 10.31659 Log likelihood -2415.170 Hannan-Quinn criter. 10.30051 F-statistic 1.471791 Durbin-Watson stat 1.987909 Prob(F-statistic) 0.230577 Relationship Between Stock Markets in Africa: A Case of Five Selected Countries 175 Table 7 VAR Analysis between CSE and NSE Coefficient Std. Error t-Statistic Prob. C(1) 0.046584 0.044095 1.056452 0.2913 C(2) 0.076471 0.044022 1.737094 0.0830 C(3) 0.056116 0.044075 1.273181 0.2035 C(4) -0.008558 0.098277 -0.087078 0.9306 C(5) 0.025213 0.098897 0.254944 0.7989 C(6) 0.119564 0.097406 1.227480 0.2202 C(7) 1.645390 7.103537 0.231630 0.8169 R-squared 0.016210 Mean dependent var 1.906506 Adjusted R-squared 0.004659 S.D. dependent var 161.9408 S.E. of regression 161.5632 Akaike info criterion 13.02109 Sum squared resid 13338460 Schwarz criterion 13.07852 Log likelihood -3365.463 Hannan-Quinn criter. 13.04359 F-statistic 1.403317 Durbin-Watson stat 1.987602 Prob(F-statistic) 0.211306 Coefficient Std. Error t-Statistic Prob. C(8) -0.011730 0.019842 -0.591200 0.5546 C(9) 0.039061 0.019809 1.971872 0.0492 C(10) -0.014749 0.019833 -0.743658 0.4574 C(11) 0.176166 0.044222 3.983633 0.0001 C(12) 0.061205 0.044502 1.375356 0.1696 C(13) -0.000653 0.043831 -0.014906 0.9881 C(14) -0.590080 3.196436 -0.184606 0.8536 R-squared 0.046126 Mean dependent var -0.801197 Adjusted R-squared 0.034926 S.D. dependent var 74.00368 S.E. of regression 72.69988 Akaike info criterion 11.42398 Sum squared resid 2700774. Schwarz criterion 11.48141 Log likelihood -2951.810 Hannan-Quinn criter. 11.44648 F-statistic 4.118342 Durbin-Watson stat 1.981709 Prob(F-statistic) 0.000475 Table 8 VAR Analysis between ESX and JSE Coefficient Std. Error t-Statistic Prob. C(1) 0.064763 0.044421 1.457926 0.1455 C(2) -0.029191 0.043686 -0.668200 0.5043 C(3) 0.063199 0.015972 3.956973 0.0001 C(4) 0.021170 0.016211 1.305930 0.1922 C(5) 12.99107 12.93924 1.004005 0.3159 R-squared 0.040679 Mean dependent var 18.46586 Adjusted R-squared 0.033110 S.D. dependent var 295.6569 S.E. of regression 290.7210 Akaike info criterion 14.19232 Sum squared resid 42850994 Schwarz criterion 14.23371 Log likelihood -3628.235 Hannan-Quinn criter. 14.20855 F-statistic 5.374669 Durbin-Watson stat 1.989281 Prob(F-statistic) 0.000303 Coefficient Std. Error t-Statistic Prob. C(6) -0.009097 0.123537 -0.073639 0.9413 C(7) 0.013389 0.121491 0.110201 0.9123 C(8) -0.034324 0.044418 -0.772757 0.4400 C(9) -0.082337 0.045083 -1.826322 0.0684 C(10) 61.72324 35.98450 1.715273 0.0869 R-squared 0.007877 Mean dependent var 55.13189 Adjusted R-squared 0.000050 S.D. dependent var 808.5257 S.E. of regression 808.5056 Akaike info criterion 16.23797 Sum squared resid 3.31E+08 Schwarz criterion 16.27936 Log likelihood -4151.920 Hannan-Quinn criter. 16.25419 F-statistic 1.006359 Durbin-Watson stat 2.012551 Prob(F-statistic) 0.403651 Table 9 Interdependencies between ESX and NGSE Coefficient Std. Error t-Statistic Prob. C(1) 0.110894 0.046372 2.391400 0.0172 C(2) -0.051656 0.046316 -1.115307 0.2653 C(3) 0.518772 0.332114 1.562028 0.1190 C(4) -0.304757 0.332492 -0.916585 0.3598 C(5) 14.06164 13.67194 1.028504 0.3042 R-squared 0.021758 Mean dependent var 15.07299 Adjusted R-squared 0.013325 S.D. dependent var 297.3441 S.E. of regression 295.3564 Akaike info criterion 14.22485 Sum squared resid 40477235 Schwarz criterion 14.26910 Log likelihood -3330.727 Hannan-Quinn criter. 14.24226 F-statistic 2.580085 Durbin-Watson stat 1.987408 Prob(F-statistic) 0.036739 Coefficient Std. Error t-Statistic Prob. C(6) -0.006362 0.006490 -0.980363 0.3274 C(7) 0.002171 0.006482 0.334918 0.7378 C(8) 0.088230 0.046478 1.898324 0.0583 C(9) -0.067828 0.046531 -1.457686 0.1456 C(10) 1.074114 1.913335 0.561383 0.5748 R-squared 0.013089 Mean dependent var 1.043220 Adjusted R-squared 0.004581 S.D. dependent var 41.42900 S.E. of regression 41.33399 Akaike info criterion 10.29185 Sum squared resid 792743.5 Schwarz criterion 10.33610 Log likelihood -2408.439 Hannan-Quinn criter. 10.30926 F-statistic 1.538481 Durbin-Watson stat 2.003774 Prob(F-statistic) 0.189890 Abdifitah Mohamed Jama DEER & Ayben KOY 176 Table 10 Interdependencies between ESX and NSE Coefficient Std. Error t-Statistic Prob. C(1) 0.086178 0.044657 1.929774 0.0542 C(2) -0.033226 0.044726 -0.742880 0.4579 C(3) 0.131392 0.184723 0.711292 0.4772 C(4) 0.011945 0.184111 0.064880 0.9483 C(5) 17.44468 13.09235 1.332433 0.1833 R-squared 0.009678 Mean dependent var 18.46586 Adjusted R-squared 0.001865 S.D. dependent var 295.6569 S.E. of regression 295.3811 Akaike info criterion 14.22413 Sum squared resid 44235733 Schwarz criterion 14.26552 Log likelihood -3636.377 Hannan-Quinn criter. 14.24035 F-statistic 1.238685 Durbin-Watson stat 1.995695 Prob(F-statistic) 0.293468 Coefficient Std. Error t-Statistic Prob. C(6) 0.016793 0.010706 1.568482 0.1174 C(7) 0.007208 0.010723 0.672220 0.5017 C(8) 0.143088 0.044286 3.230979 0.0013 C(9) 0.040409 0.044140 0.915472 0.3604 C(10) -0.138339 3.138812 -0.044074 0.9649 R-squared 0.034387 Mean dependent var 0.406465 Adjusted R-squared 0.026768 S.D. dependent var 71.78313 S.E. of regression 70.81586 Akaike info criterion 11.36776 Sum squared resid 2542547. Schwarz criterion 11.40915 Log likelihood -2905.147 Hannan-Quinn criter. 11.38399 F-statistic 4.513706 Durbin-Watson stat 2.003667 Prob(F-statistic) 0.001365 Table 11 Interdependencies between JSE and NGSE Coefficient Std. Error t-Statistic Prob. C(1) -0.047488 0.046644 -1.018095 0.3092 C(2) -0.091164 0.047347 -1.925457 0.0548 C(3) -0.053769 0.047476 -1.132544 0.2580 C(4) 1.047503 0.952680 1.099533 0.2721 C(5) -1.047110 0.950826 -1.101264 0.2714 C(6) -0.573130 0.934377 -0.613382 0.5399 C(7) 53.83033 38.54297 1.396631 0.1632 R-squared 0.018277 Mean dependent var 44.61006 Adjusted R-squared 0.005500 S.D. dependent var 831.8252 S.E. of regression 829.5345 Akaike info criterion 16.29445 Sum squared resid 3.17E+08 Schwarz criterion 16.35650 Log likelihood -3805.901 Hannan-Quinn criter. 16.31887 F-statistic 1.430452 Durbin-Watson stat 2.004189 Prob(F-statistic) 0.201111 Coefficient Std. Error t-Statistic Prob. C(8) 0.009571 0.002289 4.181122 0.0000 C(9) 0.002499 0.002323 1.075391 0.2828 C(10) 0.000829 0.002330 0.355752 0.7222 C(11) 0.059058 0.046752 1.263225 0.2071 C(12) -0.080623 0.046661 -1.727851 0.0847 C(13) 0.007439 0.045854 0.162224 0.8712 C(14) 0.444316 1.891456 0.234907 0.8144 R-squared 0.048411 Mean dependent var 0.998996 Adjusted R-squared 0.036026 S.D. dependent var 41.46225 S.E. of regression 40.70853 Akaike info criterion 10.26560 Sum squared resid 763962.1 Schwarz criterion 10.32765 Log likelihood -2395.150 Hannan-Quinn criter. 10.29001 F-statistic 3.908846 Durbin-Watson stat 2.000432 Prob(F-statistic) 0.000805 Relationship Between Stock Markets in Africa: A Case of Five Selected Countries 177 Table 12 Interdependencies between JSE and NSE Coefficient Std. Error t-Statistic Prob. C(1) -0.038387 0.044022 -0.872002 0.3836 C(2) -0.100111 0.044143 -2.267890 0.0238 C(3) -0.067616 0.044431 -1.521825 0.1287 C(4) 0.712840 0.495129 1.439705 0.1506 C(5) -0.519209 0.497380 -1.043888 0.2970 C(6) -0.748027 0.486337 -1.538082 0.1246 C(7) 62.81324 35.63379 1.762744 0.0785 R-squared 0.024138 Mean dependent var 53.35201 Adjusted R-squared 0.012680 S.D. dependent var 809.9353 S.E. of regression 804.7840 Akaike info criterion 16.23245 Sum squared resid 3.31E+08 Schwarz criterion 16.28988 Log likelihood -4197.204 Hannan-Quinn criter. 16.25495 F-statistic 2.106601 Durbin-Watson stat 1.996998 Prob(F-statistic) 0.051053 Coefficient Std. Error t-Statistic Prob. C(8) 0.013634 0.003935 3.464595 0.0006 C(9) 0.007516 0.003946 1.904845 0.0574 C(10) -0.000530 0.003972 -0.133561 0.8938 C(11) 0.156360 0.044260 3.532746 0.0004 C(12) 0.062720 0.044461 1.410648 0.1590 C(13) 0.007614 0.043474 0.175148 0.8610 C(14) -1.693260 3.185356 -0.531576 0.5953 R-squared 0.065941 Mean dependent var -0.801197 Adjusted R-squared 0.054974 S.D. dependent var 74.00368 S.E. of regression 71.94080 Akaike info criterion 11.40299 Sum squared resid 2644670. Schwarz criterion 11.46042 Log likelihood -2946.373 Hannan-Quinn criter. 11.42549 F-statistic 6.012449 Durbin-Watson stat 1.976492 Prob(F-statistic) 0.000004 Table 13 Interdependencies between NGSE and NSE Coefficient Std. Error t-Statistic Prob. C(1) 0.081690 0.048072 1.699342 0.0899 C(2) -0.098379 0.047867 -2.055249 0.0404 C(3) -0.016649 0.048085 -0.346230 0.7293 C(4) -0.002083 0.027798 -0.074922 0.9403 C(5) 0.069184 0.027998 2.471063 0.0138 C(6) 0.001038 0.027738 0.037411 0.9702 C(7) 1.113673 1.909197 0.583320 0.5600 R-squared 0.024226 Mean dependent var 0.998996 Adjusted R-squared 0.011526 S.D. dependent var 41.46225 S.E. of regression 41.22261 Akaike info criterion 10.29070 Sum squared resid 783379.1 Schwarz criterion 10.35275 Log likelihood -2401.023 Hannan-Quinn criter. 10.31511 F-statistic 1.907556 Durbin-Watson stat 1.996824 Prob(F-statistic) 0.078020 Coefficient Std. Error t-Statistic Prob. C(8) 0.167162 0.082551 2.024953 0.0434 C(9) 0.056686 0.082200 0.689607 0.4908 C(10) 0.040869 0.082574 0.494939 0.6209 C(11) 0.128628 0.047736 2.694590 0.0073 C(12) 0.024133 0.048079 0.501936 0.6160 C(13) -0.032499 0.047633 -0.682271 0.4954 C(14) -1.675492 3.278556 -0.511046 0.6096 R-squared 0.037484 Mean dependent var -1.511325 Adjusted R-squared 0.024957 S.D. dependent var 71.68950 S.E. of regression 70.78927 Akaike info criterion 11.37214 Sum squared resid 2310126. Schwarz criterion 11.43419 Log likelihood -2654.080 Hannan-Quinn criter. 11.39655 F-statistic 2.992218 Durbin-Watson stat 2.005899 Prob(F-statistic) 0.007059 3.5. Impulse Response Tests An impulse response function traces the effect of a one-time shock to one of the innovations on current and future values of the endogenous variables. This study uses cholesky decomposition to determine the influence of shocks on one variable to another. Cholesky uses the inverse of the Cholesky factor of the residual covariance matrix to orthogonalize the impulses. Abdifitah Mohamed Jama DEER & Ayben KOY 178 0 40 80 120 160 1 2 3 4 5 6 7 8 9 10 Response of D(CSE) to D(CSE) 0 40 80 120 160 1 2 3 4 5 6 7 8 9 10 Response of D(CSE) to D(ESX) 0 100 200 300 1 2 3 4 5 6 7 8 9 10 Response of D(ESX) to D(CSE) 0 100 200 300 1 2 3 4 5 6 7 8 9 10 Response of D(ESX) to D(ESX) Response to Cholesky One S.D. (d.f. adjusted) Innovations ± 2 S.E. Figure 1: Interdependencies between CSE and ESX using Cholesky decomposition As shown in Figure 1, ESX is partially affected by the shocks in CSE in the first week, however the shocks applied to ESX do not affect CSX prominently. 0 40 80 120 160 1 2 3 4 5 6 7 8 9 10 Response of D(CSE) to D(CSE) 0 40 80 120 160 1 2 3 4 5 6 7 8 9 10 Response of D(CSE) to D(JSE) 0 200 400 600 800 1 2 3 4 5 6 7 8 9 10 Response of D(JSE) to D(CSE) 0 200 400 600 800 1 2 3 4 5 6 7 8 9 10 Response of D(JSE) to D(JSE) Response to Cholesky One S.D. (d.f. adjusted) Innovations ± 2 S.E. Figure 2: Interdependencies between CSE and JSE using Cholesky decomposition It is seen in Figure 2 that both CSE and JSE do not affected prominently from the shocks on each other. 0 40 80 120 160 1 2 3 4 5 6 7 8 9 10 Response of D(CSE) to D(CSE) 0 40 80 120 160 1 2 3 4 5 6 7 8 9 10 Response of D(CSE) to D(NGSE) 0 10 20 30 40 1 2 3 4 5 6 7 8 9 10 Response of D(NGSE) to D(CSE) 0 10 20 30 40 1 2 3 4 5 6 7 8 9 10 Response of D(NGSE) to D(NGSE) Response to Cholesky One S.D. (d.f. adjusted) Innovations ± 2 S.E. Figure 3: Interdependencies between CSE and NGSE using Cholesky Decomposition It is seen in Figure 3 that both CSE and NGSE do not affected prominently from the shocks on each other. Relationship Between Stock Markets in Africa: A Case of Five Selected Countries 179 0 40 80 120 160 1 2 3 4 5 6 7 8 9 10 Response of D(CSE) to D(CSE) 0 40 80 120 160 1 2 3 4 5 6 7 8 9 10 Response of D(CSE) to D(NSE) 0 20 40 60 1 2 3 4 5 6 7 8 9 10 Response of D(NSE) to D(CSE) 0 20 40 60 1 2 3 4 5 6 7 8 9 10 Response of D(NSE) to D(NSE) Response to Cholesky One S.D. (d.f. adjusted) Innovations ± 2 S.E. Figure 4: Interdependencies between CSE and NSE using Cholesky Decomposition While the short term effect is examined, it is seen that NSE is effected from the shocks on CSE on the third week. CSE responses to shocks on NSE weakly in the fourth week. 0 100 200 300 1 2 3 4 5 6 7 8 9 10 Response of D(ESX) to D(ESX) 0 100 200 300 1 2 3 4 5 6 7 8 9 10 Response of D(ESX) to D(JSE) 0 200 400 600 800 1 2 3 4 5 6 7 8 9 10 Response of D(JSE) to D(ESX) 0 200 400 600 800 1 2 3 4 5 6 7 8 9 10 Response of D(JSE) to D(JSE) Response to Cholesky One S.D. (d.f. adjusted) Innovations ± 2 S.E. Figure 5: Interdependencies between ESX and JSE using Cholesky Decomposition Figure 5 shows that ESX is influenced by JSE shocks in the second week. The changes are felt later and the effect dies down. But JSE react immediately to own shock in the first week. Abdifitah Mohamed Jama DEER & Ayben KOY 180 0 100 200 300 1 2 3 4 5 6 7 8 9 10 Response of D(ESX) to D(ESX) 0 100 200 300 1 2 3 4 5 6 7 8 9 10 Response of D(ESX) to D(NGSE) 0 10 20 30 40 1 2 3 4 5 6 7 8 9 10 Response of D(NGSE) to D(ESX) 0 10 20 30 40 1 2 3 4 5 6 7 8 9 10 Response of D(NGSE) to D(NGSE) Response to Cholesky One S.D. (d.f. adjusted) Innovations ± 2 S.E. Figure 6: Interdependencies between ESX and NGSE using Cholesky Decomposition Figure 6 shows that, shocks in NGSE stock prices influences NGSE stock prices in the first two weeks there is not a strong effect on ESX by the shocks on NGSE. 0 100 200 300 1 2 3 4 5 6 7 8 9 10 Response of D(ESX) to D(ESX) 0 100 200 300 1 2 3 4 5 6 7 8 9 10 Response of D(ESX) to D(NSE) 0 20 40 60 1 2 3 4 5 6 7 8 9 10 Response of D(NSE) to D(ESX) 0 20 40 60 1 2 3 4 5 6 7 8 9 10 Response of D(NSE) to D(NSE) Response to Cholesky One S.D. (d.f. adjusted) Innovations ± 2 S.E. Figure 7: Interdependencies between ESX and NSE using Cholesky Decomposition As indicated in Figure 7, the effect of shocks on ESX on NSE continues gradually up to the fourth week where it recedes, however shocks in NSE do not cause prominent effects on ESX. Relationship Between Stock Markets in Africa: A Case of Five Selected Countries 181 0 200 400 600 800 1 2 3 4 5 6 7 8 9 10 Response of D(JSE) to D(JSE) 0 200 400 600 800 1 2 3 4 5 6 7 8 9 10 Response of D(JSE) to D(NGSE) 0 10 20 30 40 1 2 3 4 5 6 7 8 9 10 Response of D(NGSE) to D(JSE) 0 10 20 30 40 1 2 3 4 5 6 7 8 9 10 Response of D(NGSE) to D(NGSE) Response to Cholesky One S.D. (d.f. adjusted) Innovations ± 2 S.E. Figure 8: Interdependencies between JSE and NGSE using Cholesky Decomposition Shocks in JSE stock prices and NGSE stock prices are immediately felt by their own stock markets. However, shocks in JSE stock prices are felt by NGSE stock prices in the first to third period. Shocks in NGSE stock prices have zero influence on JSE stock prices in the first period and subsequent periods. 0 200 400 600 800 1 2 3 4 5 6 7 8 9 10 Response of D(JSE) to D(JSE) 0 200 400 600 800 1 2 3 4 5 6 7 8 9 10 Response of D(JSE) to D(NSE) 0 20 40 60 1 2 3 4 5 6 7 8 9 10 Response of D(NSE) to D(JSE) 0 20 40 60 1 2 3 4 5 6 7 8 9 10 Response of D(NSE) to D(NSE) Response to Cholesky One S.D. (d.f. adjusted) Innovations ± 2 S.E. Figure 9: Interdependencies between JSE and NSE using Cholesky Decomposition Figure 9 show that shocks in JSE stock prices and NSE stock prices are immediately felt by their own stock markets. Shocks in NSE prices have no influence on JSE prices. However, shocks in JSE stock prices are not felt by NSE stock prices in the first but the subsequent periods feels the shock. Abdifitah Mohamed Jama DEER & Ayben KOY 182 0 10 20 30 40 1 2 3 4 5 6 7 8 9 10 Response of D(NGSE) to D(NGSE) 0 10 20 30 40 1 2 3 4 5 6 7 8 9 10 Response of D(NGSE) to D(NSE) 0 20 40 60 1 2 3 4 5 6 7 8 9 10 Response of D(NSE) to D(NGSE) 0 20 40 60 1 2 3 4 5 6 7 8 9 10 Response of D(NSE) to D(NSE) Response to Cholesky One S.D. (d.f. adjusted) Innovations ± 2 S.E. Figure 10: Interdependencies between NGSE and NSE using Cholesky Decomposition Shocks in NGSE as indicated in Fig. 10 are immediately felt in the first week by NSE. And the shocks of NSE have got weak effect on NSE in the third week. 3.6. Granger causality Granger causality is used to detect presence of either unidirectional or bidirectional causality between the stock market prices of the five selected countries. The lag length is based on the predictions done earlier in the study. Table 14 Granger Causality between CSE and ESX Dependent variable: ESX Excluded Chi-sq df Prob. CSE 7.556056 3 0.0561 All 7.556056 3 0.0561 Dependent variable: CSE Excluded Chi-sq df Prob. ESX 2.434922 3 0.4872 All 2.434922 3 0.4872 With ESX as the dependent variable, the p-value is 0.0561 which is greater than the recommended value of 0.05. Thus, CSE does not Granger cause ESX. Taking CSE as the dependent variable, the probability value for ESX is 0.4872 (greater than 0.05). We therefore conclude that ESX does not Granger cause CSE. Table 15 Granger Causality between CSE and JSE Excluded Chi-sq df Prob. CSE 0.178826 1 0.6724 All 0.178826 1 0.6724 Dependent variable: CSE Excluded Chi-sq df Prob. JSE 0.109819 1 0.7404 All 0.109819 1 0.7404 The results shown in Table 29 shows that the first null hypothesis cannot be rejected as the p-value is 0.6724. Thus, CSE does not Granger cause JSE. Also, JSE does not Granger cause CSE (p-value=0.7404 >0.05). Relationship Between Stock Markets in Africa: A Case of Five Selected Countries 183 Table 16 Granger Causality between CSE and NGSE Dependent variable: NGSE Excluded Chi-sq df Prob. CSE 0.153352 1 0.6954 All 0.153352 1 0.6954 Dependent variable: CSE Excluded Chi-sq df Prob. NGSE 0.150202 1 0.6983 All 0.150202 1 0.6983 A p-value of 0.6954 indicates that the first hypothesis cannot be rejected. Therefore, CSE does not Granger cause NGSE. The p-value for the second hypothesis is 0.6983>0.05 hence NGSE does not Granger cause CSE. Table 17 Granger Causality between CSE and NSE Dependent variable: NSE Excluded Chi-sq df Prob. CSE 5.487182 3 0.1394 All 5.487182 3 0.1394 Dependent variable: CSE Excluded Chi-sq df Prob. NSE 0.765466 3 0.8577 All 0.765466 3 0.8577 Taking NSE as the dependent variable, the probability value is 0.1394> 0.05 thus we conclude CSE does not Granger cause NSE. On the other hand NSE does not Granger cause CSE as the p-value is 0.8577. Table 18 Granger Causality between ESX and JSE Dependent variable: JSE Excluded Chi-sq df Prob. ESX 0.388657 2 0.8234 All 0.388657 2 0.8234 Dependent variable: ESX Excluded Chi-sq df Prob. JSE 16.37996 2 0.0003 All 16.37996 2 0.0003 The first hypothesis of ESX does not Granger cause JSE is not rejected as the p-value exceeds 0.05. The conclusion therefore is ESX does not Granger cause JSE. The second hypothesis of JSE does not Granger cause ESX is rejected (p-value=0.0003<0.05). Thus JSE does Granger cause ESX. Table 19 Granger Causality between ESX and NGSE Dependent variable: NGSE Excluded Chi-sq df Prob. ESX 0.961717 2 0.6183 All 0.961717 2 0.6183 Dependent variable: ESX Excluded Chi-sq df Prob. NGSE 3.162612 2 0.2057 All 3.162612 2 0.2057 Abdifitah Mohamed Jama DEER & Ayben KOY 184 The first hypothesis of ESX does not Granger cause NGSE is not rejected as the p-value exceeds 0.05 (p=0.6183). The conclusion therefore is ESX does not Granger cause NGSE. The second hypothesis of NGSE does not Granger cause ESX is not rejected (p-value=0.2057<0.05). Thus NGSE does not Granger cause ESX. Table 20 Granger Causality between ESX and NSE Dependent variable: NSE Excluded Chi-sq df Prob. ESX 6.961702 2 0.0308 All 6.961702 2 0.0308 Dependent variable: ESX Excluded Chi-sq df Prob. NSE 0.962768 2 0.6179 All 0.962768 2 0.6179 The first hypothesis is rejected (p-value = 0.0308< 0.05). The conclusion is ESX does Granger cause NSE. The second hypothesis of JSE does not Granger cause ESX is accepted (p-value=0.6179>0.05). Thus NSE does not Granger cause ESX. Table 21 Granger Causality between JSE and NGSE Dependent variable: NGSE Excluded Chi-sq df Prob. JSE 18.26396 3 0.0004 All 18.26396 3 0.0004 Dependent variable: JSE Excluded Chi-sq df Prob. NGSE 3.623450 3 0.3051 All 3.623450 3 0.3051 The first hypothesis of JSE does not Granger cause NGSE is rejected as the p-value is less than 0.05 (0.0004). The conclusion therefore is JSE does Granger cause NGSE. The second hypothesis of NGSE does not Granger cause JSE is accepted (p value = 0.3051 > 0.05). Thus NGSE does not Granger cause JSE. Table 22 Granger Causality between JSE and NSE Dependent variable: NSE Excluded Chi-sq df Prob. JSE 17.59385 3 0.0005 All 17.59385 3 0.0005 Dependent variable: JSE Excluded Chi-sq df Prob. NSE 4.296359 3 0.2312 All 4.296359 3 0.2312 The probability value for the first hypothesis (JSE does not Granger cause NSE) is less than 0.05 (p-value = 0.0005) hence its rejection. The conclusion therefore is JSE does Granger cause NSE. The p-value for the second hypothesis is greater than 0.05 (p=0.2312) hence NSE does not Granger cause JSE. Table 23 Granger Causality between NGSE and NSE Dependent variable: NSE Excluded Chi-sq df Prob. NGSE 5.261217 3 0.1536 All 5.261217 3 0.1536 Dependent variable: NGSE Excluded Chi-sq df Prob. NSE 6.116412 3 0.1061 All 6.116412 3 0.1061 Relationship Between Stock Markets in Africa: A Case of Five Selected Countries 185 The p-value for the first hypothesis is greater than 0.05 (p-value = 0.1536) hence it is not rejected. The conclusion therefore is NGSE does not Granger cause NSE. The p-value for the second hypothesis is 0.1061 > 0.05 hence we conclude that NSE does not Granger cause NGSE. 4. Conclusion As the African continent has yet to play a unique role in global economic growth, the potential of African financial markets has got an increasing importance. On this content, this study focuses on the relationship of the African stock markets. In the ten years (2009 – 2018) period, Egypt, Kenya, Morocco, Nigeria and South Africa stock markets’ relationships are analyzed by VAR models and Granger Causality test. One of the main results is that Casablanca stock exchange is not influenced by other stock markets. Both in the short term and long term any significant result is not found. In the long term, the stock index of Egypt is not influenced by the indices too. But it can be used to predict the Kenyan stock market. Though Egypt also considered not part of Africa, it can have a significant influence on weaker countries. Kenya, for instance, is seen as a weaker economy compared to the other four countries under study. While the Johannesburg stock market index is not influenced by other stock prices, South African stock market can be used to predict the Egyptian, Nigerian, and Kenya stock markets. South Africa is considered the giant economy of Africa. It is therefore not expected to be swayed by the performance of other African countries. However, an influence on other countries shows its superiority. NSE and JSE stock indices have a significant influence on the performance of Nigeria stock exchange in the long run. Nigeria and Kenya are known to have a special relationship. The two countries heavily rely on each other. South African being the giant economy is expected to determine stock prices of other economies in Africa. NGSE responds to immediate changes in JSE and ESX stock performance. The stock performance of the Nairobi stock exchange is seen to be influenced by the past performance of itself, CSE, JSE, and NGSE. Kenya heavily relies on other African countries in its investment strategies. This can, therefore, explain why the performance of other countries influences it. The NSE stock price responds to immediate performances of ESX and NGSE. Kenya also relies on the daily updates of other African countries. It particularly has a special relationship with Nigeria. The Kenyan stock market cannot predict the Morocco, Egypt, Nigeria, or South Africa stock markets. Though it is a giant in the East African community, Kenya is weak compared to the four nations under study. The study concludes that Johannesburg stock exchange plays a vital role in influencing the stock prices of other African countries. African countries can, therefore, predict their future performances using South Africa stock market. The Egyptian stock market is also useful in predicting the stock performances of weaker countries. References Adjei, B. (2015). Return and Volatility spillovers among stock and Foreign Exchange Markets: Empirical Evidence from selected African Markets, Lappeenranta University of Technology, School of Business and Management, Master’s Degree Programme in Strategic Finance and Business Analytics. African Securities Exchanges Association. 2018. ASEA. Available at http://www.africanexchanges.org/ [Accessed: 15 April 2019] Agyei-Ampomah, S., 2011. Stock market integration in Africa. Managerial Finance, 37(3), pp.242-256. Alagidede, P., 2011. Return behaviour in Africa’s emerging equity markets. The Quarterly Review of Economics and Finance, 51(2), pp.133-140. Ali, S., Butt, B. Z., and Rehman, K. 2011. Comovement between emerging and developed stock markets: an investigation through cointegration analysis. World Applied Sciences Journal, 12(4), 395-403. Abdifitah Mohamed Jama DEER & Ayben KOY 186 Brandao-Marques, L., (2016). Stock Market Liquidity in Chile. International Monetary Fund Working Paper No. 16/223. Available at SSRN: https://ssrn.com/abstract=2886391 Coetzer, J. (2018). African stock exchanges hold the key to unlocking the continent’s economic growth and development. Retrieved on 12th April 2019 on https://www.inonafrica.com/2018/11/05/african-stock-exchanges- hold-the-key-to-unlocking-the-continents-economic-growth-and-development/ Diamandis, P.F., Drakos, A.A. (2010). Financial liberalization, exchange rates and stock prices: Exogenous shocks in four Latin America countries, Journal of Policy Modelling, 33(11), pp.381-394. Hoang, H. (2017). Liquidity and Stock Returns: Evidence from Taiwan Stock Market. Unpublished Master’s Thesis Jayasuriya, S.A. (2011). Stock market correlations between China and its Emerging Market neighbors, Emerging Markets Review 12, 418–431 Lischewski & Voronkova (2012). Size, Value and Liquidity. Do They Really Matter on an Emerging Stock Market? Emerging Markets Review. 13 (1), pp. 8-25. Ly, A. (2011). Capital Market-Growth Nexus in Selected SSA Countries: A Panel and Time. Economics Honors Papers. 5.http://digitalcommons.conncoll.edu/econhp/5 Mahama, A. (2013).The State of African Stock Markets. ARCADA Mohanasundaram, T. and Karthikeyan, P., (2015). Cointegration and stock market interdependence: Evidence from South Africa, India and the USA. South African Journal of Economic and Management Sciences, 18(4), pp.475-485. Næs, R., Skjeltorp, J.A. and Ødegaard, B.A., (2011). Stock market liquidity and the business cycle. The Journal of Finance, 66(1), pp.139-176. Nairobi Securities Exchange. (2018). NSE. Available at: https://www.nse.co.ke/ [Accessed:15 April 2019] Ncube, G. and Mingiri, K.F. (2015). Stock Market Integration in Africa: The Case of the Johannesburg Stock Exchange and Selected African Countries, International Business & Economics Research Journal. 14 (2) Nigeria Stock Exchange. 2018. NSE. Available at: http://www.nse.com.ng/ [Accessed: 15April 2019] Ntim, C. (2012). Why African Stock Markets Should Formally Harmonise and Integrate their Operations. African Review of Economics and Finance, 4(1), pp. 53-72 Ntim, G. C., (2011). Why African Stock Markets Should Formally Harmonise and Integrate their Operations. Managerial Finance, 4(2), 1-15. Ntim, G. C., Opong, K. K., Danbolt, J., & Dewotor, F. S. (2011). Testing the weak-form efficiency in African stock markets. Managerial Finance, 37(3), 195-218. Osamwonyi, I. O., & Kasimu, A. (2013). Stock Market and Economic Growth in Ghana, Kenya and Nigeria. International Journal of Financial Research, 4(2), pp.83-98. Vacu, N.P. (2013). Impact of Stock Market Development on Economic Growth: Evidence from South Africa