We investigate the impact of health on working hours. This is in recognition of the fact that leaving the labour market because of persistently low levels of health status, or because of new health shocks, is only one of the possible responses open to employees. We use the first six waves of the Household, Income and Labour Dynamics in Australia (HILDA) Survey to estimate the joint effect of health status and health shocks on working hours. To account for zero working hours, we use a dynamic random effects Tobit model of working hours. We follow Heckman (1981) and approximate the unknown initial conditions with a static equation that utilises information from the first wave of the data. Predicted individual health status is used to ameliorate the possible effects of measurement error and endogeneity. We conclude that overall, lower health status results in fewer working hours and that when they occur, health shocks lead to further reductions in working hours. Estimation results show that the model performs well in separating the time-persistent effect of health status and the potentially more transient health shocks on working hours. Copyright © 2013 John Wiley & Sons, Ltd.