Turnout bias in postelection surveys: political involvement, survey participation, and vote overreporting

Turnout bias is a classic flaw of postelection surveys. It has a double cause: overrepresentation of actual voters among survey respondents and vote overreporting by actual nonvoters. Previous studies have focused on the latter, measurement bias, and have failed to address the former, nonresponse bias. However, turnout is typically a topic where there may be a link between response propensity and measurement error, as factors related to political involvement are likely to influence both survey participation and misreport. To take this link into account, we resort to a Heckman selection model in which the outcome equation estimates the probability of overreporting conditional on the probability of participation in the survey. We apply such a model to a unique data set of validated votes from the canton of Geneva, Switzerland, informing about the actual turnout of both individuals who were interviewed and individuals who were sampled but could not be interviewed. Based on the same data, we put forward a solution that may correct for the nonresponse bias in cases where information regarding nonresponse is not available. Applying this correction to U.S. data shows that not correcting for response propensity may indeed lead to inaccurate estimates of the determinants of overreporting and, ultimately, of turnout