Leaders’ approval ratings provide a continuing referendum on their actions, conditioning future decision-making. Despite consensus on the drivers and cyclical dynamics of approval of executives in normal times, research on public response to leader performance during crises remains scattershot and incomplete. Seeking a more general theory, we argue that mass publics exhibit an initial, affective response to crisis threats that results in a rally of executive support, especially when the threat is perceived as external. Affective rallies tend to dissipate quickly as mass publics gain more comprehensive perspectives and, in turn, evaluate government performance more rigorously. Based on our arguments, we expect international conflict crises to generate rally effects, but not domestic terrorist attacks. Similarly, we expect that infrequent natural disasters should generate modest rally effects, but not when they occur in quick succession. And with the important exception of COVID-19’s global threat, we do not expect public health crises to generate rallies. We test these expectations in two ways. Observationally, we leverage an original crisis database consisting of security (interstate conflict and terrorism), natural disasters, and public health crises for 36 countries paired with monthly time series estimates from the Executive Approval Database 3.0. Our observational analyses largely confirm these expectations. Experimentally, we propose two designs to help unpack our theoretical mechanisms.