It has been demonstrated that in most countries electoral behaviour follows a pattern of nationalization, that is, national politics substitute local politics. Most research (both comparative and single case study) analyses the phenomenon of nationalization and its strength by focusing on a single type of election: the legislative/parliamentary election. When dealing with comparative research, this implies that the electoral offer and the context differ from one country area to another or within one country. The original contribution of this article is that it offers both a new methodological and substantive perspective. From a methodological point of view, we propose a quasi-experimental design. Nationalization will be measured for two types of elections (presidential and legislative), in the same country (France), at the same territorial level (departments), for a long span of time (the Fifth Republic), covering 9 presidential and 14 legislative elections. For this analysis, the nationalization of the electoral behaviours will be calculated by employing first the standard deviation, the Mean of Absolute Deviation and then the Bochsler index (2010). The Bochsler index allows us to measure nationalization at the country level, and is not sensitive to the size and the number of parties. Therefore, the Bochsler index is the best instrument when dealing with a comparative or a longitudinal design. From a substantive point of view, this article aims to demonstrate that the degree of nationalization depends on the type of election. In our case, which is a semi-presidential system, we expect to observe a higher degree of nationalization in presidential elections compared with legislative elections. In fact, with regard to legislative elections the local dimension of the vote (department level) should lead to a higher fragmentation of electoral behaviour. [ABSTRACT FROM AUTHOR]