PORTFOLIO DIVERSIFICATION WITH THE USE OF RISK PROFILE CLUSTERING: EVIDENCE FROM THE JAPAN NIKKEI 225 INDEX.

Abstract

Conventional theory suggests that to reap the benefits of portfolio diversification, one might need to hold as many as 40 securities in his portfolio (Chance et al., 2011). However, this involves sizeable transaction costs which would be prohibitively expensive for a retail investor. In addition, classical portfolio allocation only captures the first two moments of a return distribution which increases the portfolio risk especially during market downturns. This paper analyses how the use of clustering algorithms, together with risk measures geared towards capturing tail dependence between assets, can result in a smaller but still well-diversified portfolio. Specifically, we focus on the Japanese stock market, with NIKKEI 225 as the benchmark. Results show that with just five or fewer stocks, it is possible to outperform the market benchmark regardless of whether this strategy was adopted before, during, or after the 2008 Global Financial Crisis.