Selective under-ascertainment of cancer patients by cancer registries may bias population-based estimates of cancer patient survival. As prognosis is strongly age related for many forms of cancer, age adjustment of cancer survival rates, which is primarily used to enhance the validity of comparative analyses between cancer populations, might also effectively reduce this type of bias. We empirically assessed this potential "side-effect" of age adjustment using data from the Finnish Cancer Registry. Analyses of five-year absolute and relative survival for patients diagnosed in 1990-1994, and age adjustment to the age structure of patients diagnosed in 1985-1989, were used as examples. Various patterns of selective under-ascertainment were simulated, and the bias in crude and age adjusted five-year survival rates was compared. Age adjusted estimates were less biased in most scenarios, which may be an additional argument for application of age adjustment in the analysis and reporting of population-based cancer survival rates.