(Almost) 200 Years of News-Based Economic Sentiment

Abstract

Using text from 200 million pages of 13,000 US local newspapers and machine learning
methods, we construct a 170-year-long measure of economic sentiment at the country
and state levels, that expands existing measures in both the time series (by more than a
century) and the cross-section. Our measure predicts GDP (both nationally and locally),
consumption, and employment growth, even after controlling for commonly-used predictors,
as well as monetary policy decisions. Our measure is distinct from the information
in expert forecasts and leads its consensus value. Interestingly, news coverage has become
increasingly negative across all states in the past half-century.

Keywords: Business cycle, macroeconomic news, economic sentiment, monetary policy, textual analysis, machine learning, big data, neural networks

JEL Classification: G1, G4, E2