To the Usefulness of Multi-Level Regression in Comparative Politics

The regression is the most common data analysis technique in political science. It allows measuring the net effect of an explanatory variable on an explained variable while controlling for the disturbing effect other variables. One of its sub-technique, i.e. the multi-level regression, is dedicated to the analysis of databases regrouping information from different countries such as survey responses from various European population samples or electoral results from various elections throughout the world. The multi-level regression is however underexploited in the field of francophone comparative politics. In most of quantitative analyses published in the best journals of the discipline, classic regressions are simply run in several countries and their results are then compared. This article introduced two types of multi-level regression : fixed-effect and random-effect regressions. These regressions allow meeting two fundamental goals of francophone comparative politics, i.e. the quest of universality and the rejection of culturalist explanations.