Age of Starting School, Academic Performance, and the Impact of Non-Compliance: An Experiment within an Experiment, Evidence from Australia

This paper estimates the difference in academic performance of oldest and youngest students a given grade. We employ Queensland Department Education school administration panel data for population state years 2008-2016. Academic is measured by National standard test scores (NAPLAN) teacher assessed measures effort individuals grades 3, 5 7. The empirical analysis employs regression discontinuity design (RDD) based on administrative rules age enrolment. class assigning mechanism operates via known cut-off date results child grade being almost year older than youngest. However, as parents may anticipate disadvantage their they choose to delay timing initial lack compliance potentially creates difficulties RDD identification strategy, particular assumption exchangeability around cut-off. exploit change rule from 2008 reform which postponed starting 6 months produced large increase rate. enables one gauge importance non-compliance estimating treatment effect versus younger cohort. find that pre-reform small generally statistically insignificant. Post-reform there sizeable significant diminishes sample proceeds through grades,