This paper explores the interrelation between health and work decisions of older workers. For this, two issues are of relevance. Firstly, health and work may be endogenously related because of direct causal effects of health on work and vice versa, and because of unobservables that may affect both observed health and work outcomes. Secondly, social surveys usually contain self-assessed health measures and research indicates that these may be subject to endogenous, state-dependent reporting bias. A solution to the ‘Health and Retirement Nexus’ therefore requires an integrated model for work decisions, health production and health reporting mechanisms. We formulate such a model and estimate it on a longitudinal dataset of older Dutch males. Copyright © 2009 John Wiley & Sons, Ltd.