本文指在研究國內全市場及各產業市場在投資行為上是否存在從眾行為。台灣股票市場屬興新市場，資訊傳遞較不完整，政府干涉程度的多寡，以及台灣股市漲跌幅7%的限制等，造成投資人對股市的影響及從眾的現象應有所不同。本文中股票報酬離散度採以Chang, Cheng and Khorana (2000) 所提出的橫斷面報酬絕對誤差CSAD衡量方法；並以Hamilton (1989)提出的馬可夫轉換模型，能有效的處理偏態、厚尾及高峰的問題。適當掌握金融與經濟變數所面臨的結構改變，以及解決在報酬分配的產生的問題，較能貼切的反應經濟變數的真實狀況。 實證結果發現台灣在總市場中投資人的確存在顯著的從眾現象。綜合各產業分析，其實證結果發現除極少數產業外，其餘皆呈現明顯從眾現象。此外，台灣股市在總市場及產業市場中利用馬可夫轉換模型，發現高波動下確實在實證上是有更顯著的從眾現象。 The thesis aimed to study whether the whole domestic market and various industrial markets had herd behavior on investment behavior. Taiwan’s stock market belongs to a new market, with incomplete information transmission; the degree of interference by government as well as the restrictions of Taiwan’s stock market rising and declining by 7% causing the difference between the impact of investors on stock market and herd phenomenon. In this thesis, stock compensation dispersion took cross-sectional absolute deviation (CSAD) method proposed by Chang, Cheng and Khorana (2000) and Markov switching model proposed by Hamilton (1989) to effectively deal with skewness, fat tails and peak. Properly grasping structural changes facing the financial and economic variables as well as solving the problems about the compensation assignment can appropriately show the actual state of economic variables. Empirical results showed that significant herd phenomenon did exist among Taiwanese investors in the total market. Based on the comprehensive analysis of various industries, it was found that significant herd phenomenon did exist among various industries except a minority of industries. In addition, Taiwan stock market used Markov switching model in the total market and industrial market, finding that it indeed had a more significant herd phenomenon empirically under high volatility.