Going viral: Inflation narratives and the macroeconomy

Abstract

This paper investigates the role of media narratives in shaping household inflation expectations. We construct dynamic indicators of inflation narratives using a semi-supervised topic model in combination with Latent Semantic Scaling, applied to a large corpus of U.S. news coverage from 2018 to 2023. Using Granger causality tests and local projections, we show that narratives concerning government spending, supply chain disruptions, the war in Ukraine, and corporate profits are systematically related to changes in both short- and medium-term inflation expectations. Further, our analysis reveals substantial heterogeneity across socioeconomic characteristics, contributing to a growing literature on the determinants of heterogeneous expectations.