Questionnaire surveys are often used in epidemiology and survey research. However, investigations have shown that the method suffers from biases. If the intention is to examine differences between subgroups in a sample, studies have shown that people tend to give answers to questionnaires more according to a social norm than to the actual situation. This has been called social desirability. The aim of this study was to investigate response bias in subgroups of a sample in a large questionnaire survey (n = 9,200) and to study whether social desirability has an impact on survey results in dental research. The answers were divided into subgroups according to gender, age, and residence and were compared with corresponding data from dental insurance claims. The levels of agreement were found to vary considerably and the differences were highly statistically significant. The variation in agreement showed that bias on the group level increased when the distance from an anticipated social norm was larger. This implies that the divergence from a socially desirable mode of action influences the magnitude of the bias.