An analysis of the impact of socio-economic disadvantage and school quality on the probability of school dropout

Abstract

This paper combines the Australian Programme for International Student Assessment (PISA) micro-level data with its longitudinal continuation, the Longitudinal Survey of Australian Youth data, to measure the association between individual PISA scores and early school dropouts. We use multilevel modelling to distinguish between student and school factors when estimating school dropout propensity for 15–18-year olds. We model attrition and deal with the possibility that PISA scores are endogenous in the dropout decision. We find that PISA scores predict well early dropout, especially so for bottom achievers, and that individual and social disadvantage plays a crucial role in this relationship, both directly and indirectly.

Keywords:

• PISA
• socio-economic disadvantage
• school dropout
• student outcomes
• multilevel modelling

JEL Code:

• I24
• I21

Notes

1 Those who completed a VET course at Certificate I or II level remain accounted as dropouts since this type of training does not provide individuals with substantial extra qualifications compared to the level they would normally have after dropping out.

2 We used an alternative specification based on the simultaneous estimation of both equations (PISA scores and probability of dropout). We obtained similar results compared to the method described above. As a result we opted for presenting this latter, simpler, method. The results of the simultaneous equations specification are available upon request.

3 We tested several more general specifications of the model involving, notably, a random coefficient given to the SES variable. Results showed that the variance of the random coefficient becomes non-significant as the right-hand side of the model becomes richer by including the full set of student and school characteristics. We conclude that a random coefficient specification for SES is only statistically relevant for a restricted right-hand-side model specification which includes only the SES of the student. Marks, McMillan, and Hillman (2001) made the same observation using the 1995 cohort of LSAY on a slightly different topic, namely the tertiary entrance performance of Australian students.