Regional Difference‐in‐Differences in France Using the German Annexation of Alsace‐Moselle in 1870–1918

Abstract

The evaluation of labor market policies has been an expanding field in the last decades. This is partly due to the increasing availability of survey data and computing power and partly to the recognition that even complex phenomena such as the impact of laws and regulations on labor markets can be rigorously tested thanks to new empiricalmethodologies, such as the so‐called difference‐in‐difference approaches (also called double differences). In this approach, the outcome (education, access to employment, unemployment rate) of a treated group, that is, a group subject to a policy change, is compared to that of a control group, that is, a group made up of individuals (or any other unit of observation) as close as possible to the treated group but unaffected by the treatment. The strategy of the researchers consists in finding a sudden change in policy and building a relevant comparison group. A recent paper by Imbens and Wooldridge (2008) discusses the issues associated with the choice of the groups and surveys the literature. In countries with a federal structure such as the United States or Canada, it is straightforward to use this technique to evaluate policy changes. Indeed, many laws are specific to a state (for the United States) or a province (for Canada). When states experiencing a reform are compared to states with no change, it is straightforward to obtain inferences about the causal effects of the reform. In contrast, researchers studying French labor laws typically face cases in which there is no geographical variance in policy changes since the main law is supposed to apply equally to all the French territory. Therefore, policy evaluation has to rely on other control groups. A consequence of this “universal”character of the law is that the evaluation of the reduction in working time in France—from