We reassess the impact of health on retirement plans of older workers using a unique survey-register match-up which allows comparing the retirement effects of potentially biased survey self-reports of health to those of unbiased register-based diagnostic measures. The aim is to investigate whether even for narrowly defined health measures a divergence exists in the impacts of health on retirement between self-reported health and objective physician-reported health. Our sample consists of older workers and retirees drawn from a Danish panel survey from 1997 and 2002, merged to longitudinal register data. Estimation of measurement error-reduced and selection-corrected pooled OLS and fixed effects models of retirement show that receiving a medical diagnosis is an important determinant of retirement planning for both men and women, in fact more important than economic factors. The type of diagnosis matters, however. For men, the largest reduction in planned retirement age occurs for a diagnosis of lung disease while for women it occurs for musculo-skeletal disease. Except for cardiovascular disease, diagnosed disease is more influential in men's retirement planning than in women's. Our study provides evidence that men's self-report of myalgia and back problems and women's self-report of osteoarthritis possibly yield biased estimates of the impact on planned retirement age, and that this bias ranges between 1.5 and 2 years, suggesting that users of survey data should be wary of applying self-reports of health conditions with diffuse symptoms to the study of labor market outcomes. On the other hand, self-reported cardiovascular disease such as high blood pressure does not appear to bias the estimated impact on planned retirement. Copyright © 2009 John Wiley & Sons, Ltd.