Chapter 8
Investing on the Edge: Exploring the Opportunities for Diversification in Frontier Markets
A. Spiru
Z. Qin Lancaster University Management School, Lancaster, United Kingdom
Abstract
The increase in the degree of integration of emerging stock markets with the mature markets together with rising correlations among asset classes have posed new challenges to investors in their quest for diversification opportunities. Frontier markets have risen up to the challenge, opening avenues for greater diversification benefits compared to their emerging counterparts. This study attempts to shed light on how the characteristics of frontier markets can enable investors to reap greater benefits from portfolio diversification. To this end, we examine the extent of integration of fifteen Central and Eastern European (CEE) and Middle Eastern and North African (MENA) frontier markets, both with mature markets and within their own group, over the period from Dec. 2005 to May 2015. For this purpose, we employ correlation and cointegration techniques to assess both short- and long-run comovements. The results suggest a greater degree of integration at the regional level. At the global level, the selected CEE frontier markets are more closely linked with their developed counterparts than the MENA frontier markets considered, suggesting that the MENA markets may have greater potential for diversification gains.
Keywords
frontier markets
stock market comovements
cointegration analysis
diversification
1. Introduction
Considerable research has been devoted to the linkages across international stock markets, which have important implications for investors and policy makers. Various studies, mostly empirical in nature, have examined the extent of the interdependencies across financial markets. In his seminal work that laid the foundations of modern portfolio theory, Markowitz (1952, 1959) highlighted the benefits of diversification. Building on the mean-variance framework proposed by Markowitz, the capital asset pricing model (CAPM) (Sharpe, 1964; Lintner, 1965; Mossin, 1966), another cornerstone of modern finance, indicates that investors will be rewarded with higher expected returns only for bearing the systematic risk, since the other component of risk, the idiosyncratic (unsystematic) risk, can be reduced by holding a well-diversified portfolio. The degree of idiosyncratic risk which is diversifiable varies from country to country. Thus, international diversification is beneficial for investors, and due to the generally lower correlations between domestic and foreign stocks, it is possible to construct a portfolio that is superior to a portfolio consisting of domestic assets only (Elton et al., 1995).
The seminal work of Grubel (1968), which stresses the benefits of international portfolio diversification, has provided an impetus for the research on this topic. A large body of literature has focused on the comovements across international stock markets; see, for instance, King et al. (1994); Lin et al. (1994); Longin and Solnik (1995, 2001); Karolyi and Stulz (1996); Forbes and Rigobon (2002); Brooks and Del Negro (2005, 2006), among others. The extent of these comovements has important implications for asset allocation and risk management, representing a key issue in the quest for reaping the benefits of diversification.
As documented by the majority of extant studies that investigate it, the comovement of stock returns across markets varies over time. Moreover, the degree of comovement depends on the stage of development of the markets under scrutiny. The early studies mainly focused on the mature financial markets, such as those of the United States, United Kingdom, Germany, France, and Japan. However, financial liberalization together with advancements in technologies used for global trading and information transmission have led to growing interlinkages across markets, with global markets tending to react to some common factors. This has reduced the extent of independent market movements and created stronger comovements, which limited the potential benefits of diversification, since there were effectively fewer low-correlated assets that could be used to diversify. Many studies—Brooks and Del Negro (2004) and Kizys and Pierdzioch (2009), inter alia—have reported evidence of increasing international comovement of stock returns among the major developed countries since the mid-1990s.
With the portfolio diversification effect fading away within the group of mature markets, researchers and practitioners have shifted their focus to the emerging markets. Two to three decades ago, when emerging economies were at an earlier stage of development, their stock markets were only weakly correlated with those in the developed markets, hence being good candidates for diversification. However, over the past 25 years or so, emerging markets have become more global and, as documented in the literature (for a survey, see Section 2), have started to have common trends with their developed counterparts, diminishing their potential for diversification. This has prompted investors, in their quest to get back some of the lost diversification benefit that they used to get from investing in emerging markets, to look toward a new category of markets, the frontier markets.
The frontier markets were designated as a new category of markets by the International Finance Corporation (IFC) in 1996. They can be regarded as a special subset of the emerging markets, representing the “emerging emerging markets.”a This next generation of emerging markets consists mainly of African, Middle Eastern, and former transition economies from Central and Eastern Europe (CEE) or the former USSR, and represents about 0.2% of the total global equity investment opportunity set. They started to attract increasing interest from both investorsb and researchers. For investors, the attractiveness of the frontier stock markets stems mainly from the high returns provided in the past and from segmentation with both developed markets and emerging markets, which can serve as a useful tool to reduce the unsystematic risk of a portfolio. In addition to offering potential diversification gains, frontier markets are attractive to investors due to their growth potential, being expected to deliver higher growth than both the developed and the emerging markets.c Another point worth noting is that frontier markets have lower volatility than the emerging markets. However, the potential benefits of investing in frontier markets are accompanied by the risks these markets pose. The frontier markets are characterized by low liquidityd and small market capitalization. A lack of liquidity causes realized gains to be less than those implied by the market. Another major risk of investing in frontier markets is represented by the political and economic instability witnessed by some of them.
Having made the case for embracing the frontier markets from the point of view of the practitioners, we will now turn our attention to researchers. While the developed and emerging markets have been examined by a large body of literature, frontier markets have started to gain increasing attention in the literature exploring the linkages across financial markets after they have been designated as a separate category of developing markets. Therefore, in comparison with their mature and emerging counterparts, relatively few studies have examined the frontier stock markets, making this area of research suitable for further empirical investigation. Drawing motivation from that, in this study we set out to investigate the short- and long-run linkages among 15 frontier markets from CEE and the Middle East and North Africa (MENA) regione and three mature markets, namely Germany, the United Kingdom and the United States, over a period that spans 10 years (2005–15). By means of correlation and cointegration analysis, we examine the potential for diversification benefits from investing in the aforementioned frontier stock markets by shedding light on the extent of their integration with respect to mature markets.
Before proceeding with the other sections of this study, we present the underlying motivation for our selection of frontier markets. The first remarkable and compelling characteristic of the frontier markets examined in this study is the high speed of economic growth, as indicated by the gross domestic product (GDP) growth rate. As can be seen in Fig. 8.1, during most of the period covered by this study, even in the most difficult economic times, the CEE and MENA frontier markets displayed, on average, a faster increase in their GDPs than the three developed countries considered here (the United States, the United Kingdom, and Germany).
Figure 8.1 GDP growth rate comparison between CEE and MENA frontier markets and mature markets.
Note: The comparison stops in 2013 as the GDP growth rates for 2014 were not available for Kuwait, Oman, and Tunisia at the time of computation. (From authors’ calculations, based on World Bank data.)
Note: The comparison stops in 2013 as the GDP growth rates for 2014 were not available for Kuwait, Oman, and Tunisia at the time of computation. (From authors’ calculations, based on World Bank data.)
Furthermore, the 15 frontier markets selected for this study account for 70.48% of the total market capitalization of the whole frontier group, while their GDPs amount to 48.34% of the group (Tables A1 and A2), which indicates that the selected economies occupy an important position in the whole frontier group and can be regarded as suitable representatives for the frontier universe.
The frontier stock markets from CEE are of particular research interest, given their potential for accelerated economic growth and the attribute of regulated markets linked to the advantages of EU membership. As for the countries from the MENA region, a series of reforms aiming at liberalizing the financial services industry, such as revising the laws on stock trading, were successively launched during the past 20 years. As a result, the economic development of countries in the MENA region is now considered to have strong support from the banking sector. Despite the varying degrees of improvement of stock markets in these countries, substantive efforts are being made to boost investment in this region from both the domestic and the international financing sources (Graham et al., 2013).
The remainder of this chapter is organized as follows. Section 2 provides a selective review of the literature on international stock market linkages, placing special attention on studies that examine frontier markets. Section 3 introduces the data. Section 4 presents and discusses the results of the empirical analysis carried out. The last section concludes.
2. Selective Review of the Literature
As stated in the introductory section, the linkages among stock markets have attracted a vast amount of interest. Most of the earlier literature has focused on the developed markets like those of the United States, Japan, and major European countries (United Kingdom, Germany, France, etc.), especially after the 1987 stock market crash, which gave an impetus to this strand of studies (Kasa, 1992; Arshanapalli and Doukas, 1993; Lin et al., 1994; Longin and Solnik, 1995; Ben-Zion et al., 1996; Kanas, 1998; Gerrits and Yuce, 1999; Engsted and Tanggaard, 2004; Goetzmann et al., 2005; Carrieri et al., 2007; Tanizaki and Hamori, 2008). The evidence provided by a number of studies in this area (Brooks and Del Negro, 2004; Pukthuanthong and Roll, 2009) shows that international stock markets have become increasingly interdependent, the most important implication of this finding being that the level of the benefits from diversifying across countries has decreased (Kearney and Lucey, 2004).
Following increased integration among mature markets, the emerging markets, offering an alternative for reaping diversification benefits, have attracted increasing attention from both the research and the investment communities. As a result, a stream of literature, focusing on the linkages between developed and emerging stock markets, started to count an increasing number of contributions. Employing a variety of methodological approaches, which include vector autoregressions and cointegration (Syriopoulos, 2006; Yang et al., 2006; Diamandis, 2009, inter alia), multivariate GARCH modeling (Li and Majerowska, 2008; Beirne et al., 2010; Syllignakis and Kouretas, 2011) and the use of high-frequency data (Černý and Koblas, 2005, 2008; Égert and Kočenda, 2007), most of these studies focus on short- and long-term stock market linkages for specific regions, such as Latin America, Southeast Asia, or CEE. More recently, several studies have addressed the issues of stock market integration and linkages of the emerging markets in the MENA region (Ben Naceur et al., 2007; Cheng et al., 2010; Neaime, 2012; Graham et al., 2013). Earlier research on the linkages among the markets from CEE and the developed markets focused on three major stock markets in that part of Europe, namely the Czech Republic, Hungary, and Poland (Voronkova, 2004; Fadhlaoui et al., 2009; Syriopoulos, 2004, 2007), often with mixed findings. For instance, Syriopoulos (2004, 2007) and Voronkova (2004) document the existence of a long-term equilibrium of those markets with their mature counterparts, while no long-term relationship is found by Gilmore et al. (2008). Fadhlaoui et al. (2009) examine the potential benefits of investing in the Czech, Hungarian, and Polish stock markets. They find low correlations between these three CEE markets and the G7 developed markets and no evidence of a common stochastic trend between the two groups of markets, suggesting that in both the short run and the long run investors can realize diversification benefits by including the three CEE markets in their portfolios.
With the emerging markets becoming more and more integrated with their developed counterparts, the quest for diversification opportunities has directed increasing interest to a more recently established category of markets, the frontier markets. When they appeared in the investment landscape, frontier markets received substantive attention from the investment community, in particular from institutional investors. Following that, an emergent strand of the literature on financial market linkages has started to devote attention to these markets, transforming this category of markets from uncharted territory into the next frontier of investing.
As discussed in the introduction, frontier markets may offer promising diversification opportunities. In support of this, Speidell and Krohne (2007) provide evidence of low correlations between frontier and developed stock markets. Jayasuriya and Shambora (2009) investigate diversification benefits across market classifications by analyzing optimal portfolios composed of developed, emerging, and frontier markets. They report improved portfolio risk–return trade-offs when investors diversify their portfolios into six frontier markets. Cheng et al. (2010) use variations of the CAPM to study nine emerging and frontier equity markets within the MENA region. Their results suggest that the majority of markets within their sample exhibit low levels of integration. Berger et al. (2011) examine the diversification potential for a set of frontier markets and find that those markets exhibit low levels of integration with the world market, hence offering significant diversification benefits. The empirical analysis they carried out also suggests that there are no signs of growing integration over time between developed and frontier markets. In a later study, Berger et al. (2013) argue that when assessing the potential for diversification benefits of frontier markets, which are typically characterized by small market capitalization, thin trading, and restrictions for foreign investors, the focus should be on frontier exchange-traded funds (ETFs). Their empirical results confirm the good performance of frontier ETFs in terms of portfolio risk reduction via diversification. The analysis carried out by Marshall et al. (2015) takes the transaction cost for investment in frontier markets into consideration, establishing a model to examine the trade-off between the potential costs and benefits of including frontier market stocks in the portfolio.
As the empirical analysis carried out later in this chapter focuses on frontier markets from CEE and MENA, in what follows we will provide a selective review of studies that examine these markets. While, as discussed earlier in this section, a larger number of studies investigate the linkages between the CEE emerging markets and mature markets, rather little research has been conducted on the subset of frontier markets in Europe. Mateus (2004) provides evidence in support of partial integration of five CEE frontier stock markets (Bulgaria, Estonia, Lithuania, Romania, and Slovenia) with respect to the world market. Samitas et al. (2006) investigate the linkages between Balkan stock markets (Albania, Bulgaria, Croatia, Macedonia, Romania, Serbia, and Turkey) and developed stock markets, while Syriopoulos (2011) examines the degree of financial integration of six markets in the Balkan region (Romania, Bulgaria, Croatia, Turkey, Cyprus, and Greece). The empirical results on the long-term equilibrium between the Balkan stock markets and the developed markets suggest limited diversification benefits in the long term, while in the short term some benefits might still be feasible. Maneschiöld (2006) demonstrates that Baltic markets can provide diversification benefits for international investors over a long-run investment horizon. Wang and Shih (2013) provide evidence of time-varying integration in emerging European markets (including five frontier markets).
While the most remarkable feature of the CEE frontier markets is that their economies have undergone a complex transition from plan to market, the MENA region is known for the most abundant oil reserves and the largest oil-exporting trade volume, a feature acknowledged by the studies that focus on the frontier markets from the region (see, inter alia, Gomes and Chaisi, 2014). For instance, to get a more accurate representation of the financial linkages of MENA countries, Yu and Hassan (2008) divided them into two subsets, depending on whether they are members of the Gulf Cooperation Council (GCC). Their empirical results provide two interesting insights: On the one hand, they suggest increasingly stronger integration between the non-GCC economies and the US market, which may be attributed to the earlier openness and availability of these markets. On the other hand, the opposite finding is reported for the GCC markets, which can serve as evidence that the equity markets of Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, and United Arab Emirates offer opportunities for diversification; hence their selection into an investor’s international portfolio has the potential to lower risk. A similar conclusion is derived for the MENA markets by Graham et al. (2013), who employ wavelet analysis. Since these markets do not share a common trend with the US market, they constitute good candidates for international investors in their attempt to reduce portfolio risk by diversifying across different stock markets worldwide.
Cheng et al. (2010) examine the interdependencies among the MENA markets from the perspective of international finance. Based on the static international CAPM and a Markov switching model, they conclude that only two economies from this area, namely Israel and Turkey, exhibit close integration with other mainstream stock exchanges. Therefore, incorporating stocks from MENA markets into their portfolios can benefit international investors. More interestingly, Cheng et al. (2010) provide empirical evidence in support of a relationship between the oil price fluctuations and the level of integration of the MENA stock markets with other important stock markets worldwide. More exactly, the authors argue that the price of oil is negatively correlated with the degree of integration.
3. Data
The data set used in this study consists of weekly returns for 18 frontier and mature stock markets as follows: Bahrain, Bulgaria, Croatia, Estonia, Germany, Jordan, Kuwait, Lebanon, Mauritius, Oman, Qatar, Romania, Saudi Arabia, Slovenia, Tunisia, United Arab Emirates, United Kingdom, and United States. The frontier markets included are selected based on the Standard & Poor’s (S&P) classification of frontier markets. Five of them (Bulgaria, Croatia, Estonia, Romania, and Slovenia) are markets from CEE, while the other 10 (Bahrain, Jordan, Kuwait, Lebanon, Mauritius, Oman, Qatar, Saudi Arabia, Tunisia, and United Arab Emirates) belong to the MENAf region. Six of the ten MENA markets considered here are represented by the Gulf Cooperation Council (GCC) countries (Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, and United Arab Emirates). The mature markets are represented in our study by Germany, the United Kingdom, and the United States.
To compute the returns for the frontier markets, we use the Morgan Stanley Capital International (MSCI) country-specific indices for the markets selected. The MSCI indices, which are value-weighted and calculated with the dividends reinvested, are established in a consistent manner across countries,g thus providing an appropriate framework for examination of linkages across markets. All MSCI country indices are denominated in US dollars, which ensures additional comparability across countries and implicitly takes care of the effects of exchange rate fluctuations. In addition to the MSCI indices for the frontier markets selected for this study, we use the stock indices of Germany (DAX), the United Kingdom (FTSE 100) and the United States (S&P 500) as representative of the developed markets. The prices of the DAX and FTSE 100, which are traded in euros and sterling, respectively, are converted into US dollars using the relevant exchange rates.
The sample period under scrutiny here extends from Dec. 9, 2005 to May 22, 2015, representing the longest common time period of data availability and comprising 493 weekly observations.h To alleviate the potential issue of nonsynchronous trading, we employ weekly returns, which are computed as the logarithmic first difference of weekly stock price indices,i 
where Pit is the value of the closing price of the index for market i at time t. Graphs of the stock indices and returns can be found in the Appendix.
where Pit is the value of the closing price of the index for market i at time t. Graphs of the stock indices and returns can be found in the Appendix.Table 8.1 reports descriptive statistics for the returns of the 15 frontier markets and 3 developed markets under scrutiny. The statistics include the mean, median, maximum, and minimum returns for each country as well as skewness and kurtosis.
Table 8.1
Descriptive Statistics for the Returns of Selected Frontier and Mature Markets
| Panel A: CEE frontier markets | |||||
| Bulgaria | Croatia | Estonia | Romania | Slovenia | |
| Mean | −0.298 | 0.002 | −0.118 | −0.038 | −0.021 |
| Median | 0.117 | −0.007 | −0.194 | 0.493 | 0.055 |
| Maximum | 17.716 | 19.539 | 17.252 | 20.082 | 11.392 |
| Minimum | −41.597 | −23.984 | −25.216 | −34.665 | −22.015 |
| Std. deviation | 4.383 | 3.559 | 4.233 | 4.933 | 3.389 |
| Skewness | −2.145 | −0.623 | −0.462 | −1.365 | −1.128 |
| Kurtosis | 21.077 | 12.563 | 8.168 | 10.943 | 9.613 |
| Panel B: MENA frontier markets | |||||
| Bahrain | Jordan | Kuwait | Lebanon | Mauritius | |
| Mean | −0.45 | −0.27 | −0.14 | −0.03 | 0.179 |
| Median | −0.19 | −0.26 | 0.021 | −0.19 | 0.004 |
| Maximum | 10.924 | 16.848 | 16.751 | 14.928 | 11.111 |
| Minimum | −21.32 | −13.90 | −21.37 | −18.18 | −22.34 |
| Std. deviation | 3.066 | 2.877 | 3.223 | 3.159 | 3.014 |
| Skewness | −1.464 | −0.091 | −0.984 | 0.297 | −0.712 |
| Kurtosis | 12.721 | 7.819 | 11.60 | 10.103 | 11.679 |
| Oman | Qatar | Saudi Arabia | Tunisia | United Arab Emirates | |
| Mean | −0.06 | −0.02 | −0.10 | 0.076 | −0.15 |
| Median | 0.102 | 0.118 | 0.152 | 0.058 | 0.045 |
| Maximum | 13.519 | 12.865 | 13.491 | 8.618 | 20.723 |
| Minimum | −22.31 | −22.76 | −23.49 | −13.83 | −32.68 |
| Std. deviation | 2.99 | 3.529 | 4.072 | 2.417 | 4.771 |
| Skewness | −1.561 | −1.034 | −1.262 | −0.596 | −1.397 |
| Kurtosis | 14.628 | 9.851 | 9.065 | 7.628 | 10.824 |
| Panel C: mature markets | |||
| Germany | United Kingdom | United States | |
| Mean | 0.149 | 0.024 | 0.109 |
| Median | 0.679 | 0.377 | 0.218 |
| Maximum | 14.505 | 16.279 | 11.356 |
| Minimum | −26.555 | −27.819 | −20.084 |
| Std. deviation | 3.827 | 3.341 | 2.608 |
| Skewness | −1.085 | −1.418 | −0.961 |
| Kurtosis | 9.625 | 15.101 | 11.821 |
Notes: The values for the mean, median, maximum, and minimum are in %.
The mean weekly returns range from −0.45% (Bahrain) to 0.18% (Mauritius) and, as anticipated, are low, taking into consideration that the sample period under scrutiny includes the economic downturn of 2007–10. During the period considered in this study, 12 out of the 15 frontier markets considered exhibited negative average returns, which can be attributed to the substantive detrimental impact of the financial crisis during the period examined. For the CEE group, the average return ranges from almost −0.3% (Bulgaria) to 0.002% (Croatia), with four of the five CEE frontier markets recording negative mean returns, which indicates that the global financial crisis of 2008 exerted a significant impact on the markets from this region. On the one hand, international trade, which plays an important role in the CEE economies, was affected, as during the financial crisis the major trading partners of CEE economies, namely the major Western European countries like France, Germany, and the United Kingdom, were significantly negatively affected, and the external demand from these target exporting countries substantially declined. In this way, the 2008 financial crisis has, to a large extent, harmed one of the vital pillars for supporting the economic growth of the CEE countries. On the other hand, also due to the crisis, foreign direct investments into the CEE region decreased sharply. Through these two major transmission channels mentioned previously, the crisis exerted a large negative impact on the financial markets of the frontier CEE countries, as suggested by the negative average returns displayed in Table 8.1.
The average returns of the MENA frontier markets selected for this study are characterized by a wider range than their CEE counterparts, varying from −0.45% for Bahrain to −0.02% for Qatar and 0.18% for Mauritius, with 2 of the 10 markets displaying positive mean returns (Tunisia and Mauritius). The predominantly negative mean returns may, to some extent, be explained by the fact that the oil-producing economies from the MENA region have been affected by a decrease in the demand for crude oil during the global financial crisis of 2008, which has exerted a negative impact on their financial markets together with the plunging oil prices over certain subperiods of the sample. In particular, in the case of the GCC stock markets, which are part of the GCC group considered here, investors in those stock markets should look at changes in oil prices, which have predictive power for financial returns (Arouri and Rault, 2009).
The three mature markets considered (Germany, United Kingdom, and United States) were characterized by positive average weekly returns, which may be attributed to the faster recovery of these markets from the 2008 global financial crisis compared with the frontier markets. The highest weekly return within the sample (20.723%) has been recorded on the UAE stock market in the third week of Dec. 2009, while the minimum value (−41.597%) corresponds to the second week of Oct. 2008 in the Bulgarian stock market.
The standard deviation, a measure of variation of returns which makes it suitable for shedding light on the stability of the data over the sample under scrutiny, ranges from 2.417% for Tunisia to 4.933% for Romania. The volatility levels of the frontier markets are generally higher than those of the developed markets. With a few exceptions (such as Jordan and Tunisia), the standard deviation is quite high for the frontier markets, as expected, and it could result from various liquidity effects or from the presence of investors with heterogeneous information sets. The least volatile of the CEE frontier markets is Slovenia, while the most volatile is Romania, which is also the most volatile of all frontier markets considered in this analysis. In the MENA frontier market group, the most volatile is United Arab Emirates. The least volatile frontier market, Tunisia, has a volatility level below those of the mature markets. The standard deviation of returns is quite high for the mature markets as well, suggesting the turmoil in these markets. The distribution of the returns series is nonnormal,j with kurtosis exceeding 3 in all cases (indicative of a leptokurtic distribution) and negative skewness (except for Lebanon, where the distribution of returns is positively skewed).
4. Empirical Analysis
4.1. Correlation Analysis
As mentioned in a previous section, modern portfolio theory is based on the principle of diversification. Markowitz (1952) introduced the idea that covariances between assets, or the more intuitive correlation coefficients, should be used to form efficient portfolios. He claims that “[...] in trying to make variance small it is not enough to invest in many securities. It is necessary to avoid investing in securities with high covariances among themselves” (p. 89). Computing the cross-correlations between the return series is the most straightforward and intuitive method for investigating the linkages between markets and for assessing the potential for diversification in different markets. Moreover, due to their pairwise approach, crosscorrelations have the potential to offer detailed insights, which are specific to each individual stock market. Therefore, it has been common practice to evaluate the comovement of stock returns through the correlation coefficients.
However, it should be kept in mind that correlations may be biased due to heteroscedasticity and the results may be misleading without adjustments (Forbes and Rigobon, 2002). Nevertheless, they are used to obtain initial insights on potential comovements. Before we analyze the empirical findings for crossmarket correlations, a further concern may arise related to the nonsynchronicity across the 18 stock markets examined here. More specifically, the 5 CEE and 10 MENA frontier markets and the 3 developed markets are located in different time zones; hence there are gaps between the closing times of the stock exchanges. On the condition that the correlation coefficients are, in absolute value, bigger than zero, involving nonsynchronous data will cause the estimated correlation coefficients to be lower than their true values. Previous studies (Martens and Poon, 2001, inter alia) argued that returns series at weekly frequency are a better choice than those at daily frequency because the effect of closing prices in different time zones will be diluted, hence becoming smaller at lower frequency of the returns data.
We start by reporting the correlations between the CEE frontier markets and the three mature markets in Table 8.2.
Table 8.2
Correlation Coefficients Between the Weekly Returns of CEE Frontier and Mature Markets
| Germany | United Kingdom | United States | |
| Bulgaria | 0.527 | 0.503 | 0.455 |
| Croatia | 0.559 | 0.531 | 0.449 |
| Estonia | 0.526 | 0.505 | 0.412 |
| Romania | 0.621 | 0.563 | 0.529 |
| Slovenia | 0.577 | 0.571 | 0.459 |
Notes: The values reported are Pearson correlation coefficients between equity return series of frontier and mature markets for weekly frequencies. All sample correlation coefficients are significant at the 5% level.
The correlations between the CEE frontier markets and the mature markets, ranging from 0.412 (between Estonia and the United States) to 0.621 (between Romania and Germany), are below the correlations between the mature markets, which range from 0.825 (between Germany and the United States) to 0.889 (between Germany and the United Kingdom). The highest correlation with the German market is shown for Romania, which is also the market most correlated with the United States, whereas Slovenia is the market with the highest correlation with the United Kingdom. The lowest correlation with Germany is exhibited by Estonia, which is also the least correlated market with the United States, while the lowest correlation with the United Kingdom belongs to Bulgaria. On average, the CEE markets have the highest correlation with Germany, which is not surprising, given that Germany is one of their main trading partners, followed by the United Kingdom and the United States.
Crosscorrelations between the MENA frontier stock markets and the developed markets of Germany, the United Kingdom, and the United States are reported in Table 8.3.
Table 8.3
Correlations Between MENA Frontier Markets and Mature Markets
| Germany | United Kingdom | United States | |
| Bahrain | 0.026 | 0.019 | 0.067 |
| Jordan | 0.135 | 0.127 | 0.101 |
| Kuwait | 0.199 | 0.223 | 0.178 |
| Lebanon | 0.223 | 0.219 | 0.217 |
| Mauritius | 0.322 | 0.298 | 0.283 |
| Oman | 0.207 | 0.254 | 0.185 |
| Qatar | 0.231 | 0.225 | 0.233 |
| Saudi Arabia | 0.180 | 0.176 | 0.169 |
| Tunisia | 0.272 | 0.255 | 0.189 |
| United Arab Emirates | 0.201 | 0.232 | 0.219 |
Notes: The values reported are Pearson correlation coefficients between equity return series of frontier and mature markets for weekly frequencies. With the exception of those between Bahrain and each of the mature markets, all sample correlations are significant different from zero at the 5% level.
The correlations between the MENA frontier markets and the mature markets, ranging from 0.026 (between Bahrain and the United Kingdom) to 0.322 (between Mauritius and Germany), are much smaller than the correlations between the mature markets, which range from 0.825 (between Germany and the United States) to 0.889 (between Germany and the United Kingdom). The values displayed in Table 8.3 indicate that the market that is the least correlated with Germany is Bahrain, followed by Jordan, while the market that has the highest association in the sample with the German market is Mauritius albeit with a modest correlation coefficient. Bahrain continues to be the least correlated market with both the United Kingdom and the United States, followed by Jordan. The highest correlations with the United Kingdom and United States are exhibited by Mauritius. On average, the correlations of the MENA frontier markets with each of the three mature markets are characterized by modest values, which are much lower than the corresponding correlations for the CEE frontier economies. The GCC economies within the MENA group are major oil producers in the world and it would be reasonable to expect that they have a stronger association with the United States, which is also a major importer and producer of crude oil. However, according to the results in Table 8.3, the average correlation coefficient between the frontier markets in MENA and the US market (0.184) is the lowest among the three developed markets, a finding that is similar to that of previous studies (Hammoudeh and Choi, 2006, inter alia). Overall, the frontier markets in the CEE group display higher correlations, indicative of stronger associations with the mature markets, than the other 10 frontier markets which belong to the MENA group, which suggests that the CEE frontier markets offer fewer diversification opportunities than their MENA counterparts. The correlation analysis carried out here suggests that the developed markets are more integrated between them, but are still, to a great extent, segmented with the MENA frontier markets and, to a lesser extent, with the CEE frontier markets in the short term, which indicates that there may be some diversification benefits from investment in these frontier markets, especially the MENA ones. In what follows, we investigate further, using cointegration techniques, whether the short-term dependences suggested by the correlation analysis are appropriate indicators for international diversification benefits in long-term investment in the MENA and CEE frontier equity markets.
4.2. Cointegration Analysis
While crossmarket correlations are easy to compute and intuitively appealing, their computation is only the first step in a rigorous assessment of linkages across financial markets.k To gain more insights on the linkages across frontier and developed stock markets, we carry out a cointegration analysis, which enables us to investigate whether the markets under scrutiny share common trends in the long run, and to further assess the potential for diversification across markets. We start by assessing the stationarity properties of the series since a prerequisite for cointegration testing is that the series are integrated of order one. To this end, we apply two unit root tests, the augmented Dickey–Fuller (ADF) test (Dickey and Fuller, 1979, 1981) and the Phillips–Perron (PP) test (Phillips, 1987; Phillips and Perron, 1988), along with the Kwiatkowski–Phillips–Schmid–Shin (KPSS) (Kwiatkowski et al., 1992) test for stationarity, to both the index and the return series. While the ADF and PP tests assess the null hypothesis of a unit root against the alternative of stationarity, the KPSS test examines the null of stationarity against the alternative of a unit root, hence enabling us to conduct a confirmatory analysis of the ADF and PP test findings. The results of the ADF, PP, and KPSS tests are reported in Tables 8.4 and 8.5.
Table 8.4
Results of Unit Root and Stationarity Tests for the Index Series
| ADF | PP | KPSS | ||||
| Intercept | Trend and intercept | Intercept | Trend and intercept | Intercept | Trend and intercept | |
| Bahrain | −0.286 | −1.846 | −0.372 | −2.010 | 2.658*** | 0.239*** |
| Bulgaria | −1.055 | −1.749 | −1.039 | −1.691 | 1.975*** | 0.309*** |
| Croatia | −1.34 | −3.379* | −1.625 | −3.413* | 1.303*** | 0.164** |
| Estonia | −1.538 | −1.546 | −1.708 | −1.741 | 0.579** | 0.311*** |
| Jordan | −1.852 | −2.584 | −1.845 | −2.766 | 2.669*** | 0.236*** |
| Kuwait | −1.181 | −1.815 | −1.360 | −2.061 | 1.681*** | 0.176** |
| Lebanon | −2.245 | −3.002 | −2.228 | −3.018 | 1.387*** | 0.175** |
| Mauritius | −2.546 | −1.904 | −2.536 | −2.058 | 1.165*** | 0.212** |
| Oman | −1.672 | −3.241 | −2.117 | −2.372 | 0.813*** | 0.436*** |
| Qatar | −1.836 | −3.058 | −1.999 | −3.106 | 1.064*** | 0.239*** |
| Romania | −1.710 | −1.911 | −1.729 | −1.896 | 0.987*** | 0.348*** |
| Saudi Arabia | −2.688* | −2.455 | −2.743* | −2.510 | 0.509** | 0.481*** |
| Slovenia | −0.868 | −2.326 | −1.129 | −2.443 | 1.647*** | 0.198** |
| Tunisia | −3.365** | −3.003 | −3.358** | −2.981 | 0.881*** | 0.542*** |
| United Arab Emirates | −2.149 | −1.529 | −2.196 | −1.654 | 0.799*** | 0.574*** |
| Germany | −1.979 | −2.365 | −1.975 | −2.369 | 1.129*** | 0.255*** |
| United Kingdom | −1.903 | −1.885 | −1.789 | −1.769 | 0.393* | 0.374*** |
| United States | −0.555 | −1.379 | −0.501 | −1.322 | 1.284*** | 0.533*** |
Notes: *, **, and *** denote rejection of the null hypothesis at the 10, 5, and 1% levels of significance, respectively. The critical values for the ADF and PP tests are −3.443 (1%), −2.867 (5%), and −2.569 (10%) for the maintained regression with intercept, and −3.977 (1%), −3.419 (5%), and −3.132 (10%) for the maintained regression with intercept and linear trend. The critical values for the KPSS test are 0.739 (1%), 0.463 (5%), and 0.347 (10%) for the maintained regression with intercept, and 0.216 (1%), 0.146 (5%), and 0.119 (10%) for the maintained regression with intercept and linear trend. The Schwartz information criterion was employed to choose the optimal lag length in the ADF test; for the same purpose a Newey–West bandwidth (using Bartlett kernel) was used.
Table 8.5
Results of Unit Root and Stationarity Tests for the Returns Series
| ADF test | Phillips–Perron test | |||
| Intercept | Trend and intercept | Intercept | Trend and intercept | |
| Bahrain | −12.974*** | −12.966*** | −21.132*** | −21.116*** |
| Bulgaria | −8.987*** | −8.977*** | −20.97*** | −20.956*** |
| Croatia | −20.952*** | −21.088*** | −21.314*** | −21.334*** |
| Estonia | −20.966*** | −20.947*** | −21.081*** | −21.064*** |
| Jordan | −24.21*** | −24.225*** | −24.136*** | −24.146*** |
| Kuwait | −21.747*** | −21.727*** | −21.875*** | −21.857*** |
| Lebanon | −10.374*** | −10.356*** | −19.637*** | −19.643*** |
| Mauritius | −19.317*** | −19.427*** | −19.828*** | −19.795*** |
| Oman | −23.022*** | −22.998*** | −23.584*** | −23.566*** |
| Qatar | −20.792*** | −20.841*** | −20.807*** | −20.852*** |
| Romania | −12.593*** | −12.579*** | −20.4643*** | −20.447*** |
| Saudi Arabia | −22.018*** | −22.159*** | −22.074*** | −22.187*** |
| Slovenia | −22.497*** | −22.541*** | −23.06*** | −23.061*** |
| Tunisia | −22.747*** | −22.822*** | −22.748*** | −22.819*** |
| United Arab Emirates | −20.734*** | −20.931*** | −21.007*** | −21.063*** |
| Germany | −23.254*** | −23.23*** | −23.252*** | −23.228*** |
| United Kingdom | −24.773*** | −24.752*** | −24.756*** | −24.736*** |
| United States | −23.706*** | −23.746*** | −23.679*** | −23.743*** |
Notes: *, **, and *** denote rejection of the null hypothesis at the 10, 5, and 1% levels of significance, respectively. The critical values for the ADF and PP tests are −3.443 (1%), −2.867 (5%), and −2.569 (10%) for the maintained regression with intercept, and −3.977 (1%), −3.419 (5%), and −3.132 (10%) for the maintained regression with intercept and linear trend. The critical values for the KPSS test are 0.739 (1%), 0.463 (5%), and 0.347 (10%) for the maintained regression with intercept, and 0.216 (1%), 0.146 (5%), and 0.119 (10%) for the maintained regression with intercept and linear trend. The Schwartz information criterion was employed to choose the optimal lag length in the ADF test; for the same purpose a Newey–West bandwidth (using Bartlett kernel) was used.
We conduct each test for two specifications (intercept without deterministic trend included, and intercept and deterministic trend included) of the auxiliary regression of the test, as visual inspection may not be enough to establish the correct exogenous components in the maintained regression of the test in all cases. The results reported in Table 8.4 suggest that the null of a unit root cannot be rejected by both ADF and PP tests for any of the 18 index series at the 5% level of significance for a test conducted with an auxiliary regression including an intercept and a trend. This suggests that the index series are unit root processes at the 5% level of significance, a finding confirmed by the KPSS test. For the return series, the results of the ADF and PP tests indicate stationarity, which, together with the results reported in Table 8.4, suggests that the indices of the 18 stock markets considered in this study are integrated of order one, I(1), at the 5% level of significance.
4.2.1. The Johansen Cointegration Test
The results of the battery of unit root and stationarity tests conducted indicate that the prerequisites for examining for the presence of cointegration among the markets in the sample are satisfied. Cointegration occurs when nonstationary variables (equity market indices in this study) have one or more linear combinations that are stationary, which implies a long-run equilibrium relationship among variables. To assess the potential for cointegration, the multivariate cointegration test put forward by Johansen and Juselius (1990) and Johansen (1991) is applied to several groups of countries (markets). The existence of long-run relationships is assessed first within the CEE and the MENA frontier market groups. Subsequently, we investigate cointegration among each group of frontier markets and each mature market in turn.
Johansen (1991) proposes two different test statistics that can be used to test the hypothesis of the existence of r cointegrating vectors. The first such test statistic is the trace statistic, which tests the null hypothesis that the number of distinct cointegrating relationships is less than or equal to r against the alternative hypothesis of more than r cointegrating relationships, while the maximum eigenvalue test statistic tests the null hypothesis that the number of cointegrating relationships is less than or equal to r against the alternative of r + 1 cointegrating relationships.
The optimal lag length in the specification of the vector autoregressive (VAR) model underlying the cointegration test is selected based on several information criteria.l These suggest a lag of 1 for all eight groups considered for cointegration analysis. The specification of the cointegration test uses an intercept but no trend in the cointegrating equation and the underlying the VAR model.
The results of the Johansen (1991) multivariate cointegration test are presented in Tables 8.6–8.13, where both the trace and maximum eigenvalue statistics are reported.
Table 8.6
Johansen’s Test for Multivariate Cointegration Among CEE Frontier Markets
| H0 | H1 | Trace | 5% Critical value | Maximum eigenvalue | 5% Critical value |
| r = 0 | r > 0 | 82.18** | 69.82 | 46.67 | 33.88 |
| r ≤ 1 | r > 1 | 35.51 | 47.86 | 21.56 | 27.58 |
| r ≤ 2 | r > 2 | 13.96 | 29.79 | 7.784 | 21.13 |
| r ≤ 3 | r > 3 | 6.17 | 15.49 | 4.52 | 14.26 |
| r ≤ 4 | r > 4 | 1.65 | 3.84 | 1.65 | 3.84 |
Table 8.7
Johansen’s Test for Multivariate Cointegration Among CEE Frontier Markets and United States
| H0 | H1 | Trace | 5% Critical value | Maximum eigenvalue | 5% Critical value |
| r = 0 | r > 0 | 112.41** | 95.75 | 52.01 | 40.08 |
| r ≤ 1 | r > 1 | 60.39 | 69.82 | 26.53 | 33.88 |
| r ≤ 2 | r > 2 | 33.87 | 47.86 | 21.02 | 27.58 |
| r ≤ 3 | r > 3 | 12.85 | 29.79 | 7.69 | 21.13 |
| r ≤ 4 | r > 4 | 5.16 | 15.49 | 5.12 | 14.26 |
| r ≤ 5 | r > 5 | 0.04 | 3.84 | 0.04 | 3.84 |
Table 8.8
Johansen’s Test for Multivariate Cointegration Among CEE Frontier Markets and United Kingdom
| H0 | H1 | Trace | 5% Critical value | Maximum eigenvalue | 5% Critical value |
| r = 0 | r > 0 | 113.89** | 95.75 | 52.20 | 40.08 |
| r ≤ 1 | r > 1 | 61.68 | 69.82 | 26.32 | 33.88 |
| r ≤ 2 | r > 2 | 35.37 | 47.86 | 20.95 | 27.58 |
| r ≤ 3 | r > 3 | 14.41 | 29.79 | 9.01 | 21.13 |
| r ≤ 4 | r > 4 | 5.40 | 15.49 | 4.64 | 14.26 |
| r ≤ 5 | r > 5 | 0.76 | 3.84 | 0.76 | 3.84 |
Table 8.9
Johansen’s Test for Multivariate Cointegration Among CEE Frontier Markets and Germany
| H0 | H1 | Trace | 5% Critical value | Maximum eigenvalue | 5% Critical value |
| r = 0 | r > 0 | 100.33** | 95.75 | 48.41 | 40.08 |
| r ≤ 1 | r > 1 | 51.92 | 69.82 | 23.26 | 33.88 |
| r ≤ 2 | r > 2 | 28.66 | 47.86 | 16.00 | 27.58 |
| r ≤ 3 | r > 3 | 12.66 | 29.79 | 7.79 | 21.13 |
| r ≤ 4 | r > 4 | 4.87 | 15.49 | 4.55 | 14.26 |
| r ≤ 5 | r > 5 | 0.32 | 3.84 | 0.32 | 3.84 |
Table 8.10
Johansen’s Test for Multivariate Cointegration Among MENA Frontier Markets
| H0 | H1 | Trace | 5% Critical value | Maximum eigenvalue | 5% Critical value |
| r = 0 | r > 0 | 304.76** | 239.23 | 82.95** | 64.50 |
| r ≤ 1 | r > 1 | 221.81** | 197.37 | 54.71 | 58.43 |
| r ≤ 2 | r > 2 | 167.09** | 159.53 | 43.89 | 52.36 |
| r ≤ 3 | r > 3 | 123.21 | 125.61 | 33.34 | 46.23 |
| r ≤ 4 | r > 4 | 89.87 | 95.75 | 28.08 | 40.08 |
| r ≤ 5 | r > 5 | 61.78 | 69.82 | 26.09 | 33.88 |
| r ≤ 6 | r > 6 | 35.69 | 47.86 | 16.39 | 27.58 |
| r ≤ 7 | r > 7 | 19.30 | 29.79 | 10.36 | 21.13 |
| r ≤ 8 | r > 8 | 8.94 | 15.49 | 7.01 | 14.26 |
| r ≤ 9 | r > 9 | 1.93 | 3.84 | 1.93 | 3.84 |
Table 8.11
Johansen’s Test for Multivariate Cointegration Among MENA Frontier Markets and United States
| H0 | H1 | Trace | 5% Critical value | Maximum eigenvalue | 5% Critical value |
| r = 0 | r > 0 | 352.97** | 285.14 | 83.66** | 70.53 |
| r ≤ 1 | r > 1 | 269.31** | 239.23 | 62.02 | 64.50 |
| r ≤ 2 | r > 2 | 207.28** | 197.37 | 44.89 | 58.43 |
| r ≤ 3 | r > 3 | 162.39** | 159.53 | 37.30 | 52.36 |
| r ≤ 4 | r > 4 | 125.09 | 125.61 | 30.39 | 46.23 |
| r ≤ 5 | r > 5 | 94.69 | 95.75 | 26.22 | 40.08 |
| r ≤ 6 | r > 6 | 68.47 | 69.82 | 25.43 | 33.88 |
| r ≤ 7 | r > 7 | 43.04 | 47.86 | 16.52 | 27.58 |
| r ≤ 8 | r > 8 | 26.51 | 29.79 | 15.53 | 21.13 |
| r ≤ 9 | r > 9 | 10.98 | 15.49 | 8.31 | 14.26 |
| r ≤ 10 | r > 10 | 2.67 | 3.841 | 2.67 | 3.84 |
Table 8.12
Johansen’s Test for Multivariate Cointegration Among MENA Frontier Markets and United Kingdom
| H0 | H1 | Trace | 5% Critical value | Maximum eigenvalue | 5% Critical value |
| r = 0 | r > 0 | 366.91** | 285.14 | 86.54* | 70.53 |
| r ≤ 1 | r > 1 | 280.36** | 239.23 | 56.67 | 64.50 |
| r ≤ 2 | r > 2 | 223.69** | 197.37 | 55.21 | 58.43 |
| r ≤ 3 | r > 3 | 168.48** | 159.53 | 37.33 | 52.36 |
| r ≤ 4 | r > 4 | 131.15 | 125.61 | 32.79 | 46.23 |
| r ≤ 5 | r > 5 | 98.36 | 95.75 | 27.53 | 40.08 |
| r ≤ 6 | r > 6 | 70.82 | 69.82 | 26.66 | 33.88 |
| r ≤ 7 | r > 7 | 44.17 | 47.86 | 18.82 | 27.58 |
| r ≤ 8 | r > 8 | 25.35 | 29.79 | 15.29 | 21.13 |
| r ≤ 9 | r > 9 | 10.06 | 15.49 | 8.23 | 14.26 |
| r ≤ 10 | r > 10 | 1.83 | 3.84 | 1.83 | 3.84 |
Table 8.13
Johansen’s Test for Multivariate Cointegration Among MENA Frontier Markets and Germany
| H0 | H1 | Trace | 5% Critical value | Maximum eigenvalue | 5% Critical value |
| r = 0 | r > 0 | 370.72** | 285.14 | 85.29 | 70.53 |
| r ≤ 1 | r > 1 | 285.44** | 239.23 | 62.05 | 64.50 |
| r ≤ 2 | r > 2 | 223.39** | 197.37 | 54.92 | 58.43 |
| r ≤ 3 | r > 3 | 168.47** | 159.53 | 38.13 | 52.36 |
| r ≤ 4 | r > 4 | 130.34** | 125.61 | 34.85 | 46.23 |
| r ≤ 5 | r > 5 | 95.49 | 95.75 | 30.84 | 40.08 |
| r ≤ 6 | r > 6 | 64.64 | 69.82 | 26.00 | 33.88 |
| r ≤ 7 | r > 7 | 38.64 | 47.86 | 16.11 | 27.58 |
| r ≤ 8 | r > 8 | 22.53 | 29.79 | 12.65 | 21.13 |
| r ≤ 9 | r > 9 | 9.87 | 15.49 | 7.56 | 14.26 |
| r ≤ 10 | r > 10 | 2.31 | 3.84 | 2.31 | 3.84 |
The results reported in Tables 8.6–8.9 suggest the existence of one cointegrating relationship among the frontier markets from the CEE group, and also within the groups made up of the five CEE frontier markets and each of the mature markets, respectively.
The results reported in Tables 8.10–8.13 suggest that for the MENA group and the groups formed of the MENA frontier markets and each mature market, there is evidence of cointegration, but the trace and the maximum eigenvalue statistics do not support the same number of cointegrating relationships. For example, among the stock markets from the MENA group, the trace test statistic indicates three cointegrating vectors at the 5% level, while the maximum eigenvalue test suggests only one. Luutkepohl et al. (2001) argue that the trace test (which is generally preferred when bivariate cointegration is investigated) may suffer from severe size distortion. However, in a multivariate framework like the one employed here, the maximum eigenvalue test has more power. The results of this test indicate the presence of one cointegrating relationship among the MENA frontier markets and among MENA and Germany, MENA and the United Kingdom, and MENA and the United States. Therefore, it can be concluded that for all groups of frontier markets or frontier and developed markets considered here, the evidence points to the existence of one cointegrating relationship, as summarized in Table 8.14.
Table 8.14
Cointegration Ranks for the Eight Groups of Markets Examined
| Group | Composition | No. of cointegrating relationships |
| CEE frontier markets | Bulgaria, Croatia, Estonia, Romania, Slovenia | 1 |
| CEE frontier markets and US market | Bulgaria, Croatia, Estonia, Romania, Slovenia, and United States (S&P 500) | 1 |
| CEE frontier markets and UK market | Bulgaria, Croatia, Estonia, Romania, Slovenia, and United Kingdom (FTSE 100) | 1 |
| CEE frontier markets and German market | Bulgaria, Croatia, Estonia, Romania, Slovenia, and Germany (DAX 30) | 1 |
| MENA frontier markets | Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, United Arab Emirates, Jordan, Lebanon, Mauritius, Tunisia | 1 |
| MENA frontier markets and US market | Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, United Arab Emirates, Jordan, Lebanon, Mauritius, Tunisia, and United States (S&P 500) | 1 |
| MENA frontier markets and UK market | Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, United Arab Emirates, Jordan, Lebanon, Mauritius, Tunisia, and United Kingdom (FTSE 100) | 1 |
| MENA frontier markets and Germany market | Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, United Arab Emirates, Jordan, Lebanon, Mauritius, Tunisia, and Germany (DAX 30) | 1 |
4.3. Granger Causality Tests
Since the existence of cointegration has been documented for the eight groups of markets considered, they will be characterized by a tendency to move together in the long run even though they may experience short-run deviations from the common equilibrium path. Both the long-run comovements and the short-run interdependences among markets can be examined within the confines of vector error correction (VEC) models. By specifying a vector error correction system, attention is placed on the lagged value(s) of the first differences of the index series (ie, the returns) for the purpose of investigating the short-run interaction among variables, and on the speed of adjustment to equilibrium, which is represented by the error correction term (ECT). To start with, we investigate the long-run relationship between the five CEE frontier markets and each mature market. The results from the VEC model estimated for the group CEE–US suggest that although the ECTs of Bulgaria, Croatia, Romania, and Slovenia are statistically significant and the absolute value are located in the range between 0 and 1, their signs are positive instead of negative (Table 8.15); the only negative ECT coefficient is recorded in the case of Estonia, but it is statistically insignificant. More promising results are found for the groups CEE–UK and CEE–Germany, for which the ECTs are significant at the 5% level and have the desired negative sign, suggesting that the short-run deviations from equilibrium decrease over time; when temporary deviations away from the long-run equilibrium occur, the CEE frontier stock markets react correspondingly to remove the driving factors and restore the equilibrium.
Table 8.15
Results of Granger Causality Tests for the Group CEE–US
| Market X “Granger causes” market Y | Y | ||||||
| X | Bulgaria | Croatia | Estonia | Romania | Slovenia | United States | ECTt − 1 |
| Bulgaria | — | 0.448 | 0.027* | 0.046* | 0.112 | 0.033* | 0.006* |
| Croatia | 0.571 | — | 0.137 | 0.223 | 0.381 | 0.719 | 0.005* |
| Estonia | 0.638 | 0.849 | — | 0.371 | 0.381 | 0.011* | −0.002 |
| Romania | 0.022* | 0.693 | 0.610 | — | 0.383 | 0.001* | 0.005* |
| Slovenia | 0.454 | 0.002* | 0.797 | 0.025* | — | 0.008* | 0.004* |
| United States | 0.001* | 0.043* | 0.014* | 0.124 | 0.080 | — | 0.001 |
Notes: The last column reports the statistical significance of the (lagged) error correction term in the VEC model. * indicates significance at the 5% level.
The Granger causality test, which is a test of predictive power indicating whether lagged changes in the values of the index in market X have predictive power for changes in the index in market Y, can shed more light on the short-run interdependencies among the markets from each group considered. The results for the CEE frontier markets (reported in Tables 8.15–8.17) suggest that the US stock market Granger-causes the Bulgarian, Croatian, and Estonian markets, and the United Kingdom is found to Granger-cause only the Bulgarian stock market, while the German returns have no predictive power for any of the CEE market returns.
Table 8.16
Results of Granger Causality Tests for the Group CEE–UK
| Market X “Granger causes” market Y | Y | ||||||
| X | Bulgaria | Croatia | Estonia | Romania | Slovenia | United Kingdom | ECTt – 1 |
| Bulgaria | — | 0.563 | 0.059 | 0.074 | 0.136 | 0.052 | −0.001* |
| Croatia | 0.690 | — | 0.124 | 0.297 | 0.399 | 0.991 | −0.001* |
| Estonia | 0.824 | 0.824 | — | 0.246 | 0.341 | 0.008* | −0.0001 |
| Romania | 0.030* | 0.800 | 0.612 | — | 0.524 | 0.082 | −0.001* |
| Slovenia | 0.661 | −0.003* | 0.553 | 0.013* | — | 0.141 | −0.0007* |
| United Kingdom | 0.011* | 0.219 | 0.729 | 0.372 | 0.560 | — | 0.00002 |
Notes: The last column reports the statistical significance of the (lagged) error correction term in the VEC model. * indicates significance at the 5% level.
Table 8.17
Results of Granger Causality Tests for the Group CEE–Germany
| Market X “Granger causes” market Y | Y | ||||||
| X | Bulgaria | Croatia | Estonia | Romania | Slovenia | Germany | ECTt – 1 |
| Bulgaria | — | 0.563 | 0.059 | 0.074 | 0.136 | 0.072 | −0.0013* |
| Croatia | 0.690 | — | 0.124 | 0.297 | 0.399 | 0.837 | −0.0011* |
| Estonia | 0.824 | 0.824 | — | 0.246 | 0.341 | 0.071 | −0.0001 |
| Romania | 0.030* | 0.800 | 0.612 | — | 0.524 | 0.112 | −0.0011* |
| Slovenia | 0.661 | −0.003* | 0.553 | 0.013* | — | 0.023* | −0.0008* |
| Germany | 0.545 | 0.740 | 0.793 | 0.074 | 0.589 | — | −0.0001 |
Notes: The last column reports the statistical significance of the (lagged) error correction term in the VEC model. * indicates significance at the 5% level.
The results pertaining to the sign and significance of the ECT in the adjustment to equilibrium in the VEC model specification and the Granger causality tests for the MENA frontier markets are provided in Tables 8.18–8.20.
Table 8.18
Results of Granger Causality Tests for the Group MENA–US
| Market X “Granger causes” market Y | Y | |||||||||||
| X | Bahrain | Kuwait | Oman | Qatar | Saudi Arabia | United Arab Emirates | Jordan | Lebanon | Mauritius | Tunisia | United States | ECTt – 1 |
| Bahrain | — | 0.5239 | 0.2788 | 0.6659 | 0.9072 | 0.3252 | 0.1318 | 0.1177 | 0.0036* | 0.8833 | 0.5369 | −0.0112* |
| Kuwait | 0.0067* | — | 0.5322 | 0.1192 | 0.3340 | 0.8198 | 0.4606 | 0.9691 | 0.2266 | 0.9719 | 0.0012* | −0.0128* |
| Oman | 0.2415 | 0.8585 | — | 0.1637 | 0.4688 | 0.0125* | 0.7629 | 0.5823 | 0.6408 | 0.6253 | 0.0358* | 0.0060 |
| Qatar | 0.1567 | 0.3680 | 0.7846 | — | 0.8331 | 0.9206 | 0.1666 | 0.8835 | 0.3715 | 0.8030 | 0.1156 | 0.0159* |
| Saudi Arabia | 0.5867 | 0.5762 | 0.7231 | 0.3207 | — | 0.4237 | 0.2227 | 0.6645 | 0.0749 | 0.0572 | 0.0079* | −0.0008 |
| United Arab Emirates | 0.0349* | 0.2655 | 0.2170 | 0.6611 | 0.3649 | — | 0.9516 | 0.3022 | 0.868 | 0.1320 | 0.4414 | 0.0076 |
| Jordan | 0.0205* | 0.0072* | 0.0150* | 0.5954 | 0.3319 | 0.4600 | — | 0.0006* | 0.0283* | 0.3942 | 0.0753 | 0.0076 |
| Lebanon | 0.5360 | 0.6154 | 0.0300* | 0.2675 | 0.2348 | 0.6661 | 0.0005 | — | 0.3541 | 0.5217 | 0.2916 | −0.0342* |
| Mauritius | 0.1405 | 0.3941 | 0.0294* | 0.1996 | 0.4292 | 0.1321 | 0.4005 | 0.1645 | — | 0.3320 | 0.4621 | 0.0038 |
| Tunisia | 0.9268 | 0.5060 | 0.1254 | 0.0218* | 0.6069 | 0.7805 | 0.1437 | 0.6805 | 0.2943 | — | 0.2301 | −0.00106 |
| United States | 0.027* | 0.0303* | 0.0000* | 0.0001* | 0.0000* | 0.0000* | 0.0072* | 0.5704 | 0.0003* | 0.3763 | — | 0.00235 |
Note: The last column reports the statistical significance of the (lagged) error correction term in the VEC model. * indicates significance at the 5% level.
Table 8.19
Results of Granger Causality Tests for the Group MENA–UK
| Market X “Granger causes” market Y | Y | |||||||||||
| X | Bahrain | Kuwait | Oman | Qatar | Saudi Arabia | United Arab Emirates | Jordan | Lebanon | Mauritius | Tunisia | United Kingdom | ECTt – 1 |
| Bahrain | — | 0.5175 | 0.1779 | 0.4829 | 0.8841 | 0.2238 | 0.0749 | 0.1589 | 0.0023* | 0.9296 | 0.1747 | −0.013* |
| Kuwait | 0.0063* | — | 0.3942 | 0.2335 | 0.3552 | 0.7585 | 0.3825 | 0.6478 | 0.3109 | 0.9065 | 0.1445 | −0.013* |
| Oman | 0.4151 | 0.7701 | — | 0.1032 | 0.2354 | 0.0043* | 0.9334 | 0.5697 | 0.5078 | 0.6162 | 0.2769 | 0.0051 |
| Qatar | 0.1778 | 0.4364 | 0.9917 | — | 0.8691 | 0.6479 | 0.1042 | 0.9808 | 0.2745 | 0.8019 | 0.0099* | 0.015* |
| Saudi Arabia | 0.5023 | 0.6582 | 0.7061 | 0.2775 | — | 0.4274 | 0.2096 | 0.4745 | 0.074* | 0.0586 | 0.0034* | −0.0108 |
| United Arab Emirates | 0.0339* | 0.2554 | 0.1862 | 0.7366 | 0.3154 | — | 0.9926 | 0.3437 | 0.9287 | 0.1305 | 0.1057 | −0.0013 |
| Jordan | 0.0084* | 0.003* | 0.015* | 0.6773 | 0.1415 | 0.2896 | — | 0.0003* | 0.031* | 0.4603 | 0.0558 | 0.0009 |
| Lebanon | 0.6268 | 0.7614 | 0.026* | 0.2508 | 0.2003 | 0.5886 | 0.0007* | — | 0.4224 | 0.5070 | 0.5762 | −0.028* |
| Mauritius | 0.1601 | 0.2546 | 0.026* | 0.2069 | 0.2722 | 0.0818 | 0.3949 | 0.1221 | — | 0.3620 | 0.8606 | 0.0048 |
| Tunisia | 0.9855 | 0.7054 | 0.2038 | 0.0561 | 0.6461 | 0.8814 | 0.2542 | 0.8806 | 0.2430 | — | 0.6247 | 0.0013 |
| United Kingdom | 0.0080* | 0.4328 | 0.0001* | 0.0003* | 0.0004* | 0.0004* | 0.004* | 0.6299 | 0.003* | 0.3648 | — | 0.011* |
Note: The last column reports the statistical significance of the (lagged) error correction term in the VEC model. * indicates significance at the 5% level.
Table 8.20
Results of Granger Causality Tests for the Group MENA–Germany
| Market X “Granger causes” market Y | Y | |||||||||||
| X | Bahrain | Kuwait | Oman | Qatar | Saudi Arabia | United Arab Emirates | Jordan | Lebanon | Mauritius | Tunisia | Germany | ECTt– 1 |
| Bahrain | — | 0.537 | 0.212 | 0.459 | 0.9058 | 0.205 | 0.086 | 0.1451 | 0.002* | 0.931 | 0.2784 | −0.009* |
| Kuwait | 0.005* | — | 0.436 | 0.224 | 0.3251 | 0.728 | 0.398 | 0.676 | 0.319 | 0.887 | 0.038* | −0.001* |
| Oman | 0.269 | 0.8318 | — | 0.121 | 0.3515 | 0.006* | 0.924 | 0.5375 | 0.551 | 0.656 | 0.2043 | 0.0046 |
| Qatar | 0.1514 | 0.3663 | 0.898 | — | 0.9381 | 0.809 | 0.131 | 0.9311 | 0.3498 | 0.759 | 0.0499 | 0.011* |
| Saudi Arabia | 0.5835 | 0.5871 | 0.681 | 0.276 | — | 0.375 | 0.194 | 0.5937 | 0.0651 | 0.058 | 0.013* | −0.0055 |
| United Arab Emirates | 0.0272* | 0.2310 | 0.149 | 0.829 | 0.2491 | — | 0.914 | 0.3174 | 0.9766 | 0.123 | 0.5011 | −0.0007 |
| Jordan | 0.0101* | 0.003* | 0.014* | 0.612 | 0.1754 | 0.282 | — | 0.002* | 0.0359 | 0.491 | 0.4132 | 0.0016 |
| Lebanon | 0.5090 | 0.6918 | 0.023* | 0.302 | 0.2240 | 0.696 | 0.001* | — | 0.3216 | 0.519 | 0.7794 | −0.024* |
| Mauritius | 0.1334 | 0.3401 | 0.033* | 0.313 | 0.4276 | 0.161 | 0.453 | 0.1511 | — | 0.415 | 0.7596 | 0.0027 |
| Tunisia | 0.9868 | 0.5221 | 0.201 | 0.073 | 0.8516 | 0.879 | 0.274 | 0.8044 | 0.1530 | — | 0.9789 | 0.0018 |
| Germany | 0.0755* | 0.1586 | 0.003* | 0.00* | 0.0000* | 0.000* | 0.005* | 0.8565 | 0.0001* | 0.279 | — | 0.009* |
Notes: The last column reports the statistical significance of the (lagged) error correction term in the VEC model. * indicates significance at the 5% level.
The ECTs of Bahrain, Qatar, and Lebanon are negative in sign and statistically significant for all three groups: MENA and US, MENA and UK, and MENA and Germany. As for the short-run interrelationships between the MENA frontier markets and the mature markets, the results of the Granger causality test suggest that equity returns in the US market have predictive power for equity returns in all but two (Lebanon and Tunisia) of the MENA markets considered in this study. The same result holds true for the UK market, while the German market Granger-causes 7 of the 10 MENA frontier markets (exceptions being Kuwait, Lebanon, and Tunisia).
5. Concluding Remarks
As emerging markets have become more integrated with mature markets, investors in search of diversification gains have started to devote increasing attention to the more recently denominated frontier markets. The research community has followed suit, and now the literature on linkages across financial markets is counting a growing number of studies that focus on frontier markets. In this study we attempt to contribute to this literature by examining the comovement of selected CEE group (Bulgaria, Croatia, Estonia, Romania, and Slovenia) and MENA group (Bahrain, Jordan, Kuwait, Lebanon, Mauritius, Oman, Qatar, Saudi Arabia, Tunisia, and United Arab Emirates) frontier stock markets with three developed markets (Germany, United Kingdom, and United States), and also the interdependence across markets within the CEE and MENA groups. The 15 frontier markets listed earlier have been selected as representative of frontier markets in general because they account for a substantive share of both the market capitalization and the GDP of the entire frontier market group. We hypothesize that, while the CEE and MENA regional equity markets may be globally segmented, hence offering potential diversification gains, at a regional level they exhibit, at least in the longer run, higher levels of integration due to economic reasons. Our empirical investigation opens with correlation analysis, which lends some support to this hypothesis. Moreover, the results from correlation analysis indicate that that the markets have stronger short-run links within the CEE region than in the MENA region. Therefore, the potential diversification gains are rather limited when investing within the CEE frontier market region. Furthermore, the findings suggest that the CEE frontier markets are more correlated with the mature markets considered in this study than are the MENA markets, which implies that greater diversification benefits can be reaped by combining MENA and developed markets in the portfolio.
To shed more light on long-run linkages across the markets considered, we complement the examination of correlations with cointegration analysis. The empirical results derived from the Johansen cointegration testing approach suggest the presence of a stable relationship linking the markets within each frontier market group in the long run. Moreover, evidence of cointegration supporting the presence of a significant long-run equilibrium relationship is unveiled for the groups of markets formed of each frontier market group and every mature market considered (namely, CEE and Germany, CEE and United Kingdom, CEE and United States, MENA and Germany, MENA and United Kingdom, MENA and United States). This suggests that the degree of segmentation of the CEE and MENA frontier markets with their developed counterparts has decreased over time, since our findings contrast with those of the earlier studies that documented greater segmentation. This result does not come as a surprise, especially in the case of the CEE frontier stock markets considered here, given that in the year before the starting year of our sample (2004) two of them (Estonia and Slovenia) joined the EU, with the other three following in 2007 (Bulgaria and Estonia) and 2013 (Croatia). In addition to long-run linkages, we also address possible short-run market interactions by conducting Granger causality tests in the confines of VEC model representations. The results from these tests show that both the US and UK markets have predictive power for the returns of the CEE frontier stock markets, whereas (surprisingly) Germany does not. As for the short-run linkages among the MENA region market and developed markets, no mature market is found to Granger-cause Lebanon, Tunisia, and Kuwait, which suggests that some risk-reducing opportunities may be exploited by investing in these markets.
The overall picture from the empirical analysis conducted here suggests that the frontier stock markets in the MENA region offer greater diversification potential for international investors than the CEE frontier markets. The latter display a greater degree of integration with the mature markets, which can be explained by their being more advanced than some of the MENA markets considered here and by the economic ties with Germany and the United Kingdom, who are also part of the EU. In this respect, the degree of integration of the CEE markets may have a permanent component, which may be the result of the overall trend toward global financial market integration, but also a transitory component which may be attributed to the contagion over the global financial crisis of 2008, covered by the sample considered here, which spread faster from the developed markets to the markets in the CEE region than to their MENA counterparts. Therefore, our results make the case for devoting more attention to the less developed frontier markets, which may make better candidates in the search for diversification gains. However, as a less developed market probably means less liquidity and higher transaction costs, a more accessible tool—frontier exchange-traded funds (ETFs)—can be exploited to investigate the financial linkage between the less developed frontier markets and mature markets for the purpose of portfolio diversification. The outcomes of the comovement analysis conducted in this study may provide investors with a road map for investment decisions. The focus on frontier markets provides new and important insights into the field of portfolio diversification across markets.
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.