Index-tracking Portfolio Replication Method And Empirical Research

With the development of the securities market, market index system is gradually completed. Indexation investment philosophy is becoming one of the most important investment strategies and methods in the world. As a key technique in the indexation investment, Index-tracking portfolio replication method has gained lots of attentions and studies at home and aboard. However, most of the researches are only based on the return rate and few of them studied the time series of price. When come to stock choosing, many scholars choose stocks by the rank of market value, the industry share or just choose them randomly. These methods have not made full use of history information. Besides, according to the informed researches, most of them only considered one replication method and tracking error or multiple tracking errors. Rarely people did research on comparing among different index-tracking replication methods. Aiming at those problems mentioned above, this article raises a new methodology to replicate index-tracking portfolio. Based on the time series of stock price, the first step, extract features by FastICA and cluster stocks by their different trend and character. Then pick out the proper stocks using Non-negative LARS algorithm and weight stocks by optimizing and solving the target function. In the positive analysis part, the article compared the method with many others from multi-perspectives. The results show that:for the in-sample tracking error, the method raised by this article is superior to all the others; besides, both of the ICA-cluster method and the cluster based non-negative LARS algorithms have good effects. As for out-sample, this method show a great results both in short-term and medium-term. What’s more, in the empirical research, a larger sample space is provided when choosing stocks. It shows that the index-tracking scheme works better when stocks are chosen from HS300index stocks rather than from all the stocks. Even though, this trial is an exploration as well as a rewarding experiment for its practical reference value.