4/4 While we cannot definitively answer what causes this, something is happening that goes well beyond the scope of conventional measures of the socioeconomic and demographic factors of the places where voters live – and it is systematically and heavily clustering across space.

3/4 Also the share of catholics is kind of confusing. However, the main added value of this spatial analysis lies in uncovering a sort of "hidden" effect of geography. The elections were overwhelmingly(!) influenced by the local contexts of municipalities and their neighboring and nearby units.

2/4 To reveal what geography can tell us, we used a spatial error regression. The strongest contextual predictors appear to be education, previous voter turnout, and unemployment. Although the effect of unemployment is not as strong, it’s pretty standard stuff.