The use of proxy respondents in household surveys can be a source of information bias in epidemiological studies. This study assesses the agreement, reliability, magnitude, and direction of possible biases as well as the influence of proxy characteristics on the disagreements between self-report and proxy reports among 710 pairs of primary-secondary respondents. The influence of the proxy perspective on pair-agreement was also assessed, by comparing proxies' answers using their own perspective and that of primary respondents. Reliability was investigated in relation to lifestyle, health assessment, morbidity, and health care use. In general, the results showed good agreement and reliability. Disagreements were statistically associated with the following characteristics of proxy respondents: older age, lower schooling, and the proxy's knowledge of the primary respondent's health. Adults agreed more in evaluating the health of elderly when they used the elderly primary informant's perspective to respond. The results confirm the presence of bias when proxy respondents are used for hypertension and medical consultations.