The Importance of Observing Early School Leaving and Usually Unobserved Background and Peer Characteristics in Analysing Academic Performance

Melbourne Institute Working Paper No. 05/07

Date: February 2007

Author(s):

Guyonne Kalb
Sholeh A. Maani

Abstract

In this paper, we use a recent panel data set from New Zealand to examine the link between the academic performance and the decision by teenagers to drop out of school before exams at the end of year 10. These choices have significant lifetime economic impacts, since early school leaving in many cases closes pathways to further education. We address endogeneity and error correlation of potential performance in national examinations and school-leaving choices prior to exams. Birth month provides an instrument used in the equation for drop out, because those born in particular months can legally leave school before the exam takes place, whereas the other students cannot. The analyses incorporate the effect of academic ability (childhood IQ), parental education, family resources at different points in time while the child is growing up, and school and peer characteristics. The results show that those who drop out early are unlikely to have performed well in the exam. The predicted difference between those who drop out or continue, at least up to their exam, is almost completely explained by observed factors. Leaving out those variables which are often not available in other datasets (such as childhood IQ, childhood family resources and teenage peer effects), we find that the unobserved factors in academic performance and early school leaving are correlated. It is found that beyond childhood IQ and family resources, teenage peer and school factors have additional and significant associations with grade outcomes. This has important policy implications.

Download Paper