事件研究法的實證分析, 在過去已經被廣泛地使用在財務與會計領域上。在日報酬率偏離常態分配的情況下, 一般的有母數檢定方法似乎較易受到質疑, 而過去也有無母數與拔靴的檢定方法來改善檢定力。摺刀法是能改善統計量偏誤的方法, 因此本論文目的是利用一般常用的有母數方法、無母數方法、拔靴法(bootstrap) 與拔靴後摺刀法(jackknife-after-bootstrap)的事件研究法, 以台灣上市(櫃)日報酬資料, 利用模擬方式進行檢定, 並以人工方式增加定量事件期異常報酬率, 比較各種檢定方法的檢定大小及檢定力。本論文研究結果顯示, 在檢定大小落入信賴區間範圍之內的條件下, 拔靴事件研究法與拔靴後摺刀事件研究法之檢定力在增加定量異常報酬率時有較佳的表現。 Event studies are now in widespread use in the literature of accounting and finance of empirical studies. Many of the event study methods require the normality assumption. However, the stock returns may not be normally distributed, especially for daily returns. To avoid the issue of the normality assumption, the jackknife-after-bootstrapmethods is proposed in this research, which are free from any specific distribution assumption just like bootstrap. Simulations are carried out base on the real daily returns data in Taiwan stock market. The results show that the event studies incorporating the jackknifeafter-bootstrap methods outperform the bootstrap and traditional methods in most of the cases in terms of the size and power of the tests.
