UC Berkeley

For more than fifty years, political scientists have questioned whether citizens are sufficiently informed about politics to ensure representation and democratic accountability. On the one hand, surveys suggest that most voters are poorly informed about politics and policy, have inconsistent policy preferences, and suffer from deep-seated biases that hinder rational political judgment. On the other, aggregate studies of representation and collective opinion have produced evidence that voters do hold officials accountable when they fail to represent their interests, and that public opinion tends to reflect national events.

These micro and macro patterns present something of a puzzle: how can political accountability exist when overall levels of citizen competence are low? More recent research has argued that while citizens are not encyclopedias of political knowledge, they can often learn what they need to know to make rational choices. These studies focus on the dynamics of political learning and the role of the information environment in providing voters with useful cues and policy information. When examined through this lens, citizens often appear more capable than was previously thought.

But important questions about information effects and political learning remain. Some studies find that citizens will update their beliefs in predictable ways when they learn something new. Others find that political misperceptions are hard to erase, even in the face of efforts to correct them. Under what conditions will citizens learn from new political information? Are some messages more potent than others? And do simple cues always help voters navigate the political world?

This dissertation contributes to these debates by examining information effects across three different venues: electoral politics, attitudes toward government programs, and social policy markets. Using three different data sources, I explore whether the political environment helps citizens to become more informed and whether new information prompts them to revise their preferences and perceptions in rational ways. I begin with intuitions from a rational learning model that help to explain the divergent findings in the literature and serve as a guide for the empirical chapters that follow.

I find considerable evidence of information effects across all three venues. Citizens in the context of an election campaign are more informed about the party identification of their senator, and this party cue helps them to make inferences about the senator's issue positions. Importantly, however, this increase in accuracy is only evident on votes where the senator voted the party line. When senators defected to vote with the other party, citizens' reliance on party cues led them astray. Similarly, I show how providing citizens with information about the performance of social programs like Food Stamps and Head Start has a significant effect on their perceptions and preferences. Negative performance information appears to be more potent than positive, but the findings show no evidence of ideological bias. Finally, I find that citizen-consumers in the market for public colleges use additional information about student outcomes to update their preferences.

The dissertation concludes by returning to the rational learning model in an effort to reconcile these findings with those in the existing literature. I argue that the divergent findings are largely due to different assumptions about the precision of new political information, the strength of prior beliefs, and the probability that citizens receive political messages.