Submitted by: Industrial Fund Management Co., Ltd.

Fund performance evaluation is an important research topic in domestic fund industry at present. Evaluating the investment performance of its funds reasonably and objectively can provide pre-and post-event support and post-event performance evaluation for internal investment research; On the other hand, it can also show more convincing investment performance to customers, so that they can have a deeper understanding of the operational performance of the fund management team.

The positions of fund portfolios, especially open-ended active management fund portfolios, are usually in frequent changes, and fund managers will actively invest according to market trends; Customers frequently purchase and redeem funds, which also has a great impact on the fund portfolio. Fund managers will have to buy and sell stocks according to the amount of purchase and redemption, and the flow of purchase and redemption funds will directly lead to the change of the weight of stocks held. So, for fund management companies, what kind of performance evaluation method is adopted to reflect the management ability of fund managers?

Overseas scholars have not studied this issue for a long time, but so far, several practical multi-period attribution processing methods have been developed. At present, several studies are as follows: according to the four concept combinations of brinson model, there is an accurate solution to the multi-period ownership at the fund level (without subdividing assets); There is no exact solution at the level of corresponding assets (industries) or individual stocks.

Obviously, the multi-period ownership at the fund level can be handled by various accurate calculation methods. However, if we want to evaluate the fund manager's excess ability to subdivide assets, such as industries and individual stocks, we need to use multi-period attribution method at the category level. On the basis of the research of Arnarson et al. (2003), we extend the method proposed by DaviesandLaker(200 1) to make it applicable to the attribute analysis of category assets. We provide an accurate multi-period brinson attribution analysis framework, which is subdivided into stock dimensions and based on trading days. Through this framework, we can accurately measure the contribution of the fund portfolio in any time period and decompose the excess returns into individual stocks, thus facilitating internal performance evaluation and investment reference.

Based on the above model, we also provide an empirical analysis based on real fund performance, and analyze the management ability of fund managers in detail, in order to help domestic fund peers and other buyer institutions to promote and improve the performance evaluation of portfolio.