Affirmative action in education: Evidence from engineering college admissions in India

Publication Details

Journal of Public Economics, February 2010, v. 94, iss. 1-2, pp. 16-29. Available From:

Link to Source
Marianne Bertrand, Rema Hanna and Sendhil Mullainathan
Institutional affiliations
None specified
Grant-holding institution
None specified
South Asia
Education, Social Protection
System Reform & Capacity Building, Tertiary Education, Labor markets & Employment
System Reform & Capacity Building, Tertiary Education, Labor markets & Employment
Equity Focus
Ethnic Minorities, Gender
Evaluation design
Instrumental Variables (IV), Regression Discontinuity Design (RDD)
Journal Article


This study addresses the efficacy of an affirmative action programme in engineering colleges run by an Indian state in 1996. The investigation, based on a census of applicants and follow-up surveys conducted from 2004 to 2006, is two pronged. First, does the programme target disadvantaged-caste students? Second, does it positively affect the employment and earnings of targeted students? The authors test the theory that caste-based inequalities in labour markets can be reduced by quotas in educational institutions for students from disadvantaged backgrounds.
A regression discontinuity design is implemented, using estimated cut-off test scores that qualify students for admission from each caste category: upper caste, other backward castes and scheduled castes. The authors confirm that engineering college attendance is almost zero for students with scores below the cut-offs for their respective categories, thus ensuring that test score is a valid discontinuity. The authors also estimate a separate set of IV models on the entire sample, using dummies indicating if the applicant’s score was above or below the cutoff to instrument college attendance.
The sample consists of all 2,054 seats in engineering colleges filled by the programme. Of these, 51 per cent of seats were reserved. Follow-up surveys were conducted for 721 households to investigate post-college economic indicators. Since the true cut-off test scores are unknown, a method used by Card et al. (2008) is used to produce estimates for each category. Heterogeneity amongst households was accounted for by incorporating a socioeconomic background index into the regressions.

Main findings

The disadvantaged-caste students able to attend engineering colleges through the programme belong to households that are relatively better off than the average minority household. However, parental income of the ‘displaced’ higher caste students was US$282 compared with US$167 for the ‘displacing’ students, indicating that the beneficiaries are relatively poorer than the adversely affected students; this is a statistically significant finding. On the other hand, the programme reduced gender diversity, as women constituted 23 per cent of the students displaced by the programme and only 16 per cent of the students granted admission through the programme. Lower caste programme beneficiaries were able to earn US$70 to US$124 more every month after college; upper caste beneficiaries were found to have an even greater income increase. However, the limited sample size and large standard errors preclude emphatic conclusions for the labour market income effects.
The primary policy implication of the study is that the affirmative action programme increases the enrolment of lower caste groups and positively affects their future earnings, thus having a redistributive effect. However, it does impose costs in terms of the reduced incomes of the higher caste students who would have been admitted in the absence of reservations. The income losses of these students are larger than the income gains of those students who displace them. Although the programme manages to increase the enrolment of lower caste students, it negatively affects the gender balance. Additionally, the policy benefits students who are relatively worse off, compared with upper caste students, rather than the absolute poorest.

Scroll to Top