How do Educational Transfers Affect Child Labour Supply and Expenditures? Evidence from Indonesia of Impact and Flypaper Effects

Abstract

This study utilises a large nationally representative household survey of unusual scope and richness from Indonesia to analyse how the receipt of educational transfers, scholarships and related assistance programmes affects the labour supply of children and the marginal spending behaviour of households on children's educational goods. We found strong evidence of educational cash transfers and related assistance programmes significantly decreasing the time spent by children in income-generating activities in Indonesia. Households receiving educational transfers, scholarships and assistance were also found to spend more at the margin on voluntary educational goods. These results were stronger for children living in poor families. Our results are particularly relevant for understanding the role of cash transfers and educational assistance in middle-income countries where enrolment rates are already at satisfactory levels, but the challenge is to keep the students in school at post-primary levels.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ De Silva ([email protected]) and Sumarto ([email protected]) are with the National Team for Accelerating Poverty Reduction (TNP2K) in Indonesia. The views expressed in this paper are the sole responsibility of the authors and do not necessarily reflect the views of TNP2K.

² The proposition by Rosenbaum & Rubin (1983) can be stated as: Let P(X_i) be the probability of unit i having been assigned to treatment, defined as . Assume that , for all and for the N units in the sample. Then, Corollary: If and the assumptions of the above proposition hold, then . The proposition implies that observations with the same propensity score have the same distribution of the full vector of covariates . The propensity score will be estimated by a Probit model: .

³ We also performed tests on covariate balancing. It was found that the differences between the households between the treated and untreated groups are quite small after matching, and that matching removed any bias that existed for almost all covariates. A t-test of equality of means in the two samples of participants and non-participants revealed that there is no systematic pattern of significant differences between the covariates in the treated and non-treated groups after conditioning on the propensity score. The test results and the exact number of individuals lost due to common support requirement are available upon request from the authors.