The effects of age on educational performances at the end of primary school : cross-sectional and regression discontinuity approach applications from Reunion Island

This paper focuses on the quantitative measure of the causal relationship between age and school results of pupils at the end of primary school in Reunion Island. The effect of age is composed of at least three distinct ones : (1) age at entry effect, (2) age at test effect and (3) relative age (compared to grade peers) effect. In order to extend the knowledge about determinants of educational sucess, especially about the impact of age on scholar results and then help policy makers in their decisions about optimal policies in the education field by providing informative results ; this paper, using cross-sectional data sets, exploits an exogenous variation of the age at test within a grade induced by the date of birth to measure the causal impact of age at test on the national achievement assessment scores in grade 5 in Reunion Island. I implement additionally a regresion discontinuity design for comparison purpose. The principal findings are that the age at test have a substantial positive effect on test scores in grade 5. Also, the effects in grade 5 are heterogeneous across sex subgroups but such a pattern is difficult to draw across social category subgroups. These results would suggest at best that, in order to improve the educational results of pupils in Reunion Island meaning the age variable, policy makers could first increase the minimum age of school entry. Second, they could regulate classroom compositions such that the age distribution within a classroom does not disperse too much. Third, they could normalize national achievement assessment scores by age or making pupils with different ages within a grade pass the national assessment at different times such that they have sufficiently close ages at test to not significantly impact their results. The latter enables at the same time to correct the inequality of having a different month of birth (unchose by the pupils) which is likely to lead all else equal towards different educational outcomes. Pupils would be indeed assessed « at equal luck »