A Combination Of Sophisticated And Naive Diversification Strategy With Different Data-generating Process

Markowitz proposed the traditional mean variance model in 1952,which was considered as the beginning of the modern portfolio theory.However,the model has the problem of parameter uncertainty.Many other scholars made further studies based on his study and obtained a lot of sophisticated estimation models.Many scholars also fund that the estimated Markowitz portfolio rule and most of its extensions are underperform the naive 1/N rule.DeMiguel,Garlappi,and Uppal(2009)report that naive diversification dominates sophisticated rules in out-of-sample asset allocation tests.Tu and Zhou(2011)proposed a combination method,which combines the naive 1/N rule with the sophisticated rule,and the new strategy obtained is superior to the original one.Our analysis suggests that this is largely due to their research design.Therefore,this paper improves their conclusions based on three different types of data generation processes and the real Chinese stock data.Research implying that,firstly,when the time interval is small,naive 1/N rule outperform sophisticated rules in most cases.Secondly,when the number of assets is small,the risk aversion is small,the mispricing do not equal to zero and the idiosyncratic volatility is large,we find that ML strategy and Kan-Zhou strategy is better than naive rule.Thirdly,when the mispricing is high,the combination strategy of Tu and Zhou(2011)may not outperform their original strategy.Finally,in the study of the actual data based on Chinese stock market,1/N strategy may not be able to overcome the complex strategy,the performance of the combination strategies depend on the length of test window,that is,the amount of data.In most cases,the performance of the portfolio strategy in the Shanghai stock market is superior to that in the Shenzhen stock market.