Quantile Tracking Error Portfolio Investment Method Based On Multi-factor Stock Selection And Its Empirical Research

Portfolio investment theory has always been an important research hotspot in financial investment.Its main purpose is to spread risks and obtain returns.Modern investment strategies are mainly divided into active investment and passive investment,which have their own advantages and disadvantages.In this paper,combined with active investment and passive investment,taking CSI 300 as the benchmark index,we use the multi factor stock selection method to select the dominant stocks from the constituent stocks of CSI 300 over the years,and establish the quantile tracking error model.In the multi factor stock selection,this paper selects 8 categories of factors,including 36 specific factors,and then selects 21 effective factors through factor validity test.Based on the quarterly data from 2015 to 2018,a multi factor stock selection model is established according to the industry classification standard of CSI 300.In the multi factor stock selection model,IC mean weighting method,support vector machine and random forest are used to rank the stocks,and then equal weight weighting method is used to get the weighted stock ranking of various industries.Finally,the top 20,50 and 100 stocks are selected according to the industry proportion of CSI 300.In this paper,the traditional tracking error model and quantile tracking error model are established with 20,50 and 100 stocks respectively,and six tracking error measurement forms are selected as the objective function.The results show that the more the number of stocks selected,the lower the weight concentration.When the number of stocks selected is 20,the weight concentration of the traditional tracking error model is lower than that of the quantile tracking error model.When the number of stocks selected rises to 50 and 100,the weight concentration of the traditional tracking error model is higher than that of the quantile tracking error model.From the industry point of view,the financial real estate industry has the largest proportion of weight.The more the number of stocks selected,the more similar the industry structure is to the CSI 300.In terms of beta coefficient,tracking deviation and daily average tracking error,the risk of quantile tracking error model is higher than that of traditional tracking error model.In terms of daily average return and cumulative return,the return of quantile tracking error model is higher than that of traditional tracking error model.From the sharp ratio,we can see that the performance of quantile tracking error model is better than the traditional tracking error model in the case of comprehensive risk and return.After decomposing the excess return,it is found that the quantile tracking error model has larger excess return than the traditional tracking error model.In the rebalancing study,two models with the best performance are selected,and the periodic rebalancing method is used to optimize the model through combination rebalancing under different periods.The SAQu TER model has the best performance in the rebalancing portfolio with a cycle of 5 days,and the SAAQu TE model has the best performance in the rebalancing cycle of one month.In the robustness analysis,we select Shanghai 50 as the target index to prove that the quantile tracking error model is indeed robust.So quantile tracking error model is better than traditional tracking error model.