Measuring public support for European integration from population-level data using a Bayesian IRT model

This study proposes the use of Bayesian item response theory (IRT) models to measure public preferences towards Europe. IRT models address the limitations of single-question indicators and produce valid estimates of public latent attitudes over long time periods, even when available indicators change over time or present interruptions. The approach is compared with an alternative technique recently introduced in the study of EU public opinion, the Dyad Ratios algorithm. It shows that IRT models can both incorporate a more theoretically grounded individual-level model of response and produce more accurate estimates of latent public support for Europe. The measure is validated by showing its association both to alternative measures of EU support and to the vote share of Eurosceptic parties across Europe.