讲座通知：A Statistical Explanation to Markowitz Optimization Enigma
报告题目: A Statistical Explanation to Markowitz Optimization Enigma
The renowned Markowitz mean-varianceportfolio analysis forms the foundation of modern investment science. However,the empirical performance of estimated mean-variance efficient portfoliosoftentimes does not come close to meet the expectation when there are more thanseveral assets in the investment universe. As many have observed recently, theymay even underperform the naive diversification which simply assigns equalweights across all assets. These findings inevitably cast a shadow on theusefulness of the Markowitz theory. To re-assert the practical value ofmean-variance analysis, we show here that this ``Markowitz optimizationenigma'' (Michaud, 1998) could be resolved by carefully balancing the tradeoffbetween the estimation error and systematic error through the so-calledsubspace mean-variance analysis. In addition to the consistent improvementobserved on real and simulated data sets, we prove that in a large market, itis possible to use this strategy to construct portfolio rules whose performanceclosely resemble that of theoretical mean-variance efficient portfolios.
袁教授2004年在美国威斯康辛麦迪逊大学统计系毕业获得博士学位。之后进入美国佐治亚理工工业工程系作为助理教授。2013年起在美国威斯康辛麦迪逊大学统计系做教授。同时也是Morgridge Institute for Research的Senior Investigator。2009年获得美国NSF的CAREER Award，2015年获得IMS的Fellow.