We undertook two surveys of older adults in Ontario to estimate the proportion in need of dental treatment. Because we expected low response rates, these studies were designed to assess 1) the effect of response enhancement strategies on participation, and 2) the extent of bias in estimates of treatment needs resulting from less than acceptable response rates. Our response enhancement strategies did not improve response rates substantially. In both surveys, there were significant differences in the characteristics of responders and nonresponders. Nevertheless, there was little difference in crude estimates of the prevalence of treatment needs and adjusted estimates taking account of non-response bias. We conclude that, while high response rates should always be the aim, low response rates do not necessarily compromise the results of descriptive epidemiological studies.