Abstract
Survey-based research concerning sexual behaviour almost inevitably confronts the simultaneous problems of misreporting and non-response. These problems lead to disparities among estimates of the number and characteristics of those who engage in same-sex sexual behaviour. This paper proposes a statistical model to consistently estimate the frequency of same-sex sexual behaviour in the presence of non-ignorable misreporting and non-response. The model is fitted using 1991-2000 General Social Survey data. Frequency estimates corrected for simultaneous misreporting and non-response are reported. According to the model, 7.1% of US males and 4.1% of females - 15.8 million individuals - are not exclusively heterosexual. Allowing for misreporting and non-response increases the estimated same-sex frequency by more than four million. The model reveals new patterns between misreporting and non-response probabilities and standard demographic variables such as age and income.