Monitoring AIDS and other rare population events: a network approach

This paper replicates and extends an earlier attempt to use data from the General Social Survey (GSS) to track the distribution of AIDS across demographic subgroups. (The GSS asks respondents whether they know a person with AIDS [PWA].) The gender, racial, age, and regional composition of the set of PWAs reported by GSS respondents is compared with that of the official AIDS cases reported to the Centers for Disease Control (CDC). In an attempt to assess the accuracy of the GSS estimates, a similar analysis is performed in which GSS respondents are asked whether they know a homicide victim. Data from four consecutive GSS samples (1988, 1989, 1990, and 1991) are used, permitting a more detailed exploration of potential biases and problems with the network technique. In addition, time series data from the National Health Interview Survey on the percentage of people who know at least one PWA are used to validate the GSS data. Our earlier findings, that the GSS identifies proportionately more White and midwestern cases than are reported to the CDC, are corroborated by the additional data. Possible explanations for these discrepancies are given, and suggestions are made for improving the utility of the approach.