Quantifying Informative Signals from Political Texts

• In Event: New Methods for Text-as-Data

Abstract

How do interest groups use political communication to appeal to the general public? Classical accounts of interest group politics have assumed that interest groups send signals to inform voters about the contents of policy. However, there are two kinds of signals that are hard to separate empirically: 1) ‘informative’ signals about contents of policy and who is likely to benefit and 2) ‘goodwill’ signals which burnish a group’s reputation but don’t convey information. We create an eight-category typology of these signals and generate high-quality manual annotations of press releases about trade policy. We develop an automated coding procedure using recently-developed large language models that can classify the presence of signals at the same level as human-coders. We find that ‘goodwill’ signals are sent more frequently than ‘informative’ signals suggesting that interest group communication is not primarily geared towards increasing voters’ knowledge about policy.

Authors

• Dahyun Choi, Princeton University

• Denis Peskoff

• Brandon Michael Stewart, Princeton University