Real-Time Macro Information and Bond Return Predictability: Does Deep Learning Help?

Abstract

This paper examines whether deep/machine learning can help find any statistical and/or economic evidence of out-of-sample bond return predictability when real-time, instead of fully-revised, macro variables are taken as predictors. First, we find some statistical evidence for forecasting short-term non-overlapping excess bond returns using deep learning models. Second, for forecasting overlapping excess bond returns, more statistical evidence derives from using deep learning models and other machine learning models. However, all statistical evidence is much weaker than that found from using fully-revised macro data and generates minimal economic gains for a mean-variance investor, regardless of her level of risk aversion and whether she can take short positions.

Keywords: Deep Learning, Machine Learning, Bond Return Predictability, Real-Time Macro Data, Overlapping and Non-overlapping Returns

JEL Classification: C45, C53, G11, G12, G17