本文以台灣市場為研究標的，研究事件對台灣市場的效應。過去事件研究多以異常報酬或波動度衡量單一事件或特定事件對個別公司或市場之衝擊效應，本研究主要研究目的是以風險與報酬的關係模式分析不同事件類型及不同事件視窗下對台灣市場衝擊之差異以及其資訊內涵。 本研究以迴歸分析法針對以下本文之研究目的進行探討。研究目的：(1)分析比較不同事件類型的資訊效應差異。(2)藉由波動度與報酬的變動關係探討事件效應的影響。(3)檢視是否因為不同的事件類型和事件視窗而有不一致的表現。 實證結果發現：(1)市場報酬和市場波動度面對不同事件類型存在不同的統計特性。市場對不可預測的、壞的及政治的事件影響較為顯著。(2)在事件前、事件日和事件後視窗的風險與報酬關係是不一致的，可能源自於事件前、事件後及事件日視窗存在不同的市場行為。(3)台股市場存在資訊洩漏及延遲反應的現象，屬於非效率市場。 The study compares the event impacts of difference types of events on Taiwan market. Previous studies use either of the abnormal return or volatility to extract the impact of a specific event on an individual company or markets. However, this study suggests the pattern of the risk-return relationship should be affected by the release of event, furthermore, the event impacts will be different with the types of event or the event windows. We apply the regression model to examine the event impacts on the risk-return relationship. There are three main studies of this research: Firstly, it compares the information contents of different types of events. Secondly, it examines the impact of an event on the pattern of the risk-return relationship. Finally, it finds the difference pattern of the risk-return relationship associated with the release of different types of event and the different event windows. According to the empirical results, there are three main findings: Firstly the market returns and market volatility have distinct statistical characteristics for different types of event. In general, the impact of an event on a market associated with unpredictable, bad and political types of event is more significant than with other types of event. Secondly the risk-return relationship is inconsistent over the event period resulting from the market activities during the pre-, post-event and event day windows. Furthermore, the inconsistency conditions what the type of the event window is. Finally the efficient markets hypothesis appears to be rejected by the Taiwan stock market because of prior-event information leakage and the delayed-response effect during the post-event window.
