過去對於因子模型在台灣市場的實證研究仍未有一致的看法，由於臺灣市場特性還有投資人結構與美國不同，在美國可解釋股票報酬的因子與模型在台灣市場不一定適用。本研究使用橫斷面分析檢驗Fama-French的五因子模型在臺灣市場的解釋力，發現五因子模型中的獲利因子與投資因子大部分不顯著，且部分截距項顯著異於零，因此，本研究將五因子模型加入動能因子、流動性不足因子、投資人情緒因子，希望透過考量不同變數，提升模型解釋力，並找出適合台灣市場的模型。實證結果發現動能因子與流動性不足因子在臺灣股票市場的解釋力不高，而投資人情緒因子對臺灣市場的股票報酬有部分的解釋能力。且由於過去臺灣對於因子模型的研究大部分是討論平均報酬率的結果，因此本研究進一步以分量迴歸檢驗自行建構的八因子模型，探討不同分位數下因子係數及模型解釋力的變化，實證結果發現，在第10百分位時八因子模型有最高的解釋力且解釋力隨著百分位的提高有下降的趨勢，與美國市場的研究結果一致。;In the past, there was still no consensus on the empirical research of factor models in the Taiwan market. Due to the characteristics of the Taiwan market and the difference in investor structure from that of the United States, the factors and models that explain stock returns in the United States may not be applicable in the Taiwan market. This study uses cross-sectional analysis to test the explanatory power of Fama-French′s five-factor model in the Taiwan market. It is found that most of the profit factors and investment factors in the five-factor model are not significant, and some of the intercept terms are significantly different from zero, so In this study, the five-factor model is added to the kinetic energy factor, the lack of liquidity factor, and the investor’s sentiment factor. It is hoped that by considering different variables, the model′s explanatory power will be improved and a model suitable for the Taiwan market will be found. The empirical results found that the kinetic energy factor and the lack of liquidity factor in the Taiwan stock market are not very explanatory, while the investor sentiment factor has a part of the explanatory power for the stock returns in the Taiwan market. And since most of the research on factor models in Taiwan in the past was the result of discussing the average rate of return, this study further tested the self-constructed eight-factor model with component regression, and explored the changes in factor coefficients and model explanatory power under different quantiles. It is found that the eight-factor model has the highest explanatory power at the 10th percentile and the explanatory power tends to decrease as the percentile increases, which is consistent with the research results of the US market.
