Returns to Education in India: An Instrumental Variables Analysis Using the District Primary Education Programme as a Source of Exogenous Variation in Schooling

Abstract

Estimating the causal effect of schooling on earnings is a foundational challenge in labour economics because standard Mincerian regressions are confounded by unobserved ability bias and measurement error in years of schooling. This study addresses both concerns by exploiting the District Primary Education Programme (DPEP), launched by the Government of India in 1994 in 42 educationally-backward districts and subsequently expanded in phases to 271 districts by 2002. Following the identification strategy of Duflo (2001) for Indonesia, the analysis combines district-level and cohort-level variation in DPEP exposure to construct an instrument for individual years of completed schooling. Two-stage least squares estimates are obtained from a pooled sample of the 2018–19 and 2022–23 rounds of the Periodic Labour Force Survey (PLFS) covering wage-earners aged 23–37 years (cohorts born 1985–1995). The first-stage F-statistic of 28.4 comfortably exceeds the Stock–Yogo (2005) weak-instrument critical value, and the over-identification test (p = 0.42) supports the exclusion restriction. The IV estimate of the private return to a year of schooling is 9.6 per cent larger than the OLS estimate of 7.1 per cent consistent with the international literature surveyed by Card (1999, 2001) on the role of measurement error and on the marginal-returns interpretation of IV. The returns are substantially higher for women (12.4 per cent) than for men (8.3 per cent), and higher at post-secondary than at primary levels, indicating convexity. The findings support continued public investment in basic and secondary education, particularly for girls, and suggest that earlier estimates of the return to schooling in India may be downward biased.