The Internet and Opinion Measurement: Surveying Marginalized Populations

Objective . Problems of self‐selection, selection bias, and response rates have greatly limited the use and validity of on‐line surveys. This study addresses those issues by examining population data for a group of Internet users who responded to a series of on‐line surveys, enabling the calculation of both response rate and selection bias. Methodology . A series of surveys were posted on a gay/lesbian website. We compare demographic data collected from our study sample with national data on gays/lesbians. A logistic regression model was used to determine if differences existed between participants and nonparticipants. Results . The study sample of gays/lesbians comported well with the national sample. Demographic characteristics of those electing to participate in the surveys and nonparticipants are practically indistinguishable. The response rate to our on‐line surveys was approximately 16.4 percent, similar to that in nontargeted mail surveys. Conclusions . The results indicate that, despite its limitations, the Internet can be a valuable medium in reaching populations difficult to identify using standard survey research techniques.