All 75-year-olds born in 1914 and living in the city of Jyväskylä, central Finland (n=388) were invited to study the predictive value of exercise test for mortality. Subjects who entered the laboratory (n=295) were to have a standard pre-test evaluation and perform a cycle ergometer exercise test. Subjects with complete background, exercise-test status and mortality data (n=282) were divided into three groups according to exercise-test status: a non-exercise test group (n=79), an exercise-test termination group (n=95), and an exercise-test completion group (n=108). Mortality was followed up for 9 years. The multivariate hazard ratio (HR) for death among the non-exercise test group compared with exercise-test completion group was 1.87 (CI 1.19–2.94). The multivariate HR for death among the exercise-test termination group compared with the exercise-test completion group was 0.95 (CI 0.58–1.54). High cycling power (W/kg body weight) in the exercise-test completion group was associated with a decreased risk for death with a multivariate HR 0.14 (CI 0.05–0.38). Performing an exercise test serves information on the risk of death that is incremental to clinical data and traditional risk factors of death in elderly people.