Forecasting TAIEX and FITX with Affirmative and Doubtful Investor Sentiments

Abstract The existing literature have used media messages as a proxy variable for investor sentiment, but they mainly classify sentiment into positive or negative categories based on the words used in news articles, without much attention to the degree of affirmative or doubtful conveyed by the words used. Thus, in addition to classifying news content into positive or negative sentiment, this study also measures the degree of affirmative or doubtful expressed in the news articles in order to achieve more accurate predictive results. The study converts qualitative text to two quantitative scores (sentiment ratio and affirmative ratio) and investigates the predictive ability of these two variables on stock index returns and volatility in Taiwan’s case. The empirical findings indicate that only affirmative ratio exhibits a significant and negative impact on the one-day ahead returns of the Taiwan Stock Exchange Capitalization Weighted Stock Index (TAIEX) and the Taiwan Stock Exchange Index Futures (FITX). Additionally, the volatility of returns in both future and spot markets is significantly influenced by both sentiment ratio and affirmative ratio. JEL Classifications: C22; G10; G40. Keywords: Investor sentiment; Text-mining; TAIEX; FITX; Affirmative ratio.