本文主要目的有二：一為測試幾個常用投資組合績效衡量指標的區別能力；二為檢踏過去文獻上所載基金普遍不具擇時能力的實證結果，是否由於衡量模型的系統性偏誤所致。儘管學術界已經發展出許多投資組合績效評估方法，一般投資人在選擇基金時，多半使用未經風險調整的基金報酬率或報酬率排名去評估基金績效。我們懷疑較複雜的指標不被應用的原因可能是因為其區別能力並不顯著高於未經調整的報酬率。利用模模分析，我們發現Sharpe指標、Treynor指標、Jensen指標及未經風險調整的基金報酬率在區別經理人能力的正確率上，並無顯著差異。另外，如我們所預期，評估時間愈長，異常能力愈高，則各指標之區別正確率也愈高。儘管各指標之區別能力平均而言並不太高，但投資人還是以選擇過去相對績效較優的基金為宜。在基金擇時能力方面，我們用五十個完全投資於股票的投資組合報酬率代入Treynor& Mazuy、Chang & Lewellen及Henriksson 的實證模型中，結果發現這三個模型皆顯示出負的擇時能力及正的選股能力。我們進一步尋找這個系統性偏誤的來源，發現與投資組合組成個股的規模有關。

This paper examines two puzzling phenomena in portfolio performance evaluation. Despite numerous performance evaluation indices(methods) were developed by the aca-demics, average investors do not have access to this information. Instead, funds returns in the evaluation periods or their rankings relative to other funds are usually reported in financial press without explicit adjustments for funds risks. We suspect the reason why performance indices lack supply in practice is because the marginal value of risk ad-justmentsis low for the fund selection purpose. We use Monte-Carlo simulation to compare the abilities of four performance measures, the Sharpe index, Treynor index, Jensen index, and portfolio return (not adjusted for its risk,) to identify fund managers superior abilities. The simulation results indicate that the above four measures have very similar screening abilities. Although the persistence of performance is low, it is still best for investors to select funds with good past performance. Many prior studies reported mutual funds on average exhibit negative timing ability and positive selectivity ability. Our simulation results show that the Treynor and Mazuy(1966), Chang and Lewellen( 1984 ), and Henriksson( 1984 ) model all have strong bias for negative timing and positive selectivity abilities. We also find that the bias is related to the size of indi-vidual stocks in the portfolio.