本文以統計套利探討市場狀態對價格動能策略之影響,以驗證市場具有不效率性,並以行為財務學之處分效果與過度自信效果探討市場不效率性之原因。使用台灣上市之普通股週資料,樣本期間為民國88年1月01日至民國97年8月31日。 實證結果發現動能策略以傳統t檢定驗證時,在市場趨勢上升時具有顯著正向報酬,但當市場趨勢下降時,動能報酬呈現負向不顯著。統計套利四種模型之檢驗結果皆驗證臺灣股市動能策略週報酬具有顯著統計套利空間,且證實市場趨勢向上之動能策略較具有統計套利空間。當觀察市場趨勢不同時之處分效果與過度自信效果,發現市場趨勢向上時,各種分類下之淨過度自信效果一致皆為顯著正向,但當市場趨勢向下時,則會互相抵銷,使動能不顯著。最終以強韌度測試確認動能報酬並非來自公司規模、重疊期間、市場趨勢定義、市場摩擦等因素,臺灣股市確實具有不效率性。 In this study, we use the statistical arbitrage to analysis the influence of market states to momentum strategies in Taiwan and intend to demonstrate that if the market efficiency is tenable. We also use the “disposition effect” and “over-confidence effect” in behavior finance to explain the source of market inefficiency. Our sample is the weekly data of stock market in Taiwan from 1999 to 2008. Our result shows that when we use the traditional student’s t test, the momentum strategies have positive significant return in up-market, but negative insignificant return in down-market. In the four models of statistical arbitrage, the momentum weekly return has significant statistical arbitrage opportunity. The momentum return is more significant in the up-market than in the down market. We observe the “disposition effect” and “over-confidence effect” in different market states. The net over-confidence factors are significant positive in two classes by all categories in up-market, but offset in down-market and make the momentum insignificant. Finally, we use the robust test to confirm that the momentum return doesn’t come from firm size, overlapping, market state definition, market friction. The momentum strategies still have excess return and are inconsistent with market efficiency.