Using data from the Health and Retirement Study (HRS) and the English Longitudinal Study of Ageing (ELSA), we estimate a dynamic model of health and employment. We estimate how transitory and persistent health shocks affect employment over time. In a first step we formulate and estimate a dynamic model of health. The procedure accounts for measurement error and the possibility that people might justify their employment status by reporting bad health. We find that health is well represented by the sum of a transitory white noise process and a persistent AR(1) process. Next, we use the method of simulated moments to estimate the employment response to these shocks. We find that persistent shocks have much bigger effects on employment than transitory shocks, and that these persistent shocks are long lived. For this reason employment is strongly correlated with lagged health, a fact that the usual cross sectional estimates do not account for. We also show that accounting for the dynamics of health and employment leads to larger estimates of the effect of health on employment than what simple OLS estimates of health on employment would imply. We argue that the dynamic effect of health on employment could be generated by a model with human capital accumulation, where negative health shocks slowly reduce the human capital stock, and thus gradually causes people to exit the labor market.