In this paper, we study the effects of household shocks on the incidence of domestic violence (DV) using a unique set of microdata from the World Bank’s Living Standard Measurement Survey for Tanzania. We use idiosyncratic variation in rainfall as an exogenous shock to Tanzanian households and control for a large set of potential confounding variables on the individual, household and community levels, while exploiting intra-and inter-community rainfall variation for identification. We find that rainfall shocks substantially increase the likelihood of the DV incidence in the household. A one standard deviation negative rainfall shock increases the incidence of domestic violence by about 18.8 percentage points compared to baseline for wives. We furthermore show that rainfall shocks have an effect on physical violence, while we do not find an effect on severe physical or sexual abuse, which is consistent with the strategic use of violence. Estimates from non-linear specifications reveal that the overall effects are driven by droughts rather than floods. We furthermore show that effects are more pronounced for poorer households. In addition, we also provide evidence that female empowerment mitigates the impact of rainfall shocks on violence.