台灣共同基金績效持續性之檢定：多構面績效指標之比較

Abstract

本文旨在探討1999~2004年間國內共同基金，是否存在顯著之績效持續性。首先，以資料包絡分析法(Data Envelope Analysis, DEA)法推估技術效率，並與傳統之Sharpe's measure 及Jensen's alpha指標作爲衡量台灣共同基金績效之三項績效指標。其次，利用25%-50%-25%之分群比例，將樣本基金依前述指標之衡量值，分別區分爲低、中及高績效群組，並利用ANOVA及t檢定，據以驗證分群之適切性。最後，使用虛擬變數聯立迴歸模型，驗證樣本基金是否存在績效持續性之特徵。實證結果顯示，在技術效率的衡量下，我國股票型基金具有一年期績效持續性之現象，隱含基金投資者應採取動能(momentum)投資策略；但在Sharpe's measure及Jensen's alpha之衡量下，並未出現績效持續性之特徵。上述實證結果之差異，本文認爲應是DEA分析同時將交易成本（經理費、手續費、週轉率）及風險（標準差）訊息納入模型中，而有助於績效之可測性。

Using the data of Taiwan's domestic equity funds during the period of 1999 to 2004, this paper explores the persistence phenomenon in mutual fund performance. We apply Data Envelopment Analysis (DEA) to estimate relatively technical efficiencies among funds. For comparison, we also apply Sharpe's measure and Jensen's aplha to evaluate sample funds' performance. We then rank funds by the above three performance measures, respectively, and divide them into low (bottom 25%), middle (middle 50%), and high (top 25%) performance groups. Last, we set up dummy variables for every performance groups, and employ Iterative Seemingly Unrelated Regression to examine the persistence of funds' performance. The empirical results show that, under the measure of technical efficiency, there exists one-year performance persistence phenomenon, but not two-year and three-year persistence. As a result, the evidence suggests that fund investors should take the repeated-winner strategy. In addition, the analysis fails to find the existence of performance persistence under Sharpe's measure and Jensen's alpha. The different results among three performance measures could be due to the fact that DEA model allows us to takes transaction costs (expenses ratios, loads, turnover) and risk (standard deviation) into consideration at the same time.