France in ten slices: An analysis of aggregate data

Many causal relationships cannot be detected by national averages, whether based on regression analysis of aggregate data or on survey data, because the local social context explains an important part of the variance. An analysis by deciles of aggregate data shows the great territorial diversity of social contexts in France. The unit of analysis is the French canton. The 2,450 cantons are distributed by deciles of such variables as religious practice, social structure, and vote. In the second part, six extreme deciles are analysed: the most leftist part of France, the most rightist, the most industrialized, the most agricultural, the most devout, and the most irreligious. For a country with such a territorial diversity, generating a polymorphism of parties, the national averages based on aggregate data or on survey data are largely fallacious, particularly those concerning the causal relationships. Slicing in deciles is the most efficient method for analysing aggregate data