Evaluating China's Stock Fund Based On Random Alpha Model

After decades of development and expansion of China's fund market,mutual fund has already become one of the most crucial investing instrument in China's capital market.Until 2017,according to data from Wind,the number of fund management companies offering mutual funds has reached 115,and the number and size of all these funds are 4500 and over 9 thousand billion RMB.According to rules set by AMAC,stock mutual funds are funds with stock assets occupying over 80%of their whole values.Besides,according to categories in Wind,normal stock mutual funds are stock mutual funds not investing in stock index(such as ETFs).Normal stock mutual funds used to be the most popular investing instrument,however,they are losing their popularity from the perspectives of product diversity and whole size with respect to hybrid funds and bond funds.Normal stock mutual funds,as one of the most traditional types of investment for individual and institutional investors,are important to them.Hence,it is needless to mention the importance of finding appropriate ways to evaluate the performance of them.Traditionally,there are three measures used to evaluate the performance of a single normal stock fund:the single-factor measure,the multifactor measure and the non-benchmark measure such as VaR,ES,DEA and Bootstrap.In fact,the former two types of measure are most commonly used not only among academic empirical studies,but also in industry.Jensen's alpha,as one of the most classic ways to evaluate the absolute performance of a mutual fund due to the easiness and solid academic background of it,is found to have various flaws practically,such as the overreliance on chosen samples,descriptive statistics and regression,though used quite widely.These flaws lead to the problems such as sample bias,model overfitting and the considerable large estimation error caused by missing the cross-section information of the samples while only concentrating on time series information of the individual fund.Thus,it is urgent to find and choose a better evaluating way to improve the performance of classic Jensen's alpha model.Through reading China's and international literature concerning improvement of Jensen's alpha,it is discovered that random alpha model has been proposed and improved in recent years.The core issues of this model are that:1st,it tries to capture the characteristics of the population of sampled funds and evaluate them;2nd,it uses this information to calibrate the result of performance evaluation of a single fund.With incorporating the information from sample population,this model could enlarge the usage of information in evaluating a single fund and avoid those problems haunting the classic Jensen's alpha.Therefore,it could enhance the adaptability of classic Jensen's alpha.That is why this thesis focuses on this model to explore its performance using empirical study method.In this thesis,the monthly return data of 125 normal stock mutual funds are selected with the time horizon from February 2014 to February 2017 to conduct empirical study on evaluating fund performance using classic Jensen's alpha and random alpha model.The results of the evaluation of the performance of the funds as a whole and individual funds using the two measures are compared and come to the conclusions:1st,random alpha model could capture the information and characteristics of samples as a whole better;second,in terms of performance evaluation of individual funds,this model could use the information captured in conclusion#1 to effectively lower the estimation error in the evaluation of individual funds' performance,thus enhance the effectiveness of the performance evaluation and prevent model from overfitting.This thesis also tries to explain the empirical results of random alpha model in scope of finance and tries to conclude the ways to use this model in empirical study.