Abstract There is an enduring debate about whether, to what extent, and along which fault lines citizens’ policy preferences are ideologically structured. Some scholars maintain that public opinion is largely unstructured, with individuals adopting inconsistent and idiosyncratic positions – for instance, holding a ‘left’ stance on issue A does not necessarily imply a ‘left’ stance on issue B. Others argue that citizens’ views are shaped by coherent ideological constraints, such that a ‘left’ position on one issue is systematically accompanied by ‘left’ positions on others. This paper contributes new evidence to this debate by leveraging unprecedented big data. We draw on original data from Belgium’s widely used Voting Advice Application (VAA), known as the Vote Test , which was completed more than six million times in the run-up to the June 2024 elections. Our analysis is based on the actual log files of the application, encompassing millions of observations. The Vote Test consisted of twelve distinct VAAs designed for the seven concurrent elections held in Belgium in 2024 – including simultaneous federal, regional, and European contests across the country’s three regions – yielding millions of responses to hundreds of policy statements. Using these exceptional data, we examine the correlations among citizens’ answers and assess the dimensionality of the opinion landscape. We further compare the structure of mass opinion with that of political elites, who responded to the identical set of policy statements. Our findings reveal minimal, if any, ideological structuring among voters, especially when contrasted with the more consistent patterns observed among elites.
Mongrain et al. (2026) studied this question.