Assessment of non-response bias in a probability household survey of male same-gender sexual behavior

Objective. To assess non-participation bias in a survey ofmale sexual behavior. Material and methods. A householdsurvey was carried out in 1992-1993 using a probabilitysampling frame in Mexico City. Demographic variableswere available for all eligible men. The extent of non-participationbias was estimated using a version of the Heckmanmethod, which utilizes two equations, one to predict participationand the other to predict reports of same-gendersexual behavior. Results. A total of 8 068 of the 13 713eligible men completed a face-to-face questionnaire (responserate 59%); 173 men (2.1%) reported bisexual behaviorin their lifetime, and 37 (0.4%) reported only malepartners. Survey participation was predicted using demographicvariables: 67% of the observations were correctlypredicted by a probit regression model: 82% of participantsand 53% of non-participants (pseudo-r2=0.13). Same-gendersexual behavior was predicted by variables indicating attachmentto gay/bisexual social networks, history of sexuallytransmitted diseases, positive attitudes towards gay and bisexualmales, and lack of support from male relatives. Ninety-seven per cent of the cases was correctly predictedby the probit model (pseudo-r2=0.14). The correlation betweenthese two equations was not statistically significant.Conclusions. These results indicate that prevalence esti-of same-gender sexual behavior among Mexico Citymen were not biased by selective survey participation. Carefulselection and training of household interviewers mayhave assisted in minimizing potential bias.