This study uses data from the Korea Treasury Bond (KTB) market to empirically analyze whether the sentiment analysis index of news that is related to the interest rate has predictability for excess KTB returns. For news sentiment analysis, word to vector (Word2Vec) and machine learning techniques are applied to 19,135 articles collected from 4 economics newspapers in Korea, from January 2005 to November 2021, to quantitatively evaluate the sentiment of interest rate hikes contained in the articles. The measured news sentiment indices are converted to the news sentiment factors, on which a regression analysis is performed with the forward rate factor [Cochrane and Piazzesi, 2005], macroeconomic factor [Ludvigson and Ng, 2009], and investor sentiment factor [Baker and Wurgler, 2007]. Results confirm that the news sentiment factor has statistically significant predictability for excess returns of bonds, even after controlling for all the aforementioned factors. According to the regression analysis of news sentiments and major economic indicators, a high correlation coefficient is observed between news sentiment and real economic indicators such as short-term interest rates and the growth rate of imports and exports, indicating that news sentiments contain information that is similar to shortterm interest rates and real economic indicators
