Text-Based Fear and Bond Risk Premia

Abstract

This paper aims to explore whether text-based sentiment can perform well at predicting bond returns. We propose a sentiment trend factor based on the bond fear index that is a powerful predictor for future bond risk premia in- and out-of-sample. Notably, the text-based fear trend factor outperforms the previous effective bond return predictors and remains significant in further robustness tests, including the parametric VAR bootstrap test of Bauer and Hamilton (2018). Moreover, we find that news media has a greater impact on bond returns. The potential economic explanation is also proposed for this bond risk premia prediction.

Keywords: Bond risk premiumText-based sentimentReturn predictability