Time-Varying Relationships Between the “Sentiment” of U.S. Public Opinion and News Media

Abstract Pollsters and journalists in the United States often make generalized statements about the state of the country, in particular, its “mood” or “sentiment.” The claims often rely on responses to survey questions asking about the state of the country. Little is known about the ebb and flow of this sentiment, however. This paper develops a measure of public sentiment, pooling results of thousands of polls between 2000 and 2021. It then develops an analogous measure of media sentiment based on a content analysis of millions of network television news transcripts. Analyses explore differences across data sources (i.e., polling houses and television networks) and consider the relationship between public and media sentiment. Results suggest that the measures are correlated concurrently, and there is some evidence that public sentiment responds to earlier media coverage. However, connections between the measures wax and wane over time, highlighting the potential importance of accounting for time-varying parameters in the study of public opinion and media effects.