Using Machine-Learning Techniques to Analyse how the South European Political Space Changed from 2014 to 2019

In this paper, we compare the electoral results of the European elections of 2019 in Spain, Portugal, Italy and Greece by using the data from the Voting Advice Application launched for this election, the EU-Vox 2019. We focus on the saliency of the European dimension, and try to answer to the following questions: 1) Are the European policy dimensions more salient than the economic and social ones? 2) How voters evaluate their distance in the European policy issues (proximity/directionality/unified model)? 3) Are there any substantial difference among South European countries? 4) Were there any fundamental changes since the last European Parliamentary Election? First, we develop a machine learning algorithm to be able to capture the political space of the voters in these elections and take some measures to increase the representativity of the sample and the external validity of our results. This system was already implemented in the EU-Vox 2014 with successful results. Second, we will study the saliency and the comparison logic of the political dimensions in each country. Third, we will compare the 2014 results with those obtained in 2019. We believe that VAAs, although they suffer some biases due to lack of representative sample, are ideal to carry out this research due to the possibility of implement learning algorithms in the data to properly capture the underlying political dimensions and their saliency and logic of comparison. We expect obtaining interesting results regarding how the saliency of the European dimension has changed in the last five years, whether there are substantial differences among South European countries and whether voters perceive their distances in the European political dimensions in a distinct way than those in the other political dimensions at stake.