Measurement Invariance of Political Efficacy: The Case of the European Social Survey

Details

Supervisors

Ritter, Christian ; Baudewyns, Pierre

Faculty

Faculté des sciences

Degree label

Master [120] en statistique, orientation générale, à finalité spécialisée

Abstract

The European Social Survey is a cross-national survey organised in Europe every two years. The survey gives a great deal of importance to the methodological aspects of collecting high quality data with which longitudinal and cross-national comparisons can be made. One important concept in political sciences, that is studied in the ESS, is political efficacy defined as the citizen’s evaluation of his competence in the political sphere as well as his evaluation of the responsiveness of the political system to his demands. Researchers have demonstrated that the political efficacy concept is linked to political participation. Measurement invariance analysis checks if a latent variable is measured in the same way in different groups. In the ESS, the measurement invariance of the political efficacy has never been tested for the last three rounds of the questionnaire. The lack of study of this subject means that scientists who have done cross-national comparisons on political efficacy with the ESS data have made conclusions without verifying that the concept is measured similarly across the different cultures. This research has analysed the measurement invariance characteristics of political efficacy in the rounds 8 to 10 of the ESS. Two different types of analysis have been produced. The first one has checked if it is valid to compare latent score of one country between the different rounds. The second type of analysis has consisted in checking if in one round, it is valid to compare the latent score of the different countries. The first analysis that verified that it is correct to study the evolution over time of political efficacy in each country has shown that scalar invariance was reached for every country except Austria and Poland. This means that it is valid to compare directly differences between latent scores in the different rounds for every country except two. The cross-national analysis by rounds has demonstrated that scalar invariance was not achieved in any of the three rounds. This means that differences observed between countries in one round should not be directly compared as part of the observed differences is probably an artificial difference caused by a bias in the questionnaire. Finally, the lack of scalar invariance has been explored. The invariance has been represented graphically to let the reader evaluate the level of non-invariance that was present in the data for each round. The lack of invariance was small. By looking at the invariance graph and the latent means graph, it was possible to see that the invariance between countries was different between sub-groups of countries and that some sub-groups could have achieved scalar invariance.