1
Master Thesis
Business Economics
-
Integration of African Stock Markets and the potential
Benefits for International Diversification
Name: Petros Naziroglu
Student nr: 5742560
2 Abstract
This paper studies the short- and long-run dynamics between African and international stock markets. The long-run link is investigated using co-integration test and the short-run
dynamics using the Granger-causality test. Our co-integration results were mixed, but in general seem to suggest that the more mature African stock markets have slowly integrated into the global market over the past decade. More interesting was the discovery of linkages between the Chinese and some African stock markets. Our results for the short-term dynamics of the Granger test show a large one-sided influence of international markets on African stock markets. Only South Africa seems to have a significant influence globally. Overall, our results suggest that the opportunities for international diversification into Africa have been limited over the past decade.
3
Table of Content
1. Introduction p. 4
2. Literature review p. 5
2.1 Co-integration studies p. 7
2.2 International portfolio diversification p. 7
2.3 Market Efficiency p. 8
2.4 Crisis and cross-country market integration p. 9
3. Data p. 9
3.1 Descriptive Statistics p. 10
3.2 Summary Statistics p. 11
4. Methodology p. 12
4.1 Unit Root Tests p. 13
4.2 Co-integration Tests p. 14
4.2.1 Engle and Granger p. 15
4.2.2 Johansen p. 15
4.2.3 Gregory Hansen p. 17
4.3 Granger Causality Test p. 18
5. Results p. 18
5.1 Unit Root results p. 18
5.1.1 Augemented Dick Fuller p. 19
5.1.2 Phillip and Perron p. 19
5.1.3 Ziwot and Andrews p. 19
5.2 Co-integration results p. 20
5.2.1 Engle and Granger p. 20
5.2.2 Johansen p. 21
5.2.3 Gregory Hansen p. 22
5.3 Granger Causality results p. 25
6. Conclusion p. 27
7. References p. 29
4
1. Introduction
The past decade has seen a tremendous increase in international trade, capital flows, financial deregulation, and advancement in communication technology. This resulted in increased linkages between international equity markets, which became apparent during the recent financial crisis of 2007–2009. This crisis exposed the various risks that investors are facing because of cross-country contagion from equity markets. This enhanced the demand for international diversifications, but the opportunities were limited in traditional markets due to the increased level of market linkages. Therefore, investors, attracted by the growth potential and limited correlation with traditional markets, shifted towards emerging markets, and to a lesser extent, towards frontier markets. Despite the crisis, the increased level of market integration brought increased welfare by enhancing effective price discovery, development of financial institutions and markets, and by enabling companies to access new pools of
potential investors and economic progress.
Market integration has attracted a lot of attention from academics. A large part of the literature has focused only on traditional and emerging markets, and neglected the potential of frontier markets. Empirical findings suggest that traditional, mature markets are interlinked across countries, thus providing limited benefits for diversification by international investors. For emerging markets, the findings are more conflicting and distinguished as shown by studies focussed on Asian, Latin American, and East European markets. Most Asian
countries first showed regional integration, which was invoked by the crisis in 1990 and later followed by the global integration (Tan, 2002 and Yang, 2003). Latin American and East European equity markets show more conflicting results. However, more recent studies seem to predict a move towards further globalization of these markets. However, studies conducted on frontier markets are limited and show an tendency of segmentation rather the globalization Therefore, it would be interesting to study the potential opportunities that frontier markets provide for international investors in diversifying their portfolios, more specifically the possibilities of African stock markets which have remained neglected for rather a long period of time.
Most African stock markets were founded during the colonial era, but they have not
undergone a similar development as their more developed counterparts. However, the current decade, supported by economic growth, reforms of market structures, prioritization by
5
governments and rising commodity prices, has ignited a revival of these stock markets (Allen, 2011, p. 83). International capital flows have increased, corresponding to an increase in market capitalization and volume, but they are still very limited, in comparison with developed market levels. Traditionally, prior research shows that only South African and Egyptian stock markets play a leading regional role, while Nigeria is recognized for its potential (Allen, 2011, p. 83). However, the current crisis and increased money supply have increased investments into emerging and to lesser extent frontier markets. Furthermore, Piesse (2005) suggested that there has been much progress in institution building, which, in turn, has increased the regional integration of the equity markets, for example Southern African Customs Union (SACU). However, in general African stock markets are seen as segmented. Therefore, due to its low correlations and potential growth, Africa can play a significant role in international portfolio diversification by widening investment opportunity sets and thereby considerably reducing risks.
The recent economic crisis of 2007–2009 affected African economies through deteriorated terms-of-trade, reduced demand for exports, and falling foreign direct investments
(Kasekende, 2010, p. 3). However, this could have sparked the integration of African stock markets through contagion from developed markets. Prior literature shows that the Asian crisis of 1997–1998 had enhanced the regional and global integration of Asian emerging markets. Wang (2003) found that the linkages of African stock markets have weakened after the Asian crisis. However, Allen (2011) described how African market structures have grown and developed during the past decade, which questions the relevance of Wang’s findings.
Despite the potential that African markets provide to investors, they have long remained neglected in the academic literature. This paper aims to analyse the integration of African stock markets into the global economy and the resulting potential benefits for internationally diversifying investors.
This paper will contribute to the literature in the following order: First, it will add to the very limited research on the co-integration of African stock exchanges. Our findings could be useful for those international investors who seek to diversify in order to address both short and long-run benefits. Furthermore, the level of global integration has important implications for the ability to attract capital and investment. Therefore, it has large implications for the overall economy of markets, but the scope of this research lies beyond this particular aspect.1
1
6
However, the long-run dynamics still contain information about the development of African markets. This can be ascertained by the choice of a more recent time period of 2000–2012, in contrast to Wang’s (2003) choice as he had used a sample ranging from 1996–2002.
Moreover, Allen (2011) indicated that the African revival had predominantly occurred during the past decade with severe indications of the integration of these markets. This has
encouraged us to extend the time-period and enhance the relevance of our study.
Additionally, by implementing the co-integration methodology introduced by Engle and Granger (1987), as well as Johansen (1988), we have extended the prior studies of Collins and Biekpe (2003) and Forbes and Rigobon (2002) who used only correlation factors to study the integration of African stock markets in more recent studies for co-integration relations. In addition, we have extended the traditional co-integration test by employing the Gregory and Hansen (1996) test that allows endogenous breaks in the data. Finally, by adding India and China to our sample, we have broadened the notion that African stock markets are only influenced by traditional markets.
This paper is structured as follows: Section two describes the related literature. Section three describes and analyses the data. Section four describes the methodology of the co-integration test. Section five analyses the results. Section six concludes the paper.
2. Literature Review
The theory of market linkages has emerged out of various economic theories, such as the traditional theory of ‘law of one price’, the portfolio diversification theory of Markowitz (1952), the capital pricing model of Sharpe (1954 ), and the arbitrage pricing theory of Ross (1976). Though these theories have their differences, they share the common idea that if indeed risk drives the price, all correlations of financial assets as well as linkages between markets originate from investors’ risk aversion and their pricing of risk. Narayan and Smyth (2005) argued that prices could vary in the short run, but they will move towards the long-run equilibrium in the long run through arbitrageurs.2 From these perspectives, globalization will induce capital to flow to the optimal price-risk relation, which, in turn, would enable African
2 Arbitrageurs while extract value from the opportunity provided by international mispricing of risk and return of an similar asset, until the opportunity has been corrected with an new equilibrium price (including transfer cost)
7
companies to attract capital if they provide an attractive opportunity for international investors.
2.1 Co-integration studies
This topic has attracted a lot of attention from academics, especially after the methodology developed by Engle and Granger (1987) and Johansen (1988) came to the fore. Most studies focus on traditional and mature markets, which have been extensively studied, for example by Taylor and Tonks (1989), Kasa (1992) and Prasad (2003). They generally found most international and developed equity markets to be interlinked with each other. Additionally, the emerging markets have been extensively studied, with the focus mainly on three regions—Asia, Latin America, and to a lesser extent, on Central Europe—by Arshanapalli (1995) and Masih and Masih (2001), and Voronkova (2004).
In contrast, the integration of African stock markets has not been studied intensively. Wang (2003) studied the linkages between five largest, emerging African stock markets and the US market by using a generalized impulse response analysis. He only found a significant short and long-term relation between the US and South African markets, while other stock markets were segmented and played a very limited role due to a palpable lack of sufficient market structure. Collins and Biekpe (2003) used a correlation study, much similar to Forbes and Rigobon (2002), to compare the changes after the crash of the Hong Kong market in October 1997. Again, only South African and Egyptian stock markets exhibited some linkages, whereas other small African stock markets seemed to be segmented. These results were consequences of a small market size and liquidity, but primarily the lack of accessibility for foreign investors was responsible. However, Collins and Biekpe (2003) questioned the results of their study. This is because they had used the Hang Seng benchmark that is weighted towards financial firms, while most African listed companies are focused on commodities and consumer goods.
2.2 International portfolio diversification
Markowitz (1959) suggested that the primary goal of portfolio diversification is to diversify the specific risks of individual assets by selecting such assets in such a manner that all the specific risks are eliminated. Harvey (1995) suggested that investors often seek to minimize their market risks by diversifying towards other geographical markets with a low correlation. Further, Harvey (1995) argued that emerging markets and frontier markets provide an
8
opportunity because of their low correlation with more traditional markets. He, however, also recognized that correlation is time-variant and thus investors need to change their positions over time. Therefore, the integration test has important implications regarding the long-run potential for international diversifications (Narayan, 2005, p. 233). However, Chowdhry (2007) and Chen (2002) argued that portfolio diversification benefits wane for co-integrated markets over time as the independent variation of such markets reduces gradually. Hassan and Naka (1996) suggested that integrated markets provide benefits only in the short run. However, Byers and Peel (1993) argued that the stock market returns covariance, not stock market price covariance, affects the co-integration vectors. Therefore, co-integration does not affect the international portfolio diversification benefits for investors. The general view provided by prior literature is that diversifying into international stock markets cannot be effective if those markets share co-movements, or in other words, if they are co-integrated.
2.3 Market efficiency
Prior studies show conflicting results on whether efficient markets could be co-integrated. Narayan and Smyth (2005) and Granger (1987) suggested that co-integration is inconsistent with the weak form of market efficiency, because prices will move towards a long-run equilibrium for common factors like arbitrage activities. Therefore, efficient markets imply that there would be no long-run equilibrium across countries and thus there would be no international arbitrage opportunity. In contrast, Masih and Masih (2002) and Dwyer and Wallace (1992) suggested that co-movement between markets do not always imply that the markets are inefficient; it only indicates that prices are predictable through using information contained in other markets. Masih and Masih (2002) argued that market efficiency implies that investors can earn risk-adjusted excess rates of returns through predictability. However, predictability does not always imply risk-adjusted excess rates and thus there is not a
consistent link. Dwyer and Wallace (1992) also suggested that replacing the random walk condition with no-arbitrage condition would remove the link between market efficiency and co-integration. Therefore, co-integration depends on the underlying model.
Despite the limited attention for market integration, more attention has been paid on the price discovery (market efficiency) of African stock markets with conflicting results. Sweeney (2003) suggested that the emerging markets, in general, are less efficient than their mature counterparts. Therefore, there could be potential opportunities for investors as they could detect inefficiencies in the emerging markets. Jefferis and Okeahalam (1999a) employed a
9
unit roots test to examine the market efficiency of South Africa, Zimbabwe, and Botswana during the time-period of 1986-1996. They concluded that only South Africa and Zimbabwe have efficient markets. However, Hakkio and Rush (1991) argued that the unit roots test is not sufficient for market efficiency. Furthermore, Jefferis and Okeahalam (1999b) also studied the topic, by using an event study, to examine how the stock price moves after information announcement. In contrast, this study led to opposite results that Zimbabwe and Botswana where efficient, while their event study stresses the fact that the South African market is not efficient. Jefferis and Ryoo (1999) used the multiple variance ratios test to found that only South Africa follows a random walk, and therefore, concluded that it has an efficient market, while the markets of Egypt, Kenya, Morocco, Nigeria, Zimbabwe,
Botswana, and Mauritius seemed inefficient.
2.4 Crisis and cross-country market integration
Changing economic circumstance could have a significant effect on cross-country market integration. Granger and Morgenstern (1970) argued that an economic crisis accelerates the integration of stock markets in the long run due to contagion effects. This was supported by Karolyi and Stulz (1996) as they blamed a certain lack of enthusiasm from investors for stocks spilling over to other markets. Tan and Tse (2002) and Yang (2003) found that during the Asian financial crisis, the stock market linkages enhanced in the short run as well as in the long run. Although Chan, Gup and Pan (1997) found contradictory results for the 1987 crisis, they worked with a sample of 18 countries. They did not find any significant acceleration of market linkages. Wang (2003) found evidence that the relation between South Africa, Egypt, Morocco, Nigeria, and Zimbabwe weakened after the 1997–1998 crisis in all emerging markets around the world.
3. Data
3.1 Descriptive statistics
This paper aims to analyse the cross-country integration of African stock markets and international stock markets by covering 10 African and six international equity markets with the use of daily adjusted closing prices3 collected from DataStream. We have chosen these
3
10 specific developed markets because they have a significant trade relation with African stock exchanges through colonial ties, or because they play a leading role in global finance. The US and the UK are the largest stock markets in the world on the basis of market capitalization; they play a very important role in global finance. The German stock market is the main driver and the largest stock market of continental Europe. France once had strong colonial ties with the African continent and French companies, and it still has a strong presence in Africa. We also chose the DAX30 and CAC40 because mainly the largest companies have a presence in Africa. Next, the presence of Chinese and Indian companies is rapidly growing in Africa due to direct investments and trade relations. With the exception of established markets in South Africa and Egypt, African stock markets are thin and ill-equipped. However, the 10 chosen stock markets are the most developed and active stock markets in the region.
We will use daily data because it captures the dissemination of information, which is
essential to study market integrations for both long and short-run dynamic relations, as shown by Hassan and Naka (1996) and Voronkova (2004). Furthermore, it will increase the
robustness of our empirical results since it will avoid the common length issues in our
integration models. Furthermore, we will also use the domestic currency because it limits the impact of currency fluctuations to security prices and avoids distorting the empirical results in the face of sharp devaluations, which, for example, are common during an economic crisis (Chowdhry, 2007).
Our study covers the time-period of 1-3-2000 to 31-12-2012. Hakkio and Rush (1991) showed that to study the long-run cross country integration, it is more robust to use a long sample period instead of a short, high frequency sample. However, such a long sample period is vulnerable for a structural shift or break due to changing general economic environment, policy regime, and development of stock markets (Voronkova, 2003, p. 635). In particular, African stock markets have experienced a significant change due to a wave of modernization in the past decade. In Table 1, the characteristics of different stock markets have been
presented. Although, African stock markets are less developed than the traditional, mature markets; they do not differ much with regard to date of establishment
However, the numbers of listings give a distorted view since most international markets are capped by a certain amount of largest companies, while African stock markets incorporate all the listed companies. This is because mainly the largest companies operate on the
11
number of listed companies and market capitalization on African stock market are very limited, therefore all such companies are needed to be included.
3.2 Summary Statistics
In the following section, we will focus on the descriptive statistics of different stock markets mentioned in our sample. First, all daily adjusted closing prices (Pi,t) are converted into
natural logarithms of the stock market returns Rt:
(
) (1)
Where Pi,t are the adjusted closing prices of the stock market at t. Descriptive statistics for the returns of the stock markets are provided in Table 2. All African stock markets show a
positive mean return, while most developed markets have a negative mean-return. Moreover, on average, most African stock return means are higher than the European positive means, with some outliers such as Tanzania providing exceptional returns. Not surprisingly, the high return markets have a higher volatility. Again, Tanzania has been showing significant
volatility. Furthermore, the Jarque-Bera test results show that the return data is normally distributed for all stock markets in our sample. More surprising is the positive skewness of Mauritian, Tunisian, Tanzanian, Chinese, and German stock markets; it suggests that large
Table 1. Summary Statistics
Country Stock Market Listings Founded
US S&P 500 500 1957
UK FTSE 100 100 1984
Germany DAX 30 30 1988
France CAC 40 40 1987
India CNX 500 500 1992
China Hang Seng 40 1964
Egypt Egyptian Exchange* 184 1883
Kenya Nairobi Securities Exchange* 50 1954
S-Africa Johannesburg Stock Exchange* 410 1887
Nigeria Nigerian Stock Exchange* 223 1960
Ghana Ghana Stock Exchange* 34 1990
Tanzania Dar es Salaam Stock Exchange* 17 1998
Morocco Casablanca Stock Exchange* 81 1929
Mauritian Stock Exchange of Mauritius* 88 1988
Tunisia Bourse des Valeurs Mobilières de Tunis* 56 1969
Botswana The Botswana Stock Exchange 44 1989
12
positive returns are more common than large negative returns. The kurtosis exceeds three for all stock markets indicating that they significantly have fatter tails and higher peaks.
Table 2. Summary Stock Market
Country N.
Observations Mean
Std.
Dev Min Max Skewness Kurtosis
Jarque-Bera Test p- value US 3391 -0,0003 0,0019 -0,0140 0,0159 -0,1727 11,32 555,83 0.0000 UK 3391 -0,0007 0,0015 -0,0112 0,0113 -0,1538 9,40 467,46 0.0000 Germany 3391 0,0001 0,0019 -0,0106 0,0127 0,0039 7,52 350,32 0.0000 France 3391 -0,0019 0,0019 -0,0115 0,0130 0,0390 7,94 376,61 0.0000 India 3391 0,0050 0,0022 -0,0183 0,0186 -0,6087 9,96 661,09 0.0000 China 3391 0,0059 0,0024 -0,0177 0,0177 0,0701 8,51 410,74 0.0000 Egypty 3391 0,0072 0,0031 -0,0274 0,0281 -0,2232 11,45 571,94 0.0000 Kenya 3391 0,0020 0,0020 -0,0678 0,0630 -2,2742 685,35 2931,60 0.0000 S-Africa 3391 0,0046 0,0013 -0,0090 0,0069 -0,1842 6,75 318,29 0.0000 Nigeria 3381 0,0051 0,0014 -0,0398 0,0370 -1,5370 381,04 2388,99 0.0000 Ghana 519 0,0014 0,0015 -0,0124 0,0110 -0,1751 20,64 139,31 0.0000 Tanzania 976 6,4273 1,0705 -15,6648 15,6065 1,3588 114,64 601,48 0.0000 Morocco 2868 0,0037 0,0009 -0,0085 0,0047 -0,5953 10,24 565,84 0.0000 Mauritian 3391 0,0060 0,0010 -0,0090 0,0109 0,2385 21,98 854,03 0.0000 Tunesia 3391 0,0051 0,0007 -0,0062 0,0063 0,0306 14,10 637,21 0.0000 Botswana 3045 0,0057 0,0006 -0,0058 0,0109 2,2165 51,73 2037,85 0.0000
Notes: The Jarque-Bera statistic tests the
null hypothesis of a normal distribution and is distributes as a with 2 degrees of freedom for Kurtosis and Skewness.
It should be noticed that Ghana and Tanzania have a relatively small sample, in comparison4 with the other stock markets, which could severely affect the results of the test because the study focuses on long-term dynamics.
Overall, our sample of the developed market shows an average daily return of 0.0013, while African stock markets show a return of 0.6468. If we exclude Tanzania from the sample, the daily average return of 0.0045 is still higher than that of more traditional and mature markets. This slightly indicates why investing in African stocks could be beneficial for investors.
4. Methodology
The methodology section comprises three parts: first, the stationarity of the series will be examined by using three5 different unit root tests, similar as in Voronka (2004), to establish
4
13
the non-stationarity and order of integration of individual market indices return data, which is an essential condition before testing for co-integration. We begin with the Augmented Dick Fuller (ADF) (1979) test, followed by the Philip and Peron (1987) test controlling for serial correlation and the Ziwot and Andrews (1992) test to allow potential structural changes. Next, we examine the long-run relationship between international and African stock markets by using three different co-integration tests, similar as in Voronka (2004). First, we employ the conventional Engle and Granger (1987) and Johansen (1988) co-integration tests in order to examine the existence of a long-run equilibrium. In addition, we use the Gregory and Hansen (1996) test, which endogenously incorporates structural breaks. Therefore, economic shifts and policies are to be incorporated into studying long-run dynamics. Moreover, it will be interesting to compare the results from the previous test with that of the GH test, to get an understanding of the development of African stock markets and their dynamics on the
international stage. Finally, we examine the short-run dynamics between the stock markets by employing the Granger (1969) test. However, this particular test could provide useful
information for testing the general belief in prior literature that co-integrated markets only provide diversification opportunities on the short run, as mentioned in Section 2.2.
4.1 Unit Root test
The first unit root test was developed by the augmented version of the Dickey-Fuller test (1979). They introduced three stages for testing non-stationarity of time series: (i) only including an intercept; (ii) only including the deterministic trend; and (iii) including both the trend and intercept. This is represented by the equations below:
∑ (2)
∑ (3)
∑ (4)
where is the logarithmic stock market price levels, is the first difference of
logarithmic stock market price levels of each individual country, is the constant, is the
5
14
trend and are the estimated residuals. The optimal number of lags was found by using the Akaike Information Criterion (AIC).
Gilmore (2008) and others argued that the ADF test has some shortcomings as it assumes that the errors are independently distributed and have a constant variance. Therefore, Philips and Perron (1987) extended the models, allowing deviation of errors and proposing a non-parametric test which controls serial correlation. Their test has two stages: (i) only including an intercept; and (ii) including both the intercept and the trend. It is represented in the equation below:
(5)
(6)
where is logarithmic stock market price levels, is the first difference of logarithmic stock market price levels of each individual country, is the constant, is the trend, T is the number of observations, and are the estimated residuals.
Voronkova (2003) argued that certain economic changes can occur during a long-time interval, which could induce structural changes in the time-series. However, the ADF and Philips and Perron (1987) tests do not allow the possibility of a structural break, which could imply that these tests falsely reject the null-hypothesis, while the series are integrated to the same order. Therefore, we employ two versions of the Ziwot and Andrews (1992) test, thus allowing the possibility of structural changes. The test allows (i) changes in constant; or (ii) changes in intercept and slope. It is represented by the equations below:
∑ (7) ∑ (8) where are the logarithmic stock market price levels, is the first difference of
logarithmic stock market price levels of each individual country, is the constant, is the trend, and are the estimated residuals. Here, the dummy variables allow a change to the intercept. Next, XUt=1 if t > T , otherwise 0 ; or change in trend XTt= t - T if t > T, otherwise 0.
After establishing non-stationarity and the order of integration of individual series variables, we continue employing the co-integration test.
15
4.2 Co-integration Test
In the following section, we will examine if there exists a linear combination between two or more non-stationary series variables, which is stationary. In other words, we will examine the existence of a long-run equilibrium relation between two stock markets.
4.2.1 Engle and Granger
First, we employ the Engle and Granger (1987) co-integration test, which has two stages. In the first stage, the residual squares are estimated by using an Ordinary Least Square (OLS) regression. It is represented in the equation below:
(9)
̂ + +… ) (10)
where both and are stock market index returns. The estimated residual are temporary deviations from the long-run equilibrium and p is the number of variables in the equation. Next, the ADF test will be conducted on the estimated residual . Using the model below:
∑ (11)
where is the estimated residual. The number of optimal lags for k is determined by using AIC.
If the estimated residual is stationary, the null hypothesis will be rejected and there will be no long-run relationship between the two stock markets. The empirical T-distribution is not similar to the ADF test as described in Section 4.1, because the unit root test is now applied on the estimated residuals. By conducting an ADF test, similar problems, as mentioned in Section 4.1, will be deemed valid. We hope to overcome these problems, especially with the choice of optimal lags, by using the AIC. Overall, the test is relatively easy and sufficient for examining the co-integration between two variables. However, the test has the stringent restriction6 which assumes a common factor in the dynamics of the model. When this restriction does not hold, it could have a severe effect on the power of the model (Sjo, 2008, p. 11)
4.2.2 Johansen
16
Next, we will conduct a Johansen co-integration test which has more desirable statistical properties – all variables are endogenous and there are no restrictions as in the Engle and Granger test, as mentioned in Section 4.2.1. Furthermore, this test allows multiple co-integration vectors, although for the nature of this study a bivariate model is sufficient.
In this study, we employ the following Johansen (1988) co-integration test. We start with the following VAR-model:
(12)
where is the vector of the logarithmic level of stock market indices, and
is vector of residuals. We assume that are integrated to the order I (1), r are
linear combinations of and are stationary, therefore the co-integration vector, 0 < r < p, than the co-integration vectors can be represented as the following (Sjo, 2008, p. 14):
̀ (13)
where βi represents the i co-integration vector, and j represents the effect of each co-integrating vector on . Therefore, the numbers of stationary relations are similar to the number of co-integrations vectors in . It implies that all rows will be zero if there are non-stationary relations and vice versa, which means if there is a non-stationary relation, some rows will be non-zero. Therefore, the rank of establishes the number of independent vectors, which reflects the number of co-integration vectors. However, the rank is determined by the number of significant Eigen values found in , which can be found by using two different likelihood ratio tests— and —represented in the models below:
∑ (14)
) (15)
The null hypothesis , where only the first r Eigen values are non-zero, which rejects the existence of a stationary relation. The alternative is , where more than one r Eigen values are non-zero. Next, for the null hypothesis, there exists a co-integration vector which is tested against the alternative of r+1 vectors (Sjo, 2008, p. 15).
Sjo (2008) suggested that the trace is better since it is more robust to skewness and excess kurtosis. In our study, we will use both the tests to examine the co-integration of different stock markets.
17 4.2.3 Gregory and Hansen
The length of our sample interval makes it vulnerable for potential structural breaks (regime shift), which could significantly affect the power of our results. In order to address these problems, we will employ the Gregory and Hansen (1996) co-integration test to allow potential structural breaks to be incorporated into the model. Furthermore, the Gregory and Hansen (1996) test could be insightful if the null hypothesis of no co-integrations has not been rejected at the previous test. They develop three forms to capture the break, although they argue that there are many other forms of breaks. See the models below:
In the first model, they allow a shift in the constant (level shift model):
(16)
The second model accommodates a time trend (level shift model with trend):
(17)
In the final model, they allow the slope vector to shift (regime shift):
(18)
where y= , denotes the vector variable of interest for co-integration, denotes the constant before the shift, denotes the change acquiring in the constant due to the shift and
is the residual. Further, denotes the slope of the trend t. Finally, denotes the co-integration slope before the regime shift, while is the change in the slope coefficient. denotes the constant before the shift, denotes the change acquiring in the constant due to the shift. The GH models allow breaks through the dummy variable , which is represented below:
{
where the timing interval is taken over the whole interval by endogenously incorporating a break that is estimated sequentially for all the three models. Moreover, the non-stationarity is verified by the ADF test. By selecting the strongest evidence again, null hypothesis of no co-integration, choosing the smallest value of the ADF test (Voronka, 2004, p. 637).
18
4.3 Granger Causality Test
Narayan and Smyth (2005) suggested that not only long-term relationships are important. They suggested that the short-term relation plays an import role in international
diversification because it provides an understanding of the causality between two stock markets: price movements of one market enable you to predict prices in the other.
Therefore, we employ the Granger causality test to study the short-term interaction between different stock markets. This is done by using the models below:
∑ ∑ (19)
̇ ∑ ∑ (20)
where is the first difference of logarithmic stock market price levels of each individual country, is the coefficient and is the direction, denotes the constant and the residual. Moreover, Granger causality means that if x causes y, then x is useful in the short term to predict y (Stock and Watson, 2011, p. 580). In order to test the null hypothesis of Granger causality, an F-test is applied on the coefficient . The F-test represented below:
⁄ (21)
where RSSr denotes the constrained residual sum of squared, RSSu denotes the unconstrained residuals sum of squared, m denotes the number of observations, and k the number of
parameters.
5. Results
In following section we will analyse the long-run dynamics between African and
international stock market. Before using the co-integration test, we first need to determine the stationarity of the time series, by using a unit root test.
5.1 Unit Root test
The next section will provide an overview of the results from the augmented Dickey and Fuller (1981; ADF), Phillips and Perron (1987; PP) and Zivot and Andrews (1992) unit root tests. We will investigate the stationarity of the time-series for each level and the first difference independently.
19 5.1.1 Augmented Dick Fuller test
Our results show that the null hypothesis of a unit root of the level series cannot be rejected for any stock market at a significance level of 5%. However, the null hypothesis can be rejected for Tunisia and Botswana with a significance level of 10%. When we regress the first-difference of the series, we reject the null hypothesis of a unit root for all the stock markets in our sample. Therefore, we can conclude that all stock markets in our sample are integrated to the order one at a significant level of 1%. These findings are similar to Wang (2003), who found that the null hypothesis for unit root is only rejected at the first difference for the USA, South Africa, Egypt, Morocco, and Nigeria. Our results are represented in Appendix 1.
5.1.2 Philip and Peron
The results of the PP test confirm, in the majority of cases, the findings of the ADF test in Section 5.1.1. However, Nigerian and Botswana reject the null hypothesis at a significance of 10%. Overall, the stock price levels are stationary at their first differences at a significance level of 1%. Our results are represented in Appendix 2.
5.1.3 Ziwot and Andrews
Hakkio and Rush (1991) showed that a long interval should be deployed for testing stock market integration. However, long intervals are vulnerable for structural breaks due to changing general economic circumstances, policy shifts, and development of stock markets. Perron (1989) suggested that PP en ADF have a lower power and it may fail to detect structural breaks. Hence, we employ the ZA-test, which endogenously determines the structural breaks. The results of the ZA test are presented in Appendix 3.
The majority of our results confirm the findings of the PP and ADF tests. However, there seems to be two significant breaks in the data. First, the series of the Ghanaian stock market rejects the null hypothesis at a significance level of 1%, with a break on 19 May 2011. This could be due to the new method that they introduced for calculating the closing prices of equities on 4 January 2011. Next, the Nigerian stock market rejects the null hypothesis at a significance level of 5%; with a break on 5-23-2008. There seem to be small inconsistencies in the results of the test. However, Sjo (2008) suggested that although these conflicting results would affect the power of the test, it is still sufficient to employ them. Overall, international stock markets seem to have a break around the financial crisis, which is in line
20
with the economic reality, while the structural break for the African stock market seems to be scattered over the whole interval. Since we found that the series is non-stationary, we can proceed with the co-integration test.
5.2 Co-integration results
5.2.1 Engle and Granger results
Overall, there are conflicting results regarding the co-integration of different stock markets. The results show that Botswana and South Africa are integrated with the traditional, mature markets, except China and India. Piesse and Heam (2005) indicated that the African Stock market primarily revolves around four hubs7 with a high level of integration. Through these hubs, local African markets find linkages with international markets. This seems to be the case with Botswana which is strongly interlinked with the South African market. Further, our results show that the most mature markets of Egypt and South Africa are still interlinked with the international stock markets. Furthermore, the Kenyan and Nigerian stock markets, which have been marked for their potentials, seem to have made the next step by integrating
themselves to the US and UK markets. However, most African stock markets still seem to be segmented from the traditional and mature markets. On the other hand, the Chinese and Indian stock markets seem to have developed long-run dynamics with some African stock markets. Especially, the interlink age of the Chinese market with Egypt, Morocco, Tanzania and Mauritius is surprising. This could partly be explained by the regional integration between North African countries, amplifying the integration of individual national market with the Chinese. Furthermore, the Indian market shows a long-run dynamic with Tanzania and Mauritius.
7
North- Egypt, East- Nigeria, West- Kenya and South- South Africa
Table 3. Engle-Granger Co-integration test
USA France Germany UK China India
Nigeria -2,661*** -2,106 -1,955 -2,719*** -1,924 -1,487 South Africa -3,338** -3,079** -3,001* -3,341* -2,352 -2,706 Kenya -3,051** -2,259 -2,267 -3,208* -2,61 -2,242 Egypt -2,782*** -2,212 -2,187 -3,054* -2,841*** -2,361 Botswana 2,663*** -2,582*** -3,522* -3,378* -2,423 -1,697 Tunisia -2,345 -2,21 -2,453 -2,018 -1,664 -1,612 Morocco -2,381 -2,25 -2,483 -2,464 -3,462* -2,075 Tanzania -2,473 -2,232 -2,39 -2,356 -3,452* -3,713* Ghana -2,107 -2,472 -1,702 -2,556 -2,822 -2,731 Mauritius -2,331 -2,194 -2,317 -2,009 -3,539* -3,182*
21
Overall, there are conflicting results; most African markets are still segmented, but there seems to be a movement of integration. The number of integrated stock markets has
increased, especially in comparison with prior studies, while the movement of integration is not necessarily towards the traditional and developed markets, but also towards the Asian markets, more specifically the Chinese and Indian markets. From an investor’s perspective, it shows that they should take into account certain long-run relations, which have developed over the past decade, in times of international diversifications.
5.2.2 Johansen Co-integration results
The Johansen co-integration test was employed to test the trace statistic ( ) and the maximum value statistic ( . We perform the bivariate co-integration test because the multivariate model is limited to only indicating the rank of the co-integration vectors, without specifying the explicit vectors by country. The empirical findings, represented in Table 4a, show the evidence of the existence of a co-integration vector between the US, French, UK and German stock markets with Nigerian, Ghanaian, South African and Botswana stock markets. Two new co-integration vectors for Germany and France have been discovered with Nigeria and Ghana, respectively, while no integration vector for the UK and the US has been discovered with Egypt. The results for the South African and Botswana stock markets seem consistent with our prior findings in Section 5.2.1.
Table 4a. Johansen Co-Integration Test
USA France Germany
Trace Max Trace Max Trace Max
r = 0 r ≥ 1 r ≤ 1 r=2 r = 0 r ≥ 1 r ≤ 1 r=2 r = 0 r ≥ 1 r ≤ 1 r=2 Nigeria 20,49** 1,76 18,73* 1,76 15,43* 2,44 12,99 2,44 11,18 1,19 9,99 1,19 Tanzania 7,13 0,35 6,78 0,35 4,64 0,20 4,44 0,20 5,66 0,65 5,01 0,65 Kenya 9,14 1,02 8,11 1,02 8,53 1,69 6,84 1,69 18,29 3,50 14,79 3,50 Egypt 7,80 0,83 6,97 0,83 7,40 0,91 6,49 0,91 17,22 2,64 14,58 2,64 Botswana 51,13* 0,74 50,39** 0,74 51,14** 1,06 50,08** 1,06 52,10** 0,35 51,75** 0,35 Tunisia 5,41 0,00 5,40 0,00 4,02 0,00 4,02 0,00 6,06 0,00 6,06 0,00 Morocco 9,30 0,98 8,32 0,98 13,45 1,17 12,28 1,17 10,16 0,87 9,29 0,87 S-Africa 18,43* 0,99 17,44* 0,99 17,91* 2,31 15,60* 2,31 17,62* 1,43 16,18* 1,43 Ghana 24,61** 3,89 20,72** 3,89 27,08** 3,42 23,66** 3,42 29,04** 5,68 23,36** 5,68 Mauritius 6,93 0,34 5,71 0,34 4,05 0,30 3,76 0,30 6,27 0,78 5,49 0,78 Statistical significance is denoted; r = 0; λ (max): 14, 07 (5%) **, 18, 63 (1%)*. λ (trace): 15, 41 (5%) **, 20, 04 (1%)*.
22
Table 4b shows the existence of a co-integration vector between the Chinese stock market and the Ghanaian, Mauritanian, Tanzanian, Botswana and Moroccan stock markets. These findings partially correspond with the results of the E-G test in the previous section for Morocco, Tanzania, and Mauritius. However, they also contradict the findings for the Botswana and Ghanaian stock markets. Finally, the Indian stock markets only show a co-integration relation with the Mauritanian and Botswana markets, which corresponds with our findings in the previous section.
Overall, the Johansen test results confirm those of the E-G test; however, the Ghanaian stock market primarily tends to differ. It should be noticed that Ghana contains relatively small samples, in comparison with other African markets.
5.2.3 Gregory and Hansen results
In the integration test conducted in the previous section, we assumed that the
co-integration vectors are time-invariant. However, economic circumstances can change, which could lead to a significance loss of power of the co-integration vector if the series does not allow the structural break to be incorporated. Greg and Hansen (1996) overcame this problem by extending the Engle-Granger test by permitting possible breaks during the long-run
relationship of the co-integration series. A general overview of the co-integration test result is provided in Table 5a and 5b.The explicit results of the Gregory and Hansen test (GH-test) are represented in Appendix 5.
The GH test results detect new co-integration relations which were not discovered in our previous co-integration test. Our results for the US confirm the previous test results for South Africa, Botswana and Nigeria, while additionally detects new co-integration vectors with
Table 4b. Johansen Co-Integration Test
UK China India
Trace Max Trace Max Trace Max
r = 0 r ≥ 1 r ≤ 1 r=2 r = 0 r ≥ 1 r ≤ 1 r=2 r = 0 r ≥ 1 r ≤ 1 r=2 Nigeria 17,70* 1,85 15,85 1,85 10,21 2,24 7,97 2,24 8,89 0,53 8,35 0,53 South Africa 20,00* 1,52 18,48* 1,48 5,49 0,74 4,74 0,74 7,54 0,10 7,44 0,10 Kenya 12,30 1,52 10,78 1,52 6,40 0,60 5,80 0,60 4,76 0,15 4,61 0,15 Egypt 12,01 1,33 10,68 1,33 9,73 0,65 9,08 0,65 4,61 0,15 4,61 0,15 Botswana 55,46* 0,72 54,74** 0,72 15,08 3,29 11,79 3,29 6,13 0,31 5,83 0,31 Tunisia 6,62 0,00 6,62 0,00 10,74 0,96 9,78 0,96 20,22 1,45 18,77 1,45 Morocco 12,65 0,95 11,69 0,95 19,89* 3,40 16,50* 3,40 13,87 0,14 13,73 0,14 Tanzania 9,62 0,29 9,33 0,29 19,60* 3,29 16,31* 3,29 15,17* 0,15 15,02* 0,15 Ghana 31,22** 2,36 28,86** 2,36 27,85** 3,53 24,33** 3,53 14,34 2,85 11,48 2,85 Mauritius 6,94 0,36 6,58 0,36 46,18** 1,25 44,94** 1,25 26,21* 1,85 24,37** 1,85
Statistical significance is denoted; r = 0; λ (max): 14, 07 (5%) **, 18, 63 (1%)*. λ (trace): 15, 41 (5%) **, 20, 04 (1%)*. r ≥ 1; λ (max): 3, 76 (5%) **, 6, 65 (1%)*. λ (trace): 3, 76 (5%) **, 6, 65 (1%)*.
23
Egypt, Kenya and Ghana. The results show a break for most countries surrounding the 2007– –2009 US crises, with exception of Egypt. Furthermore, the GH test detects two new long relations of France, with Tunisia and Morocco, while confirming the prior results for Botswana. The results confirm the long-run relation between the German stock market and Botswana and South Africa, without discovering any new long-run relation. Table 5a
provides a table with all test results; the explicit test results for the GH test are represented in Appendix 5a-c.
The results of the GH test for the UK are shown in Table 5b. They reveal new co-integration vectors between the UK with Kenya and Egypt, while the Engle-Granger-test failed to discover these long-run relations. This could be due to the estimated breaks for Egypt surrounding 8-9-2002 and 3-18-2005. These breaks could partially be explained by reorganization of their market structure by removing certain stock market constraints8 . Furthermore the GH test results for the UK detect a co-integration vector with Nigeria,
Botswana and Ghana; these results are in line with our previous results. However, the GH test fails to detect a co-integration vector between Ghana and international equity markets, which contradict the findings of the Johansen test.
8
EGX introduced a new price ceiling system which removed the 5% ceiling on daily prices, with regard to the most active stocks, based on fulfilling certain criteria. Furthermore, EGX issued a new decree to reorganize the OTC market. Finally, Capital Market Authority's Board of Directors approved the new listing rules of EGX which came into effect on 1 August 2002, ( The Egyptian Exchange)
Table 5a, Overview Co-integration Test results
US France Germany
EG Johansen GH EG Johansen GH EG Johansen GH
Nigeria x x x x x Tanzania Kenya x x Egypt x x Botswana x x x x x x x x x Tunisia x Morocco x S- Africa x x x x x x x x x Ghana x x x x Mauritius
EG (Table 3): Engle-Granger test, Johansen (Table 4): Johansen test and GH (Appendix A): Gregory-Hansen test. The results are presented in Appendix A
24
More surprising is the detection of a new co-integration vector between China, with Egypt and Nigeria. The estimated break date with Nigeria seems to fall within the 2008–2009 time period during which oil prices exhibited a significant9 drop. Elyasiani (2011) suggested that a decline in oil prices induce significant increase in returns of the Nigerian stock. This partly explains the shift on the level of movements. Additionally, the GH results confirm the co-movement of the Chinese stock market with Tanzania, Morocco, Ghana and Mauritius. It should be noticed that in contrary to the traditional stock markets, China does not have long-run dynamics with South Africa and Botswana. Finally, our results detect a new
co-integration factor between India and Kenya, while confirming our prior results for Tanzania and Mauritius.
Overall, our results confirm the findings of the co-integration test used in the previous section. Additionally, the GH test discovers new long-run relationships which were not discovered in the previous section; this could be due the break in the data sample. Our results tend to show a long-term linkage between primarily Botswana and South Africa with most mature and traditional markets, with the exception of India and China. These findings do not only support the role of South Africa as the leading stock market, but also extends the notion that after regional integration has taken place, global integration will soon follow (Piesse, 2005, p. 20). Moreover, they indicated that market integration was shaped by the Southern African Customs Union (SACU) in order to enhance global competiveness and attract new
Table 5b, Overview Co-integration Test results
UK China India
EG Johansen GH EG Johansen GH EG Johansen GH
Nigeria x x x x Tanzania x x x x x x Kenya x x x Egypt x x x x Botswana x x x Tunisia Morocco x x x South Africa x x x Ghana x x x Mauritius x x x x x x
EG (Table 3): Engle-Granger test, Johansen (Table 4): Johansen test and GH (Appendix A): Gregory-Hansen test. The results GH test are presented in Appendix A.
25
capital, which explains the newly developed long-run dynamic with the traditional, mature stock markets. Furthermore, the GH test provides more insight into certain relationships which were not discovered in prior tests, prior studies, or not studied at all. For example, the French stock market seems to have long-run dynamics with their old colonies Morocco and Tunisia. Furthermore, some African stock markets seem to have developed linkages with the Chinese and Indian stock markets, while these markets do not share an co-integration vector with the transatlantic stock markets. Our results show limited opportunities for investors for international diversification into African stock markets, in comparison to the beginning of the decade.
5.2 Granger Causality results
Next, we implement a bivariate Granger causality test to analyse the short-term interaction between different stock markets. If the long-term relation is absent, the short term relation could provide useful information regarding the existence of causal relation between the stock markets. Generally, the results show that international stock markets contain useful
information to predict movements of African stock markets in the short run, while African stock markets do not contain such information. In addition, markets which show a Granger relation also tend to have long-run dynamics
Table 6A. Granger- Causality test (Pair Indices)
F Prob F Prob F Prob
Nigeria - US 0,879 0,259 Nigeria - France 3,180 0,528 Nigeria - DE 1,898 0,755 US - Nigeria 14,505 0,006* France - Nigeria 11,530 0,021 DE - Nigeria 6,103 0,192 SA - US 0,749 0.688 SA - France 7,367 0,118 SA - DE 9,425 0,051** US- SA 43,630 0,000* France- SA 35,730 0,000* DE- SA 70,362 0,000* Kenya - US 5,675 0,225 Kenya - France 4,018 0,404 Kenya - DE 20,975 0,718 US- Kenya 17,895 0,001* France- Kenya 13,721 0,008 DE- Kenya 19,201 0,001* Egypt - US 2,897 0,575 Egypt - France 2,471 0,650 Egypt - DE 6,817 0,146 US - Egypt 50,490 0,000* France - Egypt 71,411 0,000 DE - Egypt 77,729 0,000* Botswana - US 7,059 0,170 Botswana - France 6,965 0,138 Botswana -DE 11,076 0,110 US- Botswana 40,695 0,000* France- Botswana 8,124 0,087*** DE- Botswana 37,555 0,000 Tunisia - US 6,181 0,103 Tunisia - France 7,256 0,0923*** Tunisia - DE 8,610 0,220 US - Tunisia 14,751 0,002* France - Tunisia 71,411 0,000* DE - Tunisia 0,917 0,632 Morocco - US 5,653 0,227 Morocco - France 9,147 0,058*** Morocco - DE 5,675 0,225 US - Morocco 12,228 0,016** France - Morocco 16,750 0,002* DE - Morocco 15,497 0,004*
Tanzania- US 5,653 0,227 Tanzania- France 1,479 0,830 Tanzania- DE 3,677 0,452 US -Tanzania 42,512 0,000* France -Tanzania 1,057 0,901 DE -Tanzania 15,942 0,003*
Ghana-US 9,147 0,258 Ghana-France 1,635 0,442 Ghana-DE 5,336 0,269 US - Ghana 14,718 0,005* France - Ghana 40,093 0,000* DE - Ghana 45,002 0,000* Mauritius-US 7,787 0,223 Mauritius-France 0,644 0,958 Mauritius-DE 4,105 0,392 US -Mauritius 96,740 0,000* France -Mauritius 75,030 0,000* DE -Mauritius 37,657 0,000*
26
The results, represented in Table 6a, reveal a significant one-sided short term influence of the US stock market on African stock markets, while African stock markets do not have any significant effect on the US stock market. Furthermore, the French stock market shows a significant short-term influence on most African stock markets, except Tanzania and Botswana.
Also, the German and UK stock markets have a large short-term influence on African stock markets, except on the Tunisian and Nigerian stock markets. In contrast, the South African stock market shows a short-term influence on the UK stock market. It supports the view that South Africa is the leading stock market in the continent (Piesse, 2005, p. 52). Overall, there seems to be a large short-term European and US presence and influence on African stock exchanges; this could partly be explained through colonial ties (Appiah-Kusi, 2003, p. 248). More surprising is the Chinese influence on African stock exchanges, which shows causality with most African stock markets, except Tanzania, Egypt and South Africa. It must be noted that, in contrast with other mature markets, the Chinese stock market does not have a short- term relationship with the South African market. Further, the Indian stock exchange shows
Table 6b Granger- Causality test (Pair Indices)
F Prob F Prob F Prob
Nigeria - UK 4,142 0,387 Nigeria - China 0,749 0.688 Nigeria - India 2,512 0,642 UK - Nigeria 10,608 0,031** China - Nigeria 18,813 0,001* India - Nigeria 8,143 0,086***
SA - UK 12,551 0,014** SA - China 2,471 0,65 SA - India 40,639 0,000*
UK- SA 48,048 0,000* China- SA 3,3 0,509 India- SA 3,869 0,424 Kenya - UK 7,497 0,112 Kenya - China 4,033 0,402 Kenya - India 1,858 0,762 UK- Kenya 14,746 0,005* China- Kenya 14,542 0,006* India- Kenya 12,998 0,011**
Egypt - UK 6,762 0,149 Egypt - China 5,653 0,227 Egypt - India 2,471 0,65 UK- Egypt 73,959 0,000* China- Egypt 8,124 0,087** India - Egypt 1,6346 0,442 Botswana - UK 1,635 0,442 Botswana
-China 4,61 0,23
Botswana
-India 4,621 0,328
UK- Botswana 46,973 0,000* China- Botswana 9,869 0,043* India- Botswana 0,524 0,073
Tunisia - UK 7,606 0,107 Tunisia - China 5,653 0,227 Tunisia - India 0,302 0,86 UK - Tunisia 5,592 0,232 China - Tunisia 14,21 0,007* India - Tunisia 6,762 0,149 Morocco - UK 1,898 0,755 Morocco - China 7,241 0,124 Morocco - India 0,574 0,75 UK - Morocco 16,436 0,002* China - Morocco 16,791 0,002* India - Morocco 9,147 0,0578**
Tanzania- UK 6,887 0,142 Tanzania- China 7,231 0,124 Tanzania- India 0,518 0,772 UK -Tanzania 11,457 0,022** China -Tanzania 7,206 0,125 India -Tanzania 14,308 0,006*
Ghana-UK 3,544 0,17 Ghana-China 0,675 0,714 Ghana-India 1,031 0,794 UK - Ghana 23,583 0,000* China - Ghana 28,181 0,000* India - Ghana 6,887 0,142 Mauritius-UK 5,706 0,222 Mauritius-China 5,884 0,208 Mauritius-India 10,013 0,154 UK -Mauritius 36,855 0,000* China -Mauritius 64,006 0,000* India -Mauritius 48,39 0,000*
27
only short-term causality with the Tanzanian and Mauritanian stock markets, while on the contrary on the short term it is being influenced by the South African exchange. In tables 6b, the granger results are represented for UK, China and India.
Overall, our results show that most international stock markets have a short-run relationship with African stock markets, while the long-run the relation is limited. These short term causalities seem to have limited the potential diversifications benefits for international investors. Therefore, it contradicts the belief of prior literature that integrated markets could provide benefits in the short run However, the results of short-run causality and integration could be independent from each other. Additionally, all African markets which show long-run equilibriums seem to have short-long-run dynamics.
6. Conclusion
In this paper, we have studied the short- and long-run dynamics between African and international stock markets.
Results on the long-run dynamics suggest that some African stock markets have slowly integrated into the global market over the past decade. We found the long-run dynamics between Nigeria, Botswana, Kenya, Egypt, South Africa and some traditional transatlantic stock markets. In prior studies, Nigeria, Kenya and Egypt were identified for their potential and our results suggest that they capitalized on that. South Africa, being the most mature and developed among all African stock markets, has always been the exception. Our results further strengthen this view and suggest that their influence is spilling over regionally to Botswana, as was suggested by Piesse (2005). Moreover, colonial ties seem to play an important role in stock market dynamics as is shown by France’s linkages with Tunisia and Morocco. In addition, the UK stock market shows long-run dynamics with Egypt, Kenya, and Nigeria. These relations have not been studied in prior research, but they provide an
interesting notion for investors seeking to diversify from more traditional markets. More surprising is the co-movement of the Chinese stock market with Tanzania, Morocco, Ghana, Mauritius, Egypt and Nigeria which suggests a strong presence in the African continent. In addition, we discovered a long-run dynamics between Tanzania and India which also suggest a presence, although very limited.
28
Our results for the short-term dynamics of the Granger test show a large one-sided influence of international markets on African stock markets. The US market is the most influential, followed by European markets and China, while India has a very limited influence. South Africa is the only African market which influences the international market. Noticeably, markets which are co-integrated seem to also have short-run dynamics.
Overall, the international integration of African stock markets seems to have increased over the past decade. Therefore, opportunities for international investors to diversify in Africa have been limited. However, there are still opportunities provided by less integrated markets, but investors are driven away from these because of low market capitalization and inefficient market structure. Moreover, investors should not only focus on the co-movement with traditional markets, but also take into account China and to a lesser extent India.
In this study we have focused only on the dynamics between the largest African stock markets and their international counterparts. It would be interesting to broaden the narrative and imply commodities to the study which is the fundamental driver of most African
economics. Moreover, oil price seems to play an important part since it affects both the growth of the African economies as well as the international equity markets. Therefore, it would be interesting to see if this could be a potential limitation for our study.
29
7. References
Ahmed, A. D. (2011). International financial integration, investment and economic performance in Sub-Saharan African countries, Global Economy Journal, Vol. 11 (4), pp. 1-26.
Allen , F. , Otchere , I. , and Senbet , L. W. (2011). African financial systems: A review . Review of Development Finance , Vol. 1 (2), pp. 79 – 113 .
Arshanapalli, B. and Doukas, J. (1993). “International stock market linkages:
evidence from the pre- and post-October 1987 period”, Journal of Banking and Finance, pp. 193–208.
Appiah‐Kusi, J. and Menyah, K. (2003), ‘Return Predictability in African Stock Markets’, Review of Financial Economics, Vol. 12, pp. 247–70.
Byers, J. D. and Peel, D.A. (1993). “Some evidence on the interdependence of national stock markets and the gains from international portfolio diversification”, Applied Financial Economics, Vol. 3, pp. 239–42.
Cha, B. and Oh, S. (2000). “The relationship between developed equity markets and the Pacific Basin’s emerging equity markets”, International Review of Economics and Finance, Vol. 9, pp. 299–322.
Chan, K. C., Gup, B. E., and Pan, M. S. (1997). International stock market efficiency and integration: A study of eighteen nations. Journal of Business Finance & Accounting, Vol. 24(6), pp. 803−813.
Chen, G. M., Firth, M. and Rui, O.M. (2002). “Stock market linkages: evidence from Latin America”, Journal of Banking and Finance, Vol. 26, pp. 1113–1141.
Chowdhry, T, Lin, L. and Peng, K. (2007): “Common stochastic trends among Far Eastern stock prices: effects of Asian financial crisis”, International Review of Financial Analysis, Vol. 16 (3), pp. 242-61.
Collins, D., Biekpe, N. (2003). Contagion: a fear for African equity markets? Journal of Economics and Business, Vol. 55, pp. 285-297.
Dickey, D.A. and Fuller, W. A. (1979). “Distribution of the Estimator for Autoregressive Time Series with a Unit Root.”, Journal of the American Statistical Association, Vol. 74, pp. 427-31.
Dwyer, G P J and M S Wallace (1992). “Cointegration and market efficiency”, Journal of International Money and Finance, Vol. 11(6), pp. 318-327.
Elyasiani, E., Mansur, I., Odusami, B. (2011). Oil price shocks and industry stock returns. Energy Economics .Vol. 33, pp. 966–974.
Engel, C (1996). “A note on cointegration and international capital market efficiency”, Journal of International Money and Finance, Vol. 11, pp. 497-664.
30
Eun, C. S. and Shim, S. (1989). “International transmission of stock market movements” Journal of Financial and Quantitative Analysis, Vol. 24(2), pp. 241–56.
Engle, R.F. and Granger, C.W.J. (1987). “Cointegration and Error Correction: Representation, Estimation and Testing.”, Econometrica, Vol. 55, pp. 251-76.
Forbes, K.J., Rigobon, R. (2002). No contagion, only interdependence: measuring stock market comovements. The Journal of Finance, Vol. 57, pp. 2223–2261
Gerrits, R. and Yuce, A. (1999). “Short- and long-term links among European and US stock markets”, Applied Financial Economics, Vol. 9, pp 1–9.
Granger, C. W., and Morgenstern, O. (1970). Predictability of Stock Market Prices. MA: Lexington.
Gilmore, C. G., Lucey, B. M. and McManus, G. M. (2008), “The Dynamics of Central European equity market Comovements”, The Quarterly Review of Economics and
Finance, Vol. 48, pp. 605-622.
Granger, C. W. J. (1988). “Some recent developments in a concept of causality”, Journal of Econometrics, Vol. 39, pp. 199–211.
Gregory A.W., Hansen, B.E. (1996). Residual-based tests for cointegration in models with regime shifts. Journal of Econometrics, Vol.70, pp. 99-126.
Harvey, C. R. (1995). “Predictable risk and returns in emerging markets”, Review of
Financial Studies, Vol. 8, pp. 773–816.
Hassan, M. K. and Naka, A. (1996). “Short-run and long-run dynamic linkages among international stock markets”, International Review of Economics and Finance, Vol. 5(4),pp. 387-405.
Hakkio, C. S. and Rush, M. (1991. “Cointegration: how short is the long run?”,
Journal of International Money and Finance, Vol. 10(4), pp 57l–81.
Jefferis, K. and Okeahalam, C. (1999a). “International Stock Market Linkages in Southern Africa”, South African Journal of Accounting Research, Vol. 13(2), pp. 1-25.
Jefferis, K. and Okeahalam, C. (1999b). “An Event Study of the Botswana, Zimbabwe and Johannesburg Stock Exchanges”, South African Journal of Business Management, Vol. 30(4), pp. 131-140.
Johansen, S. (1988). Statistical Analysis of Cointegrating Vectors. Journal of
Dynamic and Control, Vol. (12), pp. 231-254.
Karolyi, G. A. and Stulz, R. M. (1996).Why do markets move together? An
investigation of U.S.–Japan stock return co-movements. Journal of Finance, Vol. 51(3), pp. 951−986.
31
Kasa, K (1992): “Common stochastic trends in international stock markets”, Journal of Monetary Economics, Vol. 29, pp. 95–124.
Kasekende, L., Ndikumana, L., and Brixiova, Z. (2010). Africa’s counter-cyclical policy responses to the crisis. Journal of Globalization and Development, Vol. 1(1).
Masih, R. and Masih, A. M. M. (2001).“Long and short term dynamic causal transmission amongst international stock markets”, Journal of International Money and Finance, Vol. 20, pp. 563-87.
Masih, A. M. M. and Masih, R.(2002). “Propagative causal price transmission among international stock markets: evidence from pre- and post- globalisation period”, Global Finance Journal, Vol. 13, pp. 63-91.
Markowitz, H (1952). “Portfolio selection”, Journal of Finance, Vol.7(1), pp. 77-91.
Narayan, P. K. and Smyth, R. (2005). “Cointegration of stock markets between New Zealand, Australia and the G-7 economies: searching for co-movement under structural change”, Australian Economic Papers, Vol. 44(3), pp. 231-247.
Prasad, E. S., Kenneth, R., Shang-Jin, W. and Ayhan, K. (2003). “Effects offinancial globalization on developing countries: some empirical evidence”, IMF Occasional Papers, no 220.
Sharpe, W. (1964). “Capital asset prices: a theory of market equilibrium under conditions of risk”, Journal of Finance, Vol. 19, pp. 425–42.
Sjo, B. (2008). Testing for Unit Roots and Cointegration, AERC Technical Workshop
on Time Series Econometrics, pp. 10-21.
Smith G, Jefferis, K. and Ryoo, H.J. (2002). African stock markets: Multiple variance ratio tests of random walks. Applied Financial Economics Journal, Vol.12, pp. 475-484.
Stock and Watson. (2012), “Introduction to Econometrics”, Pearson: International Edition 3rd edition.
Tan, K. B., and Tse, Y. K. (2002). The integration of the East and South-East Asian equity markets. International Center for the Study of East Asian Development, Working Paper, No. 2002-11
Taylor, M. P. and Tonks, I. (1989). “The internationalization of stock markets and abolition of UK exchange control”, Review of Economics and Statistics, Vol. 71, pp. 332–356.
Phillips, P. C. B. and Perron, P. (1988). “Testing for a Unit Root in Time Series Regressions.”, Biometrika, Vol. 75(1), pp. 335-346.
Piesse, J. and Hearn, B. (2002). Equity market integration versus segmentation in three dominant markets of the Southern African Customs Union: co-integration and causality tests, Journal of Applied Economics, Vol. 34(18), pp. 1711–22.
32
Ross, S.(1976). “The arbitrage theory of capital asset pricing”, Journal of Economic Theory, Vol. 13(3), pp. 341-360.
Wang, Z., Yang, D. and Bessler, A. (2003). “Financial crisis and African stock market integration”, Applied Economics Letters, Vol. 10, pp. 527-33.
Voronkova, S. (2004). ”Equity market integration in Central European emerging markets: a cointegration analysis with shifting regimes”, International Review of Financial Analysis, Vol. 13, pp. 633-647.
Yang, J., Kolari, J., and Min, I. (2003). Stock market integration and financial crisis: The case of Asia. Applied Financial Economics, Vol.13(7), pp. 477−486.
Zivot, E. and Andrews, D. W. K (1992), “Further evidence on the great crash, the oil-price shock, and the unit-root hypothesis”, Journal of Business & Economic Statistics, Vol.10, pp. 251-270.
