多重檢定與股票報酬率橫斷面分析

Abstract

本論文使用多重檢定方法探討實證金融文獻裡的股市異常報酬現象。為了尋找可獲得異常報酬之選股策略，投資領域專家已不斷的檢定許多模型。若使用傳統假設檢定方法，所發現的優異選股策略可能為統計推論上偏誤所造成。在第一章，我們探討多因子投資策略是否真的提供投資人附加價值。由於投資人可使用不同排列組合建構多因子投資策略，我們必須使用多重檢定方法測試該選股策略之有效性。第二章探索股市橫斷面報酬率是否存在異常現象。該章節討論當我們可以針對異常現象進行分類時，是否影響多重檢定結果。

This dissertation joins the vibrant debate in the empirical finance literature about the cross-sectional stock return anomalies. The endless effort by the finance community to find profitable stock-picking rules has raised questions of data-snooping bias in the empirical findings. The two main chapters of this dissertation investigate the cross-section of stock returns with multiple testing method to eliminate the data-snooping bias concern. The first one examines the efficiency of multi-factor investment strategies prevalent in the equity ETF market. The second one explores the relevance of group information in testing the stock market anomalies.