開放式股票型基金投資績效分類及投資報酬率預測之研究

Classifying Open-End Equity Mutual Funds and Forecasting the Rate of Return

吳宗正(Chung-Cheng Wu) ； 溫敏杰(Miin-Jye Wen) ； 黃綺年(Chi-Nien Huang)

《智慧科技與應用統計學報》 3卷2期 (2005/12) Pp. 117-139

https://doi.org/10.29807/JTITAS.200512.0007

摘要

本研究針對開放式股票型共同基金投資績效分類和其投資報酬率預測分別進行實證分析。首先，蒐集90年1月至92年12月間36個月之基金資料，根據各項基金績效評比指標作分類標準，其中績效評比指標包含淨值、週轉率、夏普指標、貝他係數、及崔納指標等；其次，蒐集88年1月至92年12月間60個月之基金報酬率資料，探討其與總體經濟指標間之關係。其中，總體經濟指標包括躉售物價指數、貨幣供給額之M1b及M2、景氣信號判斷分數、退票比率、利率、外匯淨值、進出口貿易差額等八項變數。
本研究分別以傳統統計方法及類神經網路方法進行資料分析和實證模型之比較，試圖找出最佳模式解。在分類部份，以正確分群率之高低來判斷模式好壞，結果以集群分析配合區別分析效果較佳；在預測部份，則以殘差值之大小做為判斷依據，結果以改良式倒傳遞網路效果最佳。因此，依據本研究之實證分析可知，利用類神經網路和統計方法去配適模式時，各有優點，此結論與Warner和Misra (1996)研究相似。

關鍵字

共同基金 ； 基金績效評比指標 ； 總體經濟指標 ； 統計方法 ； 類神經網路

並列摘要

Recently, diverse investment products have become more and more popular since interest income of deposit cannot catch up the inflation rate. Among all outlets for investment, mutual fund is one of investor’s favorites. Due to its characteristic of accumulation and less risk, investors having less financial support also can get a chance to make the profit from investment portfolio. Moreover, authorizing a professional manager to handle their funds could save time, so mutual fund gradually becomes a popular product in the commercial market. This study focuses on Equity Mutual Funds and includes two main directions. First, different funds are classified based on their performance. The data used in this study are collected from Jan. 2001 to Dec. 2003. The evaluation index of Mutual Funds includes net asset value, turnover rate, Sharpe Index, Beta coefficient, and Treynor Index; Secondly, based on historical data of rate of return from Jan. 1999 to Dec. 2003, this research explores the relationship between the rate of return and macroeconomic indicators including the wholesale price index, M1b and M2 of money supply, Prosperity Score, refunding rate, interest rate, net value of foreign exchange, and import and export balance of trade. This research compares traditional statistical methods and Artificial Neural Networks and to get the better result. For classification, Discriminant analysis as a result is performing better than other methods due to their low misclassification rates. For prediction, an improved BPN method reveals to be better than other models. The result shows that Artificial Neural Networks and traditional statistical methods are good techniques to do data analysis. Our studies have the similar conclusions as Warner and Misra (1996).

並列關鍵字

Mutual Fund ； Performance Evaluation Index of Mutual Funds ； Macroeconomic Indicators ； Statistical Methods ； Artificial Neural Network