Differences in the response heterogeneity of educational attainment and achievement to the same instrumental variable

ABSTRACT

The typical econometric application of instrumental variable (IV) estimates imposes a constant-treatment-effect model for all individuals. Using regression discontinuity design, we find while older school starters tend to have lower educational attainment, they are more likely to achieve qualifications at different educational levels for the UK. This finding joins the discussion on instruments for educational outcomes by offering some evidence that school starting age can affect treated individuals diversely. We highlight that identifying the treatment effect with methods similar to IV estimates should be concerned in the context of heterogeneous treatments; that is, the instrument in question has an effect on treatment for particular subpopulations only, and those subpopulations may not be the same for different treatments.

KEYWORDS:

• Instrumental variable
• heterogeneity
• educational attainment
• educational achievement

JEL CLASSIFICATION:

• I00
• J00

Acknowledgements

Authors acknowledge the Economic and Social Data Service for access to the General Household Survey. Fan gratefully acknowledges financial support from the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD), SRF for ROCS, SEM; Nanjing University of Finance and Economics Pre-research fund YYJ201413 and ‘Young Scholar’ Program. We are grateful to Paul Devereux, Nicholas C.S. Sim, Rong Zhu and two anonymous referees for their helpful comments.

Notes

¹ There is actually evidence that delaying entry tends to decrease attainment (Dobkin and Ferreia, 2010) and would also cause child to enter the labour market at an older age (Deming and Dynarski 2008).

² 31 August is always the cut-off for the UK over time, although some other date works for very few schools but not for different years. By referring relevant reports (Crawford, Dearden, and Meghir 2010) in which they review the local education authority (LEA) admission policy history over last 20 years, we find 31 August cut-off is applied by most regions and schools.

³ The compliance rate with the school starting age rule is very high in the UK, at above 95% (Bedard and Dhuey, 2006; Crawford, Dearden, and Meghir, 2010).

⁴ See Black, Devereux and Salvanes (2011) among others.

⁵ Researchers find little or no evidence of early/late starting or grade retention for the UK (Bedard and Dhuey, 2006; Devereux and Fan, 2011).

⁶ The GHS now is renamed as General Lifestyle Survey. Given Scotland has a different educational system from England and Wales (children start school on March there), we drop the Scotland sample and use England and Wales only. We are convinced that the result would not be changed much since the sample size of Scotland is quite small.

Additional information

Funding

This article was funded by Nanjing University of Finance and Economics Pre-research fund: [YYJ201413]; SRF for ROCS, SEM [48] and the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD).