Household surveys often require including proxy reporters to obtain information about other household members who cannot be interviewed. The participation of proxies can undermine survey data quality due to the fact that proxies must respond to questions thinking about other people. The objectives of the present study were to analyze the behaviour of proxy reporters and evaluate the convergence between the answers given by proxies and self-reporters by means of behaviour coding. This improves the evaluation of convergence, since only adequate (i.e., interpretable) answers given by both types of informant are taken into account. Responses to a disability questionnaire employed by an official statistical institute were analyzed. The questionnaire includes 11 questions about different limitations related to everyday activities. 16 self-reporter and 16 proxies formed 16 couples whose members lived together and supported a direct family relation. The results show a high percentage (52%) of convergence between both types of informant, although fluctuating across the questions and the couples. Proxies showed relatively more adequate behaviour during the interaction than self-reporters. From this we conclude that proxies can be considered at least as good informants as self-reporters from an interviewer-respondent interaction perspective. Future research should address the impact of proxy responses on survey validity.