周德瑋Chou, De-Wai劉俐利Liu, Li-Li2022-06-082021-07-142022-06-082021https://etds.lib.ntnu.edu.tw/thesis/detail/6671d7cf9eae72e4445e6493576ba197/http://rportal.lib.ntnu.edu.tw/handle/20.500.12235/118370投資人能在網路上透過不同來源的媒體獲取眾多資訊，並對獲取的資訊有所反應。因此本研究利用文字探勘技術，從鉅亨網、中國時報、經濟日報、自由時報、聯合報及蘋果日報的網站中，蒐集2019年7月至2021年1月與ESG主題ETF相關的新聞，並量化新聞詞語為新聞情緒。使用事件研究法，探討所有媒體來源的新聞情緒、不同媒體來源的新聞與其新聞情緒是否造成異常報酬率，並將造成異常報酬率之媒體進行差異檢定，接著使用迴歸模型分析所有媒體來源的新聞情緒與新聞數量、不同媒體來源的新聞情緒及新聞數量對異常報酬率之影響，以及探討新聞發布使市場反應的效率性。實證結果顯示，不同新聞媒體的影響力不同，財經新聞媒體如經濟日報、鉅亨網及新聞情緒強烈的蘋果日報在多數情況下比風格偏屬政治的聯合報及自由時報更具影響力，且多數新聞媒體影響當日報酬率，並不會持續影響至隔日的報酬率。Investors can obtain a lot of information from different sources of media on the Internet and react to the news in real time. Therefore, use text mining to collect news related to ESG-themed ETFs from July 2019 to January 2021, which are from the websites of Anue, China Times, Economic Daily News, Liberty Times, United Daily News, and Apple Daily, then quantitative news words as news sentiment. Use event study to explore whether the news sentiment of all media sources, the news of different media sources and their news sentiment cause abnormal return, and then use ANOVA to analyze the differences among means of the media that caused the abnormal return. Then use regression models to analyze the influence of the news sentiment and the news amount of all media sources and different media sources on the abnormal return. And discuss the efficiency of the market price react to the news. The empirical results show that the rate of return is influenced by varying degrees with different media. Financial news media such as Economic Daily News, Anue, and Apple Daily with strong news sentiment are more influential than the politically-styled media such as United Daily News and Liberty Times in most cases. Most news media affect the rate of return on the day and will not continue to affect the rate of return on the next day.文字探勘新聞情緒事件研究法迴歸分析text miningnews sentimentevent studyregression analysis學術論文