« Back to Results


Paper Session

Friday, Jan. 6, 2023 8:00 AM - 10:00 AM (CST)

Hilton Riverside, Marlborough A
Hosted By: American Economic Association
  • Chair: Peter H. Bils, Vanderbilt University

Polarization and Media Bias

A. Arda Gitmez
Bilkent University
Pooya Molavi
Northwestern University


This paper presents a model of partisan media trying to persuade a sophisticated and heterogeneous audience. We base our analysis on a Bayesian persuasion framework where receivers have heterogeneous preferences and beliefs. We identify an intensive-versus-extensive margin trade-off that drives the media's choice of slant: Biasing the news garners more support from the audience who follows the media but reduces the size of the audience. The media's slant and target audience are qualitatively different in polarized and unimodal (or non-polarized) societies. When the media's agenda becomes more popular, the media become more biased. When society becomes more polarized, the media become less biased. Thus, polarization may have an unexpected consequence: It may compel partisan media to be less biased and more informative.

Disclosure in Democracy

Matthew Denes
Carnegie Mellon University
Madeline Scanlon
University of Pittsburgh
Florian Schulz
University of Washington


Using hand-collected data on political contributions from undisclosed sources, we document novel stylized facts on "dark money" and its role in elections and politician type. Over the past decade, dark money has become a major source of campaign financing and currently comprises the largest source of capital from special interest groups. Consistent with evading disclosure, dark money is spent just before an election and often transferred to other special interest groups. We show that dark money is more likely to support candidates in competitive races and in areas with reduced information environments, lower education, greater inequality, and less poverty. Exploiting variation in exposure to television advertisements, we find that candidates supported by dark money advertisements receive an increase in votes and are more likely to win elections. While politicians supported by dark money organizations are more likely to engage in the political process by voting for and sponsoring legislation aligned with business interests, they are also more likely to be subsequently voted out of office, suggesting that they may enact an agenda focused on their donors rather than their constituents. Taken together, our results provide the first systematic evidence on the rise and impact of dark money in U.S. congressional elections, contributing to the ongoing debate about disclosure requirements of political spending.

Electoral Campaign Attacks: Theory and Evidence

Marcos Yamada Nakaguma
Sao Paulo School of Economics
Danilo Souza


This paper studies the determinants of electoral campaign attacks. We first propose a model to examine the main factors that influence candidates' decisions to attack. Our theoretical analysis yields a number of predictions which we test using information from ``right of reply" lawsuits filed in Brazil. Our empirical analysis exploits a regression discontinuity design based on virtual ties between 2nd and 3rd place candidates to show that candidates with an electoral advantage are more likely to receive an attack. We then exploit another discontinuity to show that the patterns of campaign attacks differ significantly under single and dual ballot plurality.

Identity-Based Elections

Helios Herrera
University of Warwick
Ravideep Sethi
University of Utah


We study electoral outcomes when citizens on either side of the political spectrum are exposed to different extents to mainstream news, as in the US case. Citizens gather information from mainstream news but also from possibly partisan media sources which filter news in particular predetermined ways. We assume citizens process all information they receive correctly, but choose their own media sources in a behavioral self-serving way to try to preserve their political faith/identity, namely attempting to rationally counteract mainstream news that they might view as unfavorable. This endogenous media choice generates an electoral advantage for the less exposed side, which can turn into a sure electoral victory even for the wrong candidate in a democracy. Results are robust to forms of media distrust and are stronger if citizens have biased priors. By contrast, in illiberal democracies where the government controls the media, we show how official media propaganda works only if citizens are unaware of it, but otherwise backfires entirely without censorship of other media.

Electoral Turnovers

Vincent Pons
Harvard Business School
Benjamin Marx
Sciences Po
Vincent Rollet
Massachusetts Institute of Technology


In most national elections, voters face a key choice between continuity and change. Electoral turnovers occur when the incumbent candidate or party fails to win reelection. To understand how turnovers affect national outcomes, we study the universe of presidential and parliamentary elections held since 1945. We document the prevalence of turnovers over time and we estimate their effects on economic performance, trade, human development, conflict, and democracy. Using a close-elections regression discontinuity design (RDD) across countries, we show that turnovers improve country performance. These effects are not driven by differences in the characteristics of challengers, or by the fact that challengers systematically increase the level of government intervention in the economy. Electing new leaders leads to more policy change, it improves governance, and it reduces perceived corruption, consistent with the expectation that recently elected leaders exert more effort due to stronger reputation concerns.
JEL Classifications
  • D7 - Analysis of Collective Decision-Making