International portfolio diversification: A cluster analysis approach

One of the critical aspects of modem portfolio theory is to deal with the merits of international diversification, which is closely linked with the co-movements and interdependence between stock markets. This paper aims to investigate the potential of diversifying into U.S. and some Asia Pacific markets by studying the co-movements of the monthly returns of the markets between July 1997 and Dec 2004. The markets considered are: Australia, China, Hong Kong, India, Japan, Malaysia, Singapore, South Korea, Taiwan, Thailand and the U.S. Cluster analysis is first used to classify the markets into groups such that markets within the same group move quite closely in tandem with each other while movements of markets from different groups are not quite closely related. The correlation coefficients between the monthly returns of the markets are used as the measures of similarity between the markets. Then from each group the market with the highest monthly return per unit risk is selected to form a portfolio. An optimum investment strategy is then derived using the Markowitz algorithm, i.e. weight of investment in each market in the portfolio is determined so that the risk is minimized for a given rate of return. We will also look at the stability of the cluster structure by applying the same technique over three sub-periods.