Principal leadership and student achievement: decentralising school management in Saudi Arabia

Abstract

Saudi Arabia has expanded access to secondary schooling over the past generation, while also pushing to lift quality. This includes decentralising authority out to principals, equipping them to set higher performance expectations and deploy incentives to attract and retain strong teachers. We find wide variability in the extent to which principals display leadership behaviours, based on nationally representative school samples drawn in 2003, 2011, and 2015 for the Third International Mathematics and Science Study. Pupil achievement did improve across these cohorts, markedly among girls attending schools in which principals enriched the academic climate and deployed teacher incentives, after taking into account the social class background of students and the school’s instructional resources. Greater availability of resources, including computing tools, improved over the period but did not account for higher achievement in mathematics. Lessons for other nations are discussed, as international donors press decentralised governance, at times ignoring local cultural and institutional contexts.

Keywords:

• Principal leadership
• decentralisation
• Middle East
• school
• effects

Acknowledgement

This project was funded by the Deanship of Scientific Research at King Abdulaziz University, Jeddah. The authors gratefully acknowledge this technical and financial support.

Notes

1. Two-thirds of Saudi Arabia’s population is under 30 years old, although the overall fertility rate has begun to decline (Alyaemni et al. 2013).

2. The education ministry has articulated five core domains in mathematics, which middle school teachers are to address: numbers (including application of proportions, rate of change and types of numbers); algebra; measurement; geometry; and statistics (Alshaya 2011).

3. The regression estimates required complete data for all selected variables, drawn from the TIMSS data for each cohort, including test scores and responses from principals. We found few significant differences between the small count of schools with incomplete data. Slightly more boys’ schools were excluded in the 2015 sample, relative to the share of girls’ schools (comparative table available from authors).

4. The sampling weight totwgt is used for student- level variables; schwgt is used for school-level variables when calculating mean values, standard deviations and for estimation models, as advised by TIMSS statisticians (Foy, Arora, and Stanco 2013; Martin 2005). This ensures that findings can be generalised to all Saudi middle schools.

5. We assume that each unit’s random error, ${\text{u}}_{0\text{j}}$, is distributed normally, mean of 0 and constant variance $\tau$.