We provide a framework to disentangle the role of preferences and beliefs in health behavior, and we apply it to compliance behavior during the acute phase of the COVID-19 pandemic. Using rich data on subjective expectations collected during the spring 2020 lockdown in the UK, we estimate a simple model of compliance behavior with uncertain costs and benefits, which we employ to quantify the utility trade-offs underlying compliance, to decompose group differences in compliance plans, and to compute the monetary compensation required for people to comply. We find that, on average, individuals assign the largest disutility to passing away from COVID-19 and being caught transgressing, and the largest utility to preserving their mental health. But we also document substantial heterogeneity in preferences and/or expectations by vulnerability status, gender, and other individual characteristics. In our data, both preferences and expectations matter for explaining gender differences in compliance, whereas compliance differences by vulnerability status are mainly driven by heterogeneity in preferences. We also investigate the relationship between own and others’ compliance. When others fail to comply and trust breaks down, individuals respond heterogeneously depending on their own circumstances and characteristics. When others around them comply less, those with higher risk tolerance and those without prior COVID-19 experience plan to comply less themselves, while the vulnerables plan to comply more. When a high-level public figure breaches the rules, supporters of the opposing political party plan to comply less. These findings emphasize the need for public health policies to account for heterogenous beliefs, preferences, and responses to others in citizens’ health behaviors.