Mapping Issue Salience Across Europe

What are the most important issues for electorates across Europe? And in which electoral arenas dominate specific political issues? This paper sets out to map issue salience across Europe’s multi-level governance system. This is a highly relevant endeavor as issue salience is a crucial factor to understand political processes across Europe. The recent Brexit vote, for example, demonstrates that the local salience of immigration or economic development issues substantially shape voting behavior. Hence, the main goal of this research is to adequately measure the most important issue salience divides across Europe. For this, I use a unique dataset on regional preferences for 205 EU regions from the EU-28, which allows respondents to select from a set of most important issues. Unsupervised text mining techniques on these half-open issue salience questions allow me to identify the most discriminating issues, while retaining as much as possible of the original variation in issue salience. I find substantial heterogeneity in issue salience across member states and regions. In fact, 40 percent of variation in issue salience is on the regional level, leaving a lot of variation the country level cannot account for. Strongest divergence across European regions appears on an immigration/crime/environment – unemployment/economic performance-dimension. This divide can be explained by differences in regional wealth and immigration exposure, closely mirroring a center-periphery logic. In most member states, voters in urban centers are more concerned about crime, immigration, and the environment, while the peripheral, rural regions are more concerned about economic performance. Moreover, issue salience configurations vary across member states. For example, healthcare is a more important factor to describe issue salience in Southern Europe, giving rise to a healthcare – economic performance-dimension and a healthcare – immigration-dimension. To test the validity of my results, regression analyses demonstrate that the derived salience dimensions are substantially related to a number of covariates in the expected direction. I conclude with suggestions on how to use the estimated salience dimensions for future research.