Welcome

I develop a model in which news outlets choose headlines to maximize the “information gap” relative to the full article and thereby stimulate reading. A central result is that economic incentives alone can endogenously generate belief-reversing headlines, even in the absence of ideological bias and with hard information constraints. Using a novel dataset of 100,000 U.S. news articles and LLM-based text analysis, I show that headline–content divergence increases with an article’s semantic heterogeneity, decreases with its informativeness, and that a non-negligible share of headlines meaningfully misrepresent the article’s main message.

I exploit a unique dataset in which each article appears simultaneously in two attention rankings—Most Viewed and Most Shared—to study how private and public use of news diverge for identical content. I combine these empirical patterns with a simple model of informational and signaling utility. Preliminary results reveal a systematic “scope wedge”: locally useful articles attract more private attention, while nationally framed or identity-related pieces are disproportionately shared. Moreover, contrary to common claims, shared articles are not simpler—they are more sophisticated, consistent with signaling rather than “dumbing down.”