Financial integration and portfolio diversification: Evidence from CIVETS stock markets

This paper investigates the extent of financial integration among a new group of six frontier markets called “CIVETS” by utilizing the multivariate GARCH framework of Engle and Kroner [1]. These countries are expected to show sustainable growth in productivity and domestic consumption over the next decade and are considered as potential corridor for the international investor from portfolio diversification point of view. We utilize weekly stock market return series of all the CIVIETS nations, and results exhibit significant return and volatility spillovers among all the markets under investigation. Our results reveal that there are significant linkages among CIVETS stock markets during the time of our analysis. However, the direction of relationship is asymmetric depending on the countries in the model. We believe, CIVIETS stock markets have full potential of being the future investment targets worldwide.


Introduction
In 2001, Jim O'Neill of Goldman Sachs coined the term BRIC in his report "Building Better Global Economic BRICs". Since then an emerging market boom has been witnessed and in particular, this block of four largest developing economies of the world has attracted most of the annual investment flows. However, the early mover advantage for investing in BRICs-Brazil, Russia, India and China which was present since the introduction of these emerging markets block, seems to be mitigating over time.
ing of six countries namely Colombia, Indonesia, Vietnam, Egypt, Turkey and South Africa. These countries not only provide diversity in terms of geographical location but also share some common aspects such as political stability (especially when compared to previous generations), young populations that focus on education and overall growing economic trends. We believe that despite of economic and political turmoil that may have increased the country level risks for investments all over the world, the diversification opportunities that this block of new frontier markets provides have not declined. These countries are expected to show sustainable growth in productivity and domestic consumption over the next decade. Moreover the CIVETS are not heavily dependent on external demand as compared to BRICs. Therefore, studying the financial integration and portfolio diversification among CIVETS is an essential step forward towards future investment opportunities.
Globally, the CIVETS have become a new arena for financial linkages and stock market integration academic research, besides being a topic for global business discussion. There has been widespread stream of literature on the financial markets of individual CIVETS countries along with emerging markets of different regions of the world 1 , however, the research on inter-linkages of these markets in cross-market settings among CIVETS is almost non-existent. Recently, Korkmaz et al. [15], using the Granger-Cheung-Ng-Hong causality tests for mean and variance of weekly stock market returns, have studied the return and volatility spillovers among CIVETS stock markets and found generally lower contemporaneous spillover effects among countries. They utilized and saw rather weak evidence of inter-or intra-regional interdependencies effects. Only 10 of the possible 30 country pair-wise directional casual relationships were found to be statistically significant. There is no further study found on the intra-market linkages of these six economies 2 .
We use the multivariate GARCH framework of Engle and Kroner [1] of time-varying volatility to determine the intra-market linkages of return and volatility among CIVETS markets. Both return and volatility linkages are tested in bivariate setting where bi-directional relationship of return and time-varying volatility is analyzed. The diversity in the geographical location and other trade and economic factors among CIVETS countries implies no relationship among each other because there is no significant commonalities and trade among these countries. However, since the introduction of the acronym "CIVETS" has been widely used in academic and investment circles around the world, the interest towards this block of frontier economies may have induced linkages among the stock markets due to generally high level of portfolio investments in these countries. Hence, we argue that our contribution is primarily empirical in nature and to provide the first hand evidence on the economic relationship among six emerging markets which have full potential of being the future investment targets worldwide. Our results reveal that there are significant linkages among CIVETS stock markets 1 See for example, Karim  during the time period of our analysis. However, the direction of relationship is asymmetric depending on the countries in the model. The rest of the paper is organized as follows: next section describes the specifications of model used in the analysis with a review of earlier studies. Section 3 provides a detailed outlook of the data and its descriptive characteristics. In Section 4, we present and discuss the results of the empirical analysis. Section 5 outlines the diagnostic tests to verify the results and finally Section 6 concludes the paper.

Model Specification
The Autogressive Conditional Heteroscedasticity (ARCH) process proposed by Engle [16] and the generalized ARCH (GARCH) by Bollerslev [17] are well known in volatility modelling of stock returns. In examining volatility linkages between countries, however, a multivariate GARCH approach is preferred over univariate settings.
We start our empirical specification with a bivariate VAR-GARCH (1, 1) model that accommodates each market's returns and the returns of other markets lagged one pe- where r t is an n × 1 vector of weekly returns at time t for each market. The n × 1 vector of random errors µ t represents the innovation for each market at time t with its corresponding n × n conditional variance-covariance matrix H t . The market information available at time t − 1 is represented by the information set Ω t−1 . The n × 1 vector, α, represents the constant. The own market mean spillovers and cross-market mean spillovers are measured by the estimates of matrix β elements, the parameters of the vector autoregressive term. This multivariate structure thus facilitates the measurement of the effects of innovations in the mean stock returns of one series on its own lagged returns and those of the lagged returns of other markets.
Given the above expression, and following Engle and Kroner [1], the conditional covariance matrix can be stated as: where the parameter matrices for the variance equation are defined as w 0 , which is restricted to be lower triangular and two unrestricted matrices γ 11 and δ 11 . Thus, the second moment can be represented by: where θ denotes the vector of all the unknown parameters. Numerical maximisation of Equation (4) yields the maximum likelihood estimates with asymptotic standard errors.
Finally, to test the null hypothesis that the model is correctly specified, or equivalently, that the noise terms, µt, are random, the Ljung-Box Q-statistic is used. It is assumed to be asymptotically distributed as χ 2 with (p − k) degrees of freedom, where k is the number of explanatory variables.

Data and Descriptive Statistics
The data comprise weekly price indices for the countries under investigation. More  Figure 1 clearly exhibits non-stationarity.
The first two moments of the data, i.e., mean and standard deviation, are multiplied by 52 and the square root of 52 to show them in annual terms. As one would anticipate, most of the markets offer high returns, Colombia and Egypt seems to be the most favorite investments, offering 20% and 19% returns per annum respectively, Indonesia, Turkey and South Africa offer 18%, 14% and 13% respectively while Vietnam found as the least interesting market with an annualized market return of just 0.22%. However, the high returns are associated with high risk (standard deviations) as well. All the markets under investigation found to be highly risky ranging from South Africa (lowest) at 20% to Turkey (highest) at 33% standard deviation. All the return series are, without exception, highly leptokurtic and exhibit strong negative skewness. This suggests the presence of asymmetric trends towards negative values. To check the null hypothesis of normal distribution, we calculated the Jarque-Bera test statistic and reject the null of normality in all cases.  Since we are using a GARCH process to model variance in asset returns, we also test for the presence of the ARCH effect.

Empirical Results
Our empirical results answer the theoretical questions formulated in the previous sections. First, to examine the return and volatility transmission of CIVETS stock markets, fifteen (15) pair-wise models are estimated utilizing bivariate GARCH frame work, for which a BEKK representation is adopted [1].
The results obtained from bi-variate GARCH (1,1) with BEKK specifications [1] are summarized in Tables 2-4. We first look at matrix β in the mean equation, Equation (1), captured by the parameters β ij and β ji , in order to see the relationship in terms of returns across the countries and sectors in each pair. Parameters γ ij and γ ji , captures cross-market ARCH effects and parameters δ ij and δ ji measure own and cross-market GARCH effects. LB and LB2 presents the Ljung-Box Q-statistic for standardized and standardized squared residuals.
The results shown in Tables 2-4   The parameter β represents the return spillovers. The parameter matrices for the variance equation are defined as ω, which is restricted to be lower triangular and two unrestricted matrices γ, captures own and cross-market ARCH effects and δ measure own and cross-market GARCH effects. LB and LB2 presents the Ljung-Box Q-statistic for standardized and standardized squared residuals. (*) denotes the significance level at 10%, (**) denotes the significance level at 5%. The parameter β represents the return spillovers. The parameter matrices for the variance equation are defined as ω. which is restricted to be lower triangular and two unrestricted matrices γ. captures own and cross-market ARCH effects and δ measure own and cross-market GARCH effects. LB and LB2 presents the Ljung-Box Q-statistic for standardized and standardized squared residuals. (*) denotes the significance level at 10%. (**) denotes the significance level at 5%.

Diagnostic Tests
We also estimate the Ljung-Box Q-statistic used to test the null hypothesis that the model is correctly specified, or equivalently, that the noise terms are random. We calculate both standardized and standardized squared residuals up to lag 24 for each modelled pair. Results show (not reported) no series dependence in the squared standardized residuals, indicating the appropriateness of the GARCH-BEKK model.

Summary and Conclusions
In this paper we have examined the return and volatility spillovers among a new group of six frontier markets called "CIVETS". We analyze the inter-market linkages in a more advanced setting. We use a Generalized Autoregressive Conditional Heteroske- The parameter β represents the return spillovers. The parameter matrices for the variance equation are defined as ω. which is restricted to be lower triangular and two unrestricted matrices γ. captures own and cross-market ARCH effects and δ measure own and cross-market GARCH effects. LB and LB2 presents the Ljung-Box Q-statistic for stan- These findings suggest that the portfolio investors who invest in emerging and frontier markets for better returns should take into account the correlation of risk and returns among CIVETS stock markets. The diversification benefits should be assessed keeping in view the extent of inter-market linkages of Colombia, Indonesia, Vietnam, Egypt, Turkey and South Africa.