(joint with Matthew Shum)
We study the relationship between social media and information aggregation by voting. Our experimental treatments mimic the features of social networks in the presence of media bias in order to address concerns that voters obtaining their political news via social media may become more polarized in their voting behavior. Our results suggest substantial effects of polarization at the expense of efficient information aggregation: subjects engage in “liking” behavior by selectively sharing the news that is favorable to their party and downweighting unfavorable news in their voting decisions. At the same time, social networks raise collective decision making efficiency, even in the presence of media bias. All in all, subjects behave as if information sharing and voting is expressive of their party preferences even when (by construction) their preferences have a common value component. We supplement lab results with survey evidence from Pakistan, and find that it is broadly consistent with these patterns.