Representativity and Measurement Errors in Surveys

Abstract

One of the main challenges for an accurate estimation of turnout measures based on survey data is representativeness. In fact, most recent studies show that non-response, i.e., that some groups of the population are less likely to respond to political surveys, even more strongly contributes to turnout bias in surveys than measurement errors . Another source of bias is undercoverage, which is even more rarely treated in empirical studies. The first aim of this Panel is to discuss questions, challenges and possible solutions related to the representativeness of the data: How “wrong” are we when measuring participation using for instance phone-based or online surveys, respectively? Can we improve representativeness through mixed-mode designs? In how far does a lack of representativeness affect inference based on survey data? The second aim of this Panel is to deal with questions like: How important is over-reporting and does it differ between different institutional or political contexts? What can we do to improve our surveys? What are advantages and disadvantages of official participation data? And, lastly, in how far do measurement errors impact on the substantial results we get from survey-based participation studies?