Models Selection And Empirical Analysis Of Indexing Investment Based On Tracking Error

Indexing investment is a medium-to-long-term and passive asset allocation method which is invested in index constituent stocks.Indexing investment method purchases all or part of the index constituent stocks to build a indexed portfolio,and it's purpose is to obtain roughly the same return as the target index,instead of overcoming the market to get excess returns.As a simple,efficient,and low-cost wealth management tool,index fund has an explosive growth in recent years.As of the beginning of April 2018,there were a total of 463 passive index funds issued by public funds in China,with a total fund share of 282.83 billion,and a total of 64 enhanced index fund,with a total fund share of 42.24 billion.As a result,index funds account for 58.59%of the equity funds.Index funds have the advantages of low risk,high transparency,and low costs.What's more,index funds play an important role in asset allocation,risk management,so index funds will have broader development space in the future.This paper takes indexing investment and the tracking error as the research direction to analyze the domestic securities market,and uses statistical knowledge such as correlation analysis,cluster analysis,and genetic algorithm to find the optimal investment strategy.Indexing investment is an optimization problem,this paper improves the model of the mixed integer quadratic optimization according to BIC criteria,and constructs a mixed-integer quadratic optimization model with penalties by adding a penalty for the number of constituent shares in the investment.And this model can achieve optimized sampling.At the same time,due to the adjustment of constituent stocks in the index,the indexing investment portfolio structure cannot be continued and the tracking error will be increased.Therefore,this paper calculates a long-term unadjusted SZSE 100 price index according to the construction method of the target index,and divides training samples and testing samples to compare of the tracking error of index portfolios constructed by the stratified sampling model,the optimized sampling model and the mixed-integer quadratic optimization model.And compare the advantages and disadvantages of different models and their generalization ability.The tracking error of the indexing investment portfolio constructed by optimized sampling-genetic algorithm is the smallest in tracking the unadjusted SZSE 100 price index,both in the training samples and the testing samples.Then,using three models to track the true SZSE 100 price index,it was found that the tracking error of the portfolio constructed by the optimized sampling-genetic algorithm is also the smallest.As a combination construction method that can effectively reduce the tracking error in each stage and has high generalization ability,the optimized sampling-genetic algorithm provides index fund managers with a practical and effective asset allocation method.