Around 1990 in France and 2005 in Germany, two new categories were introduced in the field of population statistics. Both “immigré” and “Person mit Migrationshintergrund” use the semantics of migration to describe a population group. My analysis shows that these two events reveal a shift in the meaning associated to migration in statistical categories in both countries. The meaning of the category changes from a mere description of mobility to the ascription of otherness within a population, a change linked to the so-called integration policy that is developing in France and Germany in the years 1990-2000. The analysis reveals the way in which statistics make migration socially relevant to the construction of otherness. The thesis thus develops a renewed approach to the interactions between statistical observation and public policy, by empirically testing the hypothesis of the “circularity of knowledge and action” developed by Alain Desrosières. These shifts in statistical nomenclatures can only be understood if we adopt a long-term perspective. It is necessary to look at the history of statistics and to compare how populations have been classified differently over time in order to understand the new categories invented at the turn of the 20th and 21st centuries. We therefore question the genesis and institutionalization of the categories of otherness and mobility in the period 1880-1914, during the formation of France and the German Kaiserreich as nation states and colonial empires. To observe these processes empirically, we chose to compare the practices of categorization of otherness and mobility (1) in France and Germany, (2) at two different time periods, 1880-1914 and 1990-2010, and (3) in the metropolitan as well as colonial context. The analytical foci in both periods are slightly different. For 1990-2010, the aim was to reconstruct the genesis of the categories “immigré” and “Person mit Migrationshintergrund”, whereas the analysis of the period 1880-1914 consists of uncovering historical ruptures and continuities in the classification principles over the entire period. The approach is neither symmetrical, nor chronological: it contrasts two historical configurations in an attempt to identify similarities and differences. Our results show that between 1880 and 1914, migration as a category was mostly associated with a mobility phenomenon in political and statistical discourse. At that time, the focus was on emigration, redefined as a geographical outward movement, i.e. people leaving the borders of the nation and the Empire. The transport of “emigrants”, a category of population that fueled public debate and statistical tables, was constructed as a political problem. Statistics on emigration were then separated from the observation of the composition of the population. In the metropolitan context, this happened through the introduction of the criterion of nationality and in the colonial context through “racial” patterns. In 1990 in France and 2005 in Germany, migration semantics were then used to statistically observe the composition of the population. Our results reveal three principles along which the construction of otherness takes place in the two countries and in the two time periods studied: a national principle, a colonial principle and a migratory principle. These findings remind us that statistics are never neutral but always steeped in their socio historical and political context – an important aspect to remember especially in times when migration policies are again a topic high on the agenda of public debates and policy makers.