Optimal Matching and Social Sciences

Les documents de travail ne reflètent pas la position de l'INSEE et n'engagent que leurs auteurs. Working papers do not reflect the position of INSEE but only the views of the authors. 1 This working paper is a reflection on the conditions required to use optimal matching (OM) in social sciences. Despite its striking success in biology, optimal matching was not invented to solve biological questions but computer science ones: OM is a family of distance concepts originating in information and coding theory were it is known under various names among which Hamming, and Levenshtein distance. As a consequence, the success of this method in biology has nothing to do with the alleged similarity of the way it operates with biological processes but with choices of parameters in accordance with the kind of materials and questions biologists are facing. As materials and questions differ in social sciences, it is not possible to import OM directly from biology. The very basic fact that sequences of social events are not made of biological matter but of events and time is crucial for the adaptation of OM: insertion and deletion operations warp time and are to be avoided if information regarding the social regulation of the timing of event is to be fully recovered. A formulation of substitution costs taking advantage of the social structuration of time is proposed for sequences sharing the same calendar: dynamic substitution costs can be derived from the series of transition matrices describing social sub-rhythms. An application to the question of the scheduling of work is proposed: using data from the 1985-86 and 1998-99 French time-use surveys, twelve types of workdays are uncovered. Their interpretability and quality, assessed visually through aggregate and individual tempograms, and box plots, seem satisfactory. Ce document de travail se veut une réflexion sur les conditions d'utilisation des méthodes d'appariement optimal en sciences sociales. En dépit de ses succès retentissant en biologie, l'appariement optimal n'a pas été inventé pour résoudre des problèmes en biologie mais en informatique et en théorie du codage où il est connu notamment sous les noms de distance de Hamming et de Levenshtein. Par conséquent, le succès de cette méthode en biologie n'a rien à voir avec la possible ressemblance de son mode opératoire avec des processus biologiques mais provient de choix de paramètres cohérents avec le type de matériel et de questions auxquels les biologistes font face. Parce qu'en …