Do congruence estimates derived from different sources converge? A growing number of scholars are studying party–voter congruence in a comparative perspective, increasingly focusing on specific issues that depart from the general left–right dimension. When doing so, however, they often encounter a missing-data problem: Researchers must combine citizen preferences and party positions derived from distinct sources, frequently presenting different question formats, wordings, and scales. This exacerbates perceptual biases and requires demanding assumptions regarding the common understanding of issues between voters and experts. While multiple techniques have been proposed to address these issues, they remain burdensome and rarely used. Consequently, this article focuses on the most widely employed approaches to estimating party–voter issue congruence. I confront measures constructed by matching European Election Study voter preferences with expert (Chapel Hill Expert Survey) and Euromanifesto party position estimates against those generated by voting advice applications (VAAs)—specifically the euandi dataset—which present the benefit of measuring parties and voters on the same scale, with the same wording. Focusing on 60 party–voter dyads across 12 Western European countries in 2014, the results indicate clear divergence between estimates derived from manifesto data and those from VAAs. Correlations between VAAs and experts are consistently positive and significant, though moderate. These estimates do not seem to be driven by different sampling methods or quality filtering. Finally, an exploratory analysis on the sources of divergence finds that manifesto length, expert uncertainty, and party newness help explain some of the differences between congruence estimates.
Mattia Gatti (Wed,) studied this question.