News Sentiment as Leading Indicators for Recessions

TL;DR

This paper introduces a novel news sentiment indicator derived from unstructured news data, which, when combined with traditional economic indicators, enhances the accuracy of recession prediction models.

Contribution

The paper develops a new sentiment metric from news data and demonstrates its effectiveness in improving recession forecasts when integrated with existing indicators.

Findings

The new sentiment indicator improves recession prediction accuracy.

Combining news sentiment with traditional indicators enhances model performance.

The approach provides a real-time measure of information polarity affecting economic outlooks.

Abstract

In the following paper, we use a topic modeling algorithm and sentiment scoring methods to construct a novel metric that serves as a leading indicator in recession prediction models. We hypothesize that the inclusion of such a sentiment indicator, derived purely from unstructured news data, will improve our capabilities to forecast future recessions because it provides a direct measure of the polarity of the information consumers and producers are exposed to. We go on to show that the inclusion of our proposed news sentiment indicator, with traditional sentiment data, such as the Michigan Index of Consumer Sentiment and the Purchasing Manager's Index, and common factors derived from a large panel of economic and financial indicators helps improve model performance significantly.