What do people's estimates of smoking related risk mean?

Data is presented to show that ratings individuals make about the percentage risk or probability of dying of a smoking related illness, do not obey the laws of probability. Samples of smokers from four broadly representative population surveys were asked about the risks of dying of a smoking related illness using one of two similar questions with quite different response formats. The two formats produced markedly different estimates. Responses were also inconsistent (in terms of true probabilities) with other questions assessing risk. Risk estimates were lower when the risks of smoking were judged relative to other risks. Based on this finding, it is suggested that the majority of smokers underestimate rather than overestimate risk. It is also concluded that the majority of respondents do not respond to risk questions in terms of probabilities and that it is generally inappropriate to assume that they do.