We study changes in political discourse during campaigns, using a novel dataset of candidate websites for US House elections, 2002–2016, and manifestos for French parliamentary and local elections, 1958–2022. We find that candidates move to the center in ideology and rhetorical complexity between the first round (or primary) and the second round (or general election). This convergence reflects candidates' strategic adjustment to their opponents, as predicted by Downsian competition: Using an RDD we show that candidates converge to the platform of opponents who narrowly qualified for the last round as opposed to those who narrowly failed to qualify.

We leverage plausibly exogenous variation in regional exposure to corruption to provide causal estimates of the impact of local political corruption on terrorist activity for a sample of 175 countries between 1970 and 2018. We find that higher levels of corruption lead to more terrorism. This result is robust to a variety of empirical modifications, including various ways in which we probe the validity of our instrumental variables approach. We also show that corruption adversely affects the provision of public goods and undermines counter-terrorism capacity. Thus, our empirical findings are consistent with predictions from a game-theoretical representation of terrorism, according to which corruption makes terrorism relatively more attractive compared to peaceful contestation, while also decreasing the costs of organizing and carrying out terrorist attacks.