此篇論文藉由一階幾乎隨機優越(Almost First-Degree Stochastic Dominance, AFSD)選股法則建構投資組合，採取反向策略買進過去表現較差(AFSD dominated)、放空過去表現較佳(AFSD dominant)的股票以檢視其股價報酬是否有顯著之超額報酬。相較於最為廣泛使用的平均數-變異數準則(Mean-Variance framework)，幾乎隨機優越法則不需要假設特定的報酬類型或是效用函數，僅以報酬率排序分析的方式判斷風險性資產的優越性。我們以台灣股票市場的上市公司作為研究標的，樣本期間從2010年1月至2014年12月，探討由過去6個月排序期為基礎的幾乎隨機優越投資策略，在各個持有期之下是否具有統計顯著的正報酬。 實證結果顯示，無論持有期為3個月、6個月、9個月或是一年，幾乎隨機優越投組皆能獲得統計顯著的超額報酬，並經Carhart的四因子迴歸模型進一步解釋台灣股票報酬的異常性。此外，我們亦針對一階幾乎隨機優越的篩選準則進行穩健性分析，隨著AFSD臨界值(critical value)的改變，每個移動窗格內投組所含的股數也跟著變動，然而，其績效表現仍具統計顯著的正報酬。 ;The purpose of this study is to construct zero-cost portfolios based on almost first-degree stochastic dominance (AFSD) rules and then examine the performance of these arbitrage portfolios. By longing dominant stocks and short selling dominated stocks, our investment strategy hypothesize that past losers will outperform in the future. Compared with the most widely accepted Mean-Variance framework, AFSD rules require neither a specific return distribution nor a specific utility function. It allows small violation of stochastic dominance and considers the preference of “most” investors. Our empirical treatment targets public firms in Taiwan stock market from 2010/01/01 to 2014/12/31. We form portfolios through the previous 6-month ranking period and hold them up to 12 months. All results show that AFSD portfolios produce statistically, significant excess returns. Moreover, these returns are robust with Carhart four-factor model under various criteria of the AFSD investment strategy.
