Portfolios Resampling and International Diversification: A Non-Parametric Stochastic Dominance Approach

Abstract

The aim of this paper is to consider instability and ambiguity problems on portfolio selection. We examine the impact of estimation errors on financial portfolios optimization processes. We investigate the controversy problem of international and domestic optimal diversification strategies choices using a non parametric stochastic dominance approach (NPSDA) based on Monte Carlo (MC) and Bootstrap p-values from an American point of view. Based on a data set consisting of daily closing prices of U.S. stocks and index and Asian and Latin American stock market indices for the period from August 03, 1993 to August 31, 2007, estimation errors visualisation results show that small changes in input parameters imply large changes in the optimized portfolio composition and a considerably modification of mean-variance efficient frontiers shape. Though mean-variance optimization could not be used to draw any preference between international and domestic diversification, empirical findings reveal the utility of the SD approach to define an optimal strategy. Further, results show that there exists substantial evidence of international global diversification benefits for domestic investors. More precisely, for risk levels higher than 30 percent, risk-adverse American investor having an increasing utility function prefers global international to domestic resampled diversification strategy. Besides, we find that domestic resampled diversification strategy beats all international major and emerging markets diversification strategies. For risk levels lower than 30 percent and 23.98, respectively, U.S. investor having a high risk-aversion coefficient prefers domestic to international major and emerging markets diversification and invests 77 percent and 82 percent, respectively, of his wealth’ locally. SD analysis suggests that global international diversification dominates entirely major and emerging markets diversification strategies for a U.S. risk-adverse investor having an increasing utility function. Finally, the findings of the SD tests reveal that risk-adverse U.S. investor having an increasing utility function prefers to diversify 45 percent of his wealth in major markets rather than in emerging markets.

Keywords: Optimal portfolios choices, Estimation Errors, Portfolio Resampling, Non parametric stochastic dominance approach, Monte Carlo and bootstrap p-values simulations