Cointegration has important implications for portfolio diversification. One of these is that in order to spread risk it is advisable to invest in markets that are not cointegrated. Over the last several decades communication technology has made the world a smaller place and hence cointegration in equity markets has become more prevalent. The bulk of research into cointegration focuses on developed and Asian markets, with little research been done on African markets. This study compares the Engle-Granger and Johansen tests for cointegration and uses them to calculate the level of cointegration between South African and other global equity markets. Each market is compared pair-wise with South Africa and the results have been that in general South Africa is cointegrated with other emerging markets but not really with African nor developed markets. Short-run analysis with the error correction was carried out and showed that in general markets respond slowly to any disequilibrium. Innovation accounting methods showed that the country placed first in Cholesky ordering dominates the other one. Multivariate cointegration was carried out using three selections of 4, 6 and 8 market portfolios. One of the markets was SA and the others were all chosen based on the criteria that they are not pair-wise cointegrated with SA. The level of cointegration varied depending on the portfolios, as did the error correction rates, impulse responses and variance decomposition. The one constant was that the USA dominated any portfolio where it was introduced. Recommendations were finally made about which market portfolio an investor should consider as most favourable.

Research studies on portfolio diversification have tended to focus on developed markets and paid less attention to emerging markets. Traditionally, correlation analysis has been used to determine potential benefits from diversification but current studies have shifted focus from correlation analysis to exploring cointegration analysis and other forms of tests such as the Vector Error Correction Methodology. The research seeks to find if it is beneficial for a South African investor to diversify their portfolio of emerging market equities over a long-term period. Daily weighted share indices for the period of January 1996 to November 2008 were collected and analysed through the application of the Johansen cointegration technique and Vector Error Correction Methodology. Granger Causality tests were also performed to established whether one variable can be useful in forecasting another variable. The study found that there was at least one statistically significant long-run relationship between the emerging markets. After testing for unit roots for all the share indices and their first difference using the Augmented Dickey-Fuller test (ADF), Philips-Perron and Kwiatkowski, Phillips, Schmidt, and Shin (KPSS) unit root tests, similar conclusions were m~de. All the unit root tests and their levels could not be rejected for all the series. However, unit root tests on the first differences were rejected, meaning that all series are of order 1(1) - evidence of cointegration. Simply put, emerging markets tend not to drift apart over time. This suggests that emerging markets offer limited benefits to investors who are looking to add some risk to their portfolios. In addition, the study also found evidence of both unidirectional and bidirectional causality (Granger-Cause tests) between markets. This implies that the conditions for a particular market are exogenous of the other market. The study concludes that emerging markets are gradually adopting the same profile as developed markets.