Expectation Error

Abstract

I backcast expectation errors of credit spreads via machine learning. I use newspapers over the past century to construct text-based expectations of credit spreads and study the relationship between expectation errors and business cycles. The main result is that overoptimism about future credit spreads predicts lower GDP growth and higher unemployment over the medium run, even after controlling for past and prevailing credit spreads. This finding suggests credit-market sentiment is an important driver of economic fluctuations. Consistent with this story, I also find both the amount of net debt issuance and the ratio of debt to equity issuance increase following periods of overoptimism.

Keywords: Expectation, Textual Analysis, Behavioral Finance

JEL Classification: G4