We utilized a unique dataset of Montreal residents to estimate the rel
ationship between employment and mental health, controlling for endoge
neity. We applied a maximum likelihood, simultaneous equation generali
zed probit model to estimate jointly the determinants of an individual
's latent index of employability and their mental health as measured b
y the Psychiatric Symptom Index (PSI). The likelihood function was adj
usted to account for the fact that individuals were sampled based on t
heir employment status, and also for the fact that repeated observatio
ns of individuals in different periods were used in the analysis. We f
ound tangible beneficial effects of mental health on employability. In
addition, employment appears to improve mental health. The ML estimat
es of the endogenous relationship between employment and mental health
indicate that OLS estimates are biased upwards, but the effects of un
employment on deteriorating mental health are not spurious. (C) 1997 b
y John Wiley & Sons, Ltd.