BLANC - Blanc

Econométrie des redoublements – graderepetition

Submission summary

Available empirical studies of grade repetitions are rather naive as far as statistical modelling is concerned. They make very little use of the recent statistical literature on treatment effects. Our objective is to fill this gap. A further objective is to study grade repetition, and other policy variables, such as class-size, in an integrated, structural model of individual student records and school-histories. This structural model would possess dynamic elements and rely on the idea of skill and knowledge accumulation by the student. We plan to exploit several panels of pupils that have been made available to us by the DEP (Départment de l'évaluation et de la prospective, Ministère de l'éducation Nationale). Secondary school panels from 1989 and 1995, and the primary school Panel from 1997. - - For instance, in the primary panel, pupils' school performances are evaluated at grade 1 (CP), grade 3 (CE2) and grade 6 (6ème). We observe test scores in grade 3 for the pupils who did not repeat a class between grade 1 and 3 (say Y2(0)). We also observe the test scores of those who did repeat a class (say Y2(1)). Clearly enough, we never observe Y2(0) and Y2(1) simultaneously. Yet, to answer the question of the effectiveness of grade repetition, we have to characterize Y2(1) Y2(0), which is by definition the only fully ceteris paribus measure of the effect of grade repetition. - - Computing the mean of Y2(1) among those pupils who have repeated a class (R=1) minus the mean of Y2(0) among those who did not (R=0) will not work because only the least able pupils should have repeated a class. So the two subpopulations (R=0) and (R=1) are not comparable. What has been done in the afore-mentioned empirical literature was to condition on the initial test scores (say Y1). There are two problems there. First Y1 may measure the true pupil's ability with error. Second, and more importantly, some new knowledge may have been accumulated between CP and CE2 (grades 1 and 3) which both conditions R and Y2, even conditional on Y1 or the information it contains on pupils' initial school ability. There are severe endogeneity problems affecting the estimation of the effect of grade repetitions on outcomes. - - Understanding the conditions under which this type of model is identified from panel data, and developing appropriate estimation techniques, is important to provide a rigorous and impartial evaluation of grade repetition. We plan to explore various approaches, including matching estimators, instrumental variable estimation, latent-variable, endogenous dummy-variable techniques, and so on. However, the interest of our project goes well beyond this particular application, as treatment-effect models with dynamic factors (here accumulated knowledge) arise in many other applications of social sciences (for example, individual characteristics like IQ vary little over time but others like self-esteem of social skills are both variable and can be easily targeted by educational policy; see the numerous evaluations of the Perry-School experiment). - - Preliminary work on the data has shown that naïve ordinary least squares regressions of test scores at grade 9 (3ème) on test score at grade 6 and the occurrence of a grade repetition yields a negative return of grade retention. However, attempts at instrumenting grade repetition show that the sign of the coefficient on the grade repetition dummy can change. This motivates our research: these preliminary results show that more work is needed on this delicate question. - - Our most ambitious goal will be an attempt at modelling several transitions simultaneously (class transitions or repetitions) in junior high-school, using panel data, matched with observations relative to the environment (i.e., family background, class-size, other information from the fichiers établissements, etc.), to estimate a dynamic structural model of human-capital accumulation, in which the combined effects of class size and gr...

Project coordination


The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.



Help of the ANR 73,000 euros
Beginning and duration of the scientific project: - 24 Months

Useful links

Explorez notre base de projets financés



ANR makes available its datasets on funded projects, click here to find more.

Sign up for the latest news:
Subscribe to our newsletter