The aim of this paper was to evaluate the accuracy of data on death certificates for occupation and main cause of death. Measure of agreement was assessed comparing data from death certificates with those from both medical records and next-of-kin interviews, analyzing information for 552 residents of Botucatu, Southeast Brazil, who died in 1997. Kappa coefficients of 0.31 (95% C.I. 0.29-0.34) and 0.76 (95% C.I. 0.75-0.76) were obtained for data on occupation and main cause of death, coded by a Brazilian two-digit classification and the three-digit ICD-10 classification, respectively. One can conclude that, although quality of the main cause of death is acceptable for pilot studies, data on occupation taken only from death certificates is not accurate enough to be used in epidemiological research.