with Dr. Gabriel Rossman, Associate Professor of Sociology, UCLA
Repeated cross-section opinion surveys like GSS and ANES hold out the promise that we can not only measure trends in particular questions but patterns among question, such as consensus. While some issues, such as abortion, are perennially contentious, other issues that are initially controversial achieve consensus, and conversely, some consensus issues become controversial. The desire of survey researchers to maintain comparability over time is irreducibly in tension with the desire to capture emerging controversies rather than flog the dead horse of now obscure issues that were highly salient in wave one. We use a simulation to show that whether responses change by secular trends or random walks, question selection that maximizes variance in the initial wave of a survey will lead to a measurement artifact of decreasing controversy in subsequent waves. We further show that the GSS tends to drop questions as the variance they yield declines, but also that mean variance is higher if using all questions in a wave than if only using a stable set of perennial items. We note that this issue should generalize to any situation where a small number of measurements are used to capture an underlying shifting reality with many dimensions.
Where: SNF Agora Conference Room and Zoom