為了檢測投資人在市場極端波動下的從眾行為與風格投資，本文將橫斷面報酬標準差指標分別以市場風險、市值規模大小及帳面市值比做拆解，並且採用日和月頻率的資料做檢測，而實證結果仍支持理性資產定價模型，亦即市場處於越極端波動的的情況下，股票報酬對基準值的敏感性越高，導致其共移程度越小。然而，將市場從極端空頭到極端多頭排序後，所有不同類別、不同頻率的橫斷面報酬標準差指標皆呈現U型曲線，而這也是過去的文獻所未曾提及的。此外，相較於過去的文獻，本文加入了一月效應和市場週轉率的概念捕捉投資人的從眾行為，結果發現在市場極端波動時，投資人在一月偏好小型股及高帳面市值比(價值股)的股票，也隱含了投資人的風格投資行為；而上期市場週轉率越低，亦即投資人的意見較一致，則可預測下期較高的從眾程度。 To detect investors' herd behavior under extreme market condutions, I decompose the cross-sectional variance based on the market risk, size and BM using both the daily and monthly frequency. The results support the rational asset pricing model, which implies there are higher dispersions when the market is volatile for its higher sensitive toward its benchmark value. However, by ranking all the market conditions from the extremely down market to the extremely up market, this thesis identified a unique U-shaped curve for cross-sectional dispersion measure. Different from previous studies, after considering the January seasonality and the market turnover ratio, I find that herding is stronger for stocks within small size portfolio and high BM portfolio in January, which leads to the higher comovement of equity returns with groups. Besides, the lower level of past market turnover can predict the higher degree of herd behavior in following period.
