Season of birth and school success in the early years of primary education

Abstract

Several studies have reported significant relationships between children’s season of birth and measures of their academic success (i.e., the ‘season of birth effect’). Whereas most of these studies were cross‐sectional, the current study uses growth curve modelling to analyse longitudinal data on 3,187 children in Flemish primary education. The results indicate season of birth effects on both grade retention and mathematics achievement during the first two years of primary school. Because the Flemish cut‐off date is 31 December, children born in the fourth quarter (October‐November‐December) invariably are among the youngest in their grade age group. Almost 20% of these children were found to have been retained or referred to special education by the end of Grade 2, whereas for children born in the first quarter (January‐February‐March), this was only 6.34%. First quarter‐born children also showed moderately higher mathematics achievement at the start of first grade. During the next two school years, this achievement gap between children born in the first and the fourth quarter narrowed significantly. Finally, differentiated instruction was not found to be related to the decrease of the season of birth effect.

Notes

1. ‘SiBO’ is a Dutch acronym for ‘Schoolloopbanen in het BasisOnderwijs’ (School Careers in Primary Education).

2. Flemish elementary education is composed of three pre‐primary and six primary grades. Children normally enter first grade (i.e., the first year of primary school) on September 1 of the calendar year they turn six years old. Consequently, children in the last year of kindergarten are typically between five and six years old.

3. Because the mathematics scores used in the present study were obtained through an IRT analysis, they were assumed to represent children’s true mathematical abilities. Additionally, the IRT model produced estimated standard errors for all individual mathematics scores. In all models in this study, it was assumed that the level‐1 variance was known and equal to the square of those standard errors.