Abstract
I explore how electoral campaigns affect the market for partisan news in Spain. I use machine learning and large language models (LLMs) to build a novel slant index that I match to high-frequency audience-meter data on television consumption. This allows me to compare how the same story is framed across outlets and how many people watched it. I integrate these measures into a structural model of news demand and supply. Outlets choose the political framing of the news and viewers select their preferred information source based on it. To identify viewers’ preferences for political content, I exploit exogenous changes in the mix of political events that constrain what outlets can cover. During the campaign period, demand becomes more polarized: viewers strongly screen out favorable coverage of the party they oppose. On the supply side, outlets specialize and face lower costs of producing slanted coverage that aligns with their political stance. I evaluate the effects of a proportional airtime requirement, the standard rule in television regulation during campaigns. Outlets comply by becoming more partisan, resulting in a more polarized media environment.
For past presentations, click here.