本文採用Brown and Warner（1980，1985）的方法，以臺灣股票市場（RSE）日資 料為樣本，研究各種事件研究法（event study）常用的期望報酬率模型及檢定方法的優劣 。研究結果顯示，就期望報酬率模型而言，以市場模式調整法（market model）表現最佳， 市場指數調整法（market-adjusted rturn）在許多設定下皆產生拒絕過少的系統性偏誤， 而平均數調整法（mean-adjusted return）的檢定力則相當低；就檢定方法而言，由於TSE 之日資料嚴重偏離常態，無法符合有母數檢定法之常態假設；而Corrado（1989）所提出的 無母數序列檢定（rank test）檢定力則明顯高於另外三種有母數檢定法。

　　　　　Using daily data from Taiwan Stock Exchange, this paper investigates the performance of various models and tests frequently used in event studies based on the Monte Carlo simulation design of Brown and Warner（1980，1985）. As has been documented in the literature, we find that the "market model" outperforms other models. Since the daily return data significantly deviate from normality, our results also show that Corrado's（1989）rank test performs better than traditional parametric tests.