Portfolio Re-sampling Methods And Their Effectiveness

Traditional research has put much effort on finding the unbiased and less-constricted parameters of Mean-Variance (M-V) Portfolio Theory. On one hand Mean-Variance is perfect in its mathematic sense. But on the other hand, M-V analysis is sensitive to changes from input parameters. Small changes in input will often lead to large changes in weight of the optimized portfolios.In the first two chapters, we introduce the development path and framework of M-V analysis. By Monte Carlo simulation, we illustrate that M-V analysis is sensitive to changes from input parameters. Small changes in input will often lead to large changes in weight of the optimized portfolios.To solve the problem of unrobustness, we introduce a new method in chapter 3. After detailed analysis of resampling method, we propose an improved resampling method in portfolio construction. The empirical analysis in Shanghai Stock Market, and the empirical results illustrate that resampling method, together with the improved resampling method, have more power when compared with the average of classical M-V analysis.In Chapter 4, we further our research in resampling method, introduces the definition of resampling efficiency and resample efficient region proposed by Michaud (1998). We make an improvement on Michaud's method of constructing the resampled efficient region. The empirical analysis on Chinese stock market has shown its applicability in Chinese stock market. We also propose to replace the normality assumption with the log-normality when constructing resampling.We summarize the paper in the Chapter 5.