Large-Scale Education Reform in General Equilibrium: Regression Discontinuity Evidence from India: Comment

TL;DR

This paper revises the impact estimates of India's 1990s primary schooling expansion using improved regression discontinuity methods, revealing smaller effects and highlighting the importance of methodological precision.

Contribution

It introduces methodological revisions, including clustering and error adjustments, that significantly alter the estimated impacts of the education reform.

Findings

Impact on schooling attainment reduced by two-thirds

Wage impact reduced by one-third

Estimates of general equilibrium effects are highly variable

Abstract

Mainly through regression discontinuity designs, Khanna (2023a) studies the impacts of a primary schooling expansion in India in the 1990s. Absent from the data set are four districts close to the modeled treatment discontinuity. Incorporating them cuts the impact of intention to treat on schooling attainment by two-thirds and the impact on wages by a third. Methodological revisions, including clustering by the geographic unit of treatment, double or triple standard errors, bringing estimates within a standard error of zero. These findings are robust to varying the location of the discontinuity, the bandwidth, and the radius of an exclusionary "donut." The estimates of the general equilibrium effect on the skill premium, as well as elasticities of substitution across age and skill groups, have high variance. One cause is that the treatment discontinuity does not occur quite where modeled. Moving the assumed cutoff pro-duces sharper but wrong-signed results.