Are Web Surveys Better at Measuring Political Attitudes and Voting Behavior?

• In Event: Measurement in Survey Research

Abstract

CRONOS panel has the unique feature of a cross-country web panel, which was designed using as 'sample frame' the respondents to the European Social Survey (ESS). Three countries participated in CRONOS, the United Kingdom, Slovenia, and Estonia. The aim of this paper is two-fold. It firstly explores the methodological issues related to drawing a web panel from the respondents to a nationally representative face-to-face survey, and then it focuses on how reporting political attitudes and voting behavior depend on survey mode and question wording. The first part of the paper analyses whether the profile of CRONOS respondents is similar to the one of ESS respondents. Selection probabilities and non-response patterns are studied within- and between- the three countries included in CRONOS. Although sampling design data will be used, priority will be given to the latent characteristics of respondents, their values and attitudes, as they can be extracted from the ESS Main Questionnaire, with the scope of identifying what drives people's decision to participate in a web survey. Alternatively and where possible, comparisons are made with other web (sub-)panels drawn from face-to-face/telephone surveys. The second part of the paper examines if political attitudes and voting behavior are reported differently online compared to a face-to-face CAPI interview by CRONOS versus ESS respondents. As CRONOS panel is literally extracted from the ESS respondents, the paper also explores how the various layers of nonresponse hidden in CRONOS web-panel affect the accuracy of point estimates related to reported voting behavior. Once an appropriate non-response adjustment has been used, the paper makes use of the various experiments incorporated in CRONOS survey and tests how gendered views to politics can be revealed by question wording. The paper also screens whether modelling voting behavior using ESS data can be transposed on data collected from the web panel, when using the same predictors. The third part of the paper discusses the results and the expected implications of using web-surveys in political science. The paper contributes to political methodology literature by providing new insights regarding the challenges of analyzing web-survey data regarding political attitudes and voting behavior, when the sample frame of the web-panel is strictly the group of respondents to a nationally representative face-to-face survey.

Author

• Gina H. Anghelescu, University of Sydney