Mental health and employment: a bounding approach using panel data

The disability employment gap is an issue of concern in most Western developed economies. This paper provides important empirical evidence on the influence of mental health on the probability of being in employment for prime age workers. We use longitudinal data and recently developed techniques, which use selection on observable characteristics to provide information on selection along unobservable factors, to estimate an unbiased effect of changes in mental health. Our results suggest that selection into mental health is almost entirely based on time-invariant characteristics, and hence fixed effects estimates are unbiased in this context. Our results indicate that transitioning into poor mental health leads to a reduction of 1.6 percentage points in the probability of employment. This is approximately 10 per cent of the raw employment gap. This effect is substantially smaller than the typical instrumental variable estimates, which dominate the literature, and often provide very specific estimates of a local average treatment effect based on an arbitrary exogenous shock. These findings should provide some reassurance to practitioners using fixed effects methods to investigate the impacts of health on work. They should also be useful to policy makers as the average effect of mental health on employment for those whose mental health changes is a highly relevant policy parameter.