Does unauthorized school absenteeism accelerates the dropout decision? – Evidence from a Bayesian duration model

Abstract

School absenteeism (or truancy) may be a signal of an ongoing process of student attrition that eventually leads to early school leaving. This article estimates how unauthorized school absenteeism accelerates the dropout decision. In particular, the timing of the dropout decision of truants is compared with the timing of regular school attendees using administrative data with insights into wrongly specified truancy spells. We correct in a Bayesian duration model for issues on data uncertainty and show that only a Bayesian structure on the data yields insightful and consistent results. The results indicate that the risk of truants to leave school early before the end of the compulsory education age increases with as much as 37.4%.

Keywords:

• Bayesian duration model
• school absenteeism
• early school leaving

JEL Classification:

• I21
• I28

Acknowledgements

We are grateful to Wim Groot, Henriёtte Maassen van den Brink as well as to TIER, NICIS and IFS (University College London) seminar participants for their useful comments and discussions. We would like to thank Orm Muilwijk of DMO Amsterdam and the Dutch Ministry of Education for providing the data. The usual caveat applies.

Notes

¹ It should be noted that some students may possibly return to school after being dropped out. This might be the case, for example, in the subsequent school year after their first dropout decision. The average number of dropouts per student is 1.14 with a very low SD of 0.46, indicating that the majority of students only dropout once.

² These findings are confirmed by plotting the Martingale residuals, indicating that the baseline hazard is no longer constant over time after the age of 17. A plot of the estimated Martingale residuals is available upon request. Also see the discussion on the proportional hazard assumption in Section III.