TL;DR
This paper evaluates the impact of affirmative action in Indian engineering colleges, finding that the policy successfully targets socio-economically disadvantaged students who see a massive 40-70% increase in earnings upon admission. However, this redistribution comes with an efficiency cost: the upper-caste students displaced by the policy would have realized nearly double the absolute earnings gain compared to the lower-caste students who took their place.
What Problem Does This Paper Address?
Affirmative action is one of the most debated social policies globally, yet empirical evidence on its actual costs and benefits is surprisingly thin. Critics often raise two major concerns:
- The "Creamy Layer" Argument: That these programs only benefit the relatively wealthy members of disadvantaged groups, essentially swapping one elite for another.
- The "Mismatch" Hypothesis: That students admitted with lower test scores are unprepared for rigorous curricula, leading to poor academic performance and negligible labor-market returns.
The authors use the Indian caste-based "reservation" system—where over 50% of seats in state-run colleges are set aside for Scheduled Castes (SC), Scheduled Tribes (ST), and Other Backward Castes (OBC)—to test these hypotheses in the high-stakes world of engineering admissions.
Methods
The researchers employed a Regression Discontinuity (RD) design, treated as an Instrumental Variables (IV) approach.
In the studied Indian state, admission is a deterministic function of a standardized entrance exam score and caste category. By comparing applicants just above and just below the admission cutoffs for each caste, the authors could isolate the causal effect of attending an engineering college.
What sets this study apart is its data collection: rather than just looking at those who graduated, the authors tracked down a census of applicants 8 to 10 years after the exam, including those who were rejected. This allowed them to observe the "counterfactual"—what happens to the people who "lost out" due to affirmative action.
Datasets
- 1996 Entrance Exam Census: Data on all individuals who took the engineering entrance exam in one Indian state (scores, ranks, and caste).
- Primary Follow-up Survey (2004–2006): Detailed interviews with ~700 households (both applicants and parents) to measure income, occupation, and social attitudes a decade later.
- National Sample Survey (NSS): Used as a benchmark to compare the applicant pool against the general urban population of the state.
Key Insights and Ideas
- Targeting is Effective: The policy does not just help the "creamy layer." Displaced upper-caste students came from significantly wealthier backgrounds (avg. parental income Rs 14,088) compared to the lower-caste students who displaced them (Rs 8,340).
- Refuting "Mismatch": Despite entering with much lower test scores, lower-caste students derived a "strong, positive economic return" from admission. The education was not "wasted" on them; it fundamentally altered their economic trajectory.
- The Efficiency-Equity Trade-off: The study quantifies a stark trade-off. While lower-caste students gain significantly, the marginal upper-caste student would have gained even more in absolute terms. For every 1 Rupee "lost" by an upper-caste applicant, the lower-caste displacer gained only 0.4 to 0.6 Rupees.
- Social Peace: Contrary to fears that affirmative action breeds deep resentment, the authors found no evidence that rejected upper-caste applicants held more negative attitudes toward lower castes or the policy itself.
Key Results
- Income Gains: Admission increased lower-caste members' monthly income by Rs 3,700 to Rs 6,200, representing a 40% to 70% increase.
- Absolute Return Gap: The marginal upper-caste entrant enjoyed nearly twice the earnings level gain as the marginal lower-caste entrant.
- Job Quality: Lower-caste admits were less likely to end up in "advanced technical" roles (using computers or speaking English at work) compared to their upper-caste counterparts, explaining part of the earnings gap.
- Gender Impact: Interestingly, the policy appeared to slightly hurt female applicants, as the share of women was higher in the upper-caste applicant pool than in the lower-caste pool.
Limitations and Future Work
The authors acknowledge several limitations:
- Scope: The study is limited to one state and one field (engineering). The "educational production function" might look very different in medicine or the humanities.
- Sample Size: While the census was large, the follow-up survey of 700 households led to some "noisy" estimates, particularly in the narrowest discontinuity samples.
- Long-term Dynamics: The data captures outcomes 10 years post-exam; it doesn't account for how these trajectories might diverge or converge over a full 40-year career.
Future research is needed to see if these findings hold across different regions of India and whether other redistributive tools (like income-based aid) could achieve similar equity goals with lower efficiency costs.
Summary updated