Are people under economic stress more or less likely to vote, and why? With large observational datasets and a survey experiment involving unemployed Americans, we show that unemployment depresses participation. But it does so more powerfully when the unemployment rate is low, less powerfully when it is high. Whereas earlier studies have explained lower turnout among the unemployed by stressing the especially high opportunity costs these would-be voters face, our evidence points to the psychological effects of unemployment and of campaign messages about it. When unemployment is high, challengers have an incentive to blame the incumbent, thus eliciting anger among the unemployed. Psychologists have shown anger to be an approach or mobilizing emotion. When joblessness is low, campaigns tend to ignore it. The jobless thus remain in states of depression and self-blame, which are demobilizing emotions.

In most democracies, citizens who identify with a political party are more likely than non-partisans to turn out to vote. But why is this the case? Does voting foster partisanship, as prominent models of political learning and cognitive dissonance postulate? Or does partisanship encourage voting, as expressive voting models and social identity theory suggest? I introduce the concept of partisan duty to capture the role of partisan social identities in the turnout decision and present new empirical tests of the relationship between partisanship and voting. I leverage a unique institutional arrangement in Chile to establish the direction of causality with a regression discontinuity, and I implement a novel survey design with behavioral outcomes to identify causal mechanisms. Data from the US confirm that the main findings generalize beyond Chile. Electoral participation does not generate partisanship. Instead, partisanship mobilizes voters: it increases the expressive benefits to voting and generates a sense of duty to support one’s partisan group.