Domestic abuse is a common yet complex phenomenon. We use the universe of police reports filed in Finland since 2006, matched to rich register data, to identify the impact of cohabiting with an abusive spouse on labour market outcomes. We show that women’s labour market outcomes worsen significantly upon cohabitation, even if physical violence is only reported a number of years later. Using a within-individual across-relationship design and a matched control event study design, we rule out that this deterioration is due to negative labour market shocks that cause women to enter into bad relationships or due to income effects deriving from cohabitation. We build a model of coercive control that rationalises these findings. In the model, women have imperfect information about their partner’s type and abusive spouses have an incentive to strategically suppress women’s labour supply to reduce her outside option. The model yields new insights into the role of break-up for determining exposure to abusive partners. Taking the predictions of the model to data leads to a policy relevant reinterpretation of previous findings about the relationship between women’s outside options and the prevalence of domestic violence.