2026 Theses Doctoral

Essays on Political and Organizational Economics

Ciccarone, Andrea

This dissertation studies questions in political and organizational economics using methods at the intersection of economics and machine learning. The following three chapters develop evidence and tools on how information and incentives shape behavior and outcomes in economic and political environments.

The first chapter examines how partisanship operates in modern video news. Political content is increasingly consumed in short, visually rich, low-attention formats, yet existing work on media bias is almost entirely text-based. I develop a multimodal measure of partisanship and show that images transmit partisan content rapidly through emotional cues, while text conveys it more slowly through substantive information. A survey experiment with real immigration news confirms the asymmetry: visuals move emotions and behavior immediately, whereas text shifts policy attitudes only under sustained exposure. Standard accounts of persuasion and media quality policy must therefore be reconsidered for a video-first environment.

The second chapter, joint with Dan Biderman, David Blei, Wei Cai, Amir Feder, and Andrea Prat, develops a topic model that recovers broad, context-agnostic themes from multi-source text. Standard topic models conflate what people talk about with how they talk, so estimated topics absorb industry jargon or sentiment polarity rather than substantive content. The Fixed Effects Topic Model (FETM) embeds environment fixed effects directly in the generative process, separating global themes from environment specific language. Applied to a large corpus of Glassdoor reviews, FETM recovers corporate culture topics that are stable across industries and outperforms conventional models on predictive fit and downstream causal inference, providing portable text-based measures for both organizational and political economic applications.

The third chapter, joint with Luigi Caloi, studies the design of performance-based intergovernmental transfers. Conditioning transfers on local outcomes can raise efficiency by incentivizing municipalities, but it can also redistribute funds toward higher capacity governments, generating equity costs. Using administrative data from Brazilian states that tied intergovernmental transfers to educational performance, we find large test score gains and limited equity losses from the incentive component, and negligible effects from contemporaneous changes to passive transfers. We cast the problem as a principal-agent model in which performance-based transfers simultaneously shift incentives and endowments, and argue that the optimal transfer mix depends on the misalignment between municipal and central government preferences.