News sentiment and bond risk premia

Abstract

This study examines whether and how news sentiment regarding interest rate changes predicts bond risk premia in Korea, a leading emerging market. Using machine learning techniques, we construct the Bond News Sentiment Index (BNSI) from news articles on interest rates. Our analysis reveals that the BNSI has a hump-shaped relationship with bond risk premia in the following month. This result suggests that a stronger consensus of news articles on interest rate changes signals decreases in uncertainty, leading to lower bond risk premia. We propose a BNSI-based forecasting factor, which exhibits predictive power beyond existing factors both in- and out-of-sample