本論文主要研究動能投資策略，第一個部分探討以個股為操作對象之動能策略，及以投組為操作對象之動能策略，兩者報酬結構之差異，並解釋造成個股動能與投組動能在長短期報酬表現不同之原因。更明確地說，藉由將投組動能報酬拆解為個股報酬的時間可預測性(而非投組報酬的時間可預測性)以及期望報酬的橫斷面變異，本文得以清楚比較造成投組動能與個股動能策略報酬差異的原因。利用CRSP資料實證的結果顯示，個股橫斷面相關對投組動能報酬的影響，區分為兩個部分，不管個股橫斷面相關的符號為何，此兩部分都將彼此相互抵銷部份效果，此外，投組動能會大幅稀釋個股自我相關對動能報酬之影響，同時降低期望報酬橫斷面變異的效果。投組動能的上述特性加上個股時間數列可預測性的型態，是造成個股動能與投組動能策略報酬差異之原因。本論文的第二個部分是探討跨國動能投資策略，隨著全球化的持續進行，各國資本市場的日益開放與整合，跨國投資成為常態，因此跨國投資策略的研究益形重要。利用Datastream資料庫全球股價資料進行的實證結果顯示，動能報酬存在於視全世界為一整體的股票市場內，且在此單一的整體市場中，不管是個股動能或投組動能皆能獲利。在動能報酬來源方面，個股動能的報酬主要來自期望報酬的橫斷面變異，國家投組與產業投組的報酬來源有相當大的比例來自投組時間數列的可預測性，但兩者又有所區別，時間數列可預測性對國家投組動能的貢獻主要來自國家投組自我相關的部分，而產業投組動能則主要來自產業投組橫斷面相關部分。 This thesis is consisted of two essays that investigate various aspects of the momentum investment strategies. The first essay investigates the sources of individual stock-based and portfolio-based momentum profits by decomposing the portfolio-based momentum profits into time-series predictability of individual stock returns, and brings out the explanations of different performances of the two different kinds of momentum strategies in different investment horizons. Empirical findings reveal that the effects of cross-autocorrelation on portfolio-based momentum are divided into two parts; regardless signs of the cross-autocorrelations, this two parts will cancel out each other partly. Besides, the formation of portfolio will dilute the effects of autocorrelation on portfolio momentums and reduce the cross-sectional variation in expected returns. The characteristics of portfolio-based momentum mentioned above, and the time-series characteristics of individual returns, are the main reasons that cause different performance. The second essay analyzes global momentum strategies. Using Datastream database, we found that momentum profits exist in global stock markets as a whole, no matter individual stock-based or portfolio-based momentum. The profits of individual stock-based momentum mostly come from cross-sectional variation in expected returns, but a large part of portfolio-based momentum returns comes from time-series predictability. For country momentum, it is autocorrelation that contributes momentum profits, and for industry momentum, it is cross-autocorrelation that contributes momentum profits.
