台灣股票市場之因子動能與動能因子實證研究

Abstract

因子投資領域於財務金融學術界已蓬勃發展多年，相比之下，因子動能於近幾年才逐漸被學界廣為討論，Ehsani and Linnainmaa (2022)提出有別於過去學者之論點，他們認為因子動能並非一個獨立於其他因子存在的風險因子，其為總和了所有其他因子的自相關性現象，因此，因子動能並非與其他因子無關，而是與所有因子都相關。故本研究旨在檢驗Ehsani and Linnainmaa (2022)所提出之新論點在台灣股票市場中是否同樣成立。 本研究採用台灣股票市場資料，並遵循Ehsani and Linnainmaa (2022)之研究方法探討因子動能現象。研究分為四個部分，第一，分別使用等權重以及市值加權權重兩種方法組成因子，並比較台灣股票市場十一個因子之報酬率表現；第二，透過時間序列迴歸分析檢驗因子前後期報酬率之相關性，探討台灣股票市場十一個因子之因子動能現象；第三，分別透過資產定價、時間序列迴歸模型角度，分析因子動能與動能因子之關係；最後，分析個股動能因子與其他因子之非條件、條件相關性。 台灣股票市場之實證結果大致符合Ehsani and Linnainmaa (2022)使用美國股票市場實證研究後所得出理論，第一，多數因子的報酬率呈現前後期正相關之因子動能現象。第二，因子動能可以有效解釋個股動能因子之變異，但是個股動能因子無法有效解釋因子動能之變異，因此，兩者之關係為因子動能為因，個股動能因子為果。第三，雖然各個因子與個股動能因子在時間序列上之非條件相關性極低，但透過控制因子過去報酬率表現後，各個因子與個股動能因子變成擁有高度相關性，故個股動能因子實際上與其他因子顯著相關，而由於因子動能是個股動能因子背後主要驅動力來源之一，因此可以得出因子動能並非一個獨立於其他因子存在之風險因子，而是其他所有因子前後期報酬率正自相關性總和之結果，故因子動能是一個可以用來針對所有其他因子的市場擇時策略。

Factor investing has flourished in the financial academic community for many years. Recently, however, the concept of factor momentum has garnered increasing discussion. Contrary to previous scholarly assertions, Ehsani and Linnainmaa (2022) proposed that factor momentum is not a distinct risk factor but rather a phenomenon that aggregates the autocorrelation of all other factors. Thus, factor momentum is not unrelated to other factors but is instead correlated with all of them. Accordingly, this study aims to examine whether the new perspective offered by Ehsani and Linnainmaa (2022) also holds true within the Taiwanese stock market. This research utilizes data from the Taiwanese stock market, following Ehsani and Linnainmaa's (2022) methodology to investigate factor momentum. The study is divided into four sections: Section I constructs factors using equal-weighted and market-value-weighted approaches and compares the performance of eleven factors. Section II measures the autocorrelations in the returns of these eleven factors. Section III uses asset pricing models and time-series regression models to analyze the relationship between factor momentum and momentum factors. Section IV analyzes the unconditional and conditional correlations with the momentum factor. The empirical results generally corroborate the findings derived from Ehsani and Linnainmaa's empirical study in the U.S. market. The returns of most factors are positively autocorrelated, indicative of the factor momentum phenomenon. Factor momentum effectively explains individual stock momentum; however, individual stock momentum does not explain factor momentum. Lastly, although the unconditional correlations between the momentum factor and other factors are low, the conditional correlations increase significantly based on whether the mean of the factor’s return over the prior year was positive or negative. Therefore, individual stock momentum is significantly related to other factors. Since factor momentum is a primary driver of individual stock momentum, it is not an independent risk factor but rather a result of the positive autocorrelation across all factors' returns. Consequently, factor momentum can be used as a market timing strategy for all other factors.