Using longitudinal data from the Canadian National Population Health Survey (NPHS), we study the relationship between health and employment among older Canadians. We focus on two issues: (1) the possible problems with self-reported health, including endogeneity and measurement error, and (2) the relative importance of health changes and long-term health in the decision to work. We contrast estimates of the impact of health on employment using self-assessed health, an objective health index contained in the NPHS – the HUI3, and a ‘purged’ health stock measure. Our results suggest that health has an economically significant effect on employment probabilities for Canadian men and women aged 50–64, and that this effect is underestimated by simple estimates based on self-assessed health. We also corroborate recent US and UK findings that changes in health are important in the work decision.