The UK has experienced substantial increases in both the number of claimants of work incapacity benefit and the proportion of claims for mental health reasons. However, identifying causality between experiencing mental health problems and claiming benefits is complex as claiming disability benefits may be the instigator or result of having a mental health problem. We use longitudinal data from the British Household Panel Survey (1991–2008) to disentangle the relationship between claiming benefits and mental health problems. We model the probability of claiming disability benefits using a dynamic random‐effects probit model, which reduces the endogeneity bias in modelling the effects of health on claiming disability benefits. Mental health has the largest effect of all health conditions modelled on the probability of disability benefit claiming. We find that the causal effect of mental health on claiming disability benefits is substantially over‐estimated by cross‐sectional associations. The effect decreases in magnitude from 12.5 to 0.5 percentage points for males and 8 to 0.7 percentage points for females. Policymakers should pay attention to other important influences on claiming disability benefits as well as interventions to improve the health of the working‐age population.