有鑑於過去研究臺灣股票報酬決定變數之文獻結果分歧，本文 除了審慎決定參數估計方法、樣本選擇、投資組合建構以及影響股 票報酬的相關變數之外，尚將找出影響時間序列與橫斷面股票報酬 的因素，並同時探討過去文獻分歧的原因。本研究的實證結果顯 示，當使用不同公司特徵建構投資組合、不同投資組合分組數以及 樣本差異時，時間序列與橫斷面的迴歸結果皆會受到影響，故後續 研究應採用不同的方法進行穩健性檢測，以避免實證推論產生偏 誤。此外，本文亦發現，Nelson 四因子模式對時間序列股票報酬具 有不錯的解釋力；而情緒指標、波動性風險因子、權益帳面對市值 比、益本比、營收市值比、個股週轉率、個股成交量、6 個月期之 動能與股票報酬標準差皆能解釋臺灣橫斷面股票報酬。

In order to clarify the inconsistent findings in the Taiwanese asset pricing literature, this study cautiously determines the adequate methodologies used for parametric estimation, sample selection, portfolio construction, and model specification. The main purposes of this study are to explore what factors can better explain the time-series and cross-sectional stock returns in Taiwan and to propose some possible reasons for the inconsistent findings in prior studies. Our empirical evidence indicates that the factors that explain time-series or cross-sectional stock returns will change when firm’s characteristics used to build portfolios, the number of portfolios used in grouping stocks, or sample selection rules are different. Therefore, in order to avoid biases in making empirical conclusions, a robustness check is indispensable for any asset pricing study in Taiwan. The evidence herein shows that the Nelson four-factor model is more powerful in explaining the time-series stock returns. In contrast, the sentiment index, volatility risk factor, book-to-market ratio, earnings-to-price ratio, sales-to-price ratio, stock turnover, trading volume, six-month momentum, and standard deviation of stock returns can all well explain the cross-sectional stock returns.