Cash or In-kind Transfers? Evidence from a Randomised Controlled Trial in Delhi, India

Abstract

This article examines a randomised intervention in Delhi, India, that provided unconditional cash transfers to a group of households as a replacement for the food security offered by a below-poverty-level card. The experimental approach can differentiate beneficial effects due to either unconditional cash transfers or newly opened bank accounts. The unconditional cash transfer does not induce a decline in food security; rather, it provides opportunities for households to shift to other nutritious options in non-cereal product categories.

Acknowledgements

The authors would like to thank two anonymous referees, Tobias Klein, colleagues at IDF, participants of the development economics seminar at Tilburg University, CERMI (Brussel) and Leuven University (Belgium), for constructive comments.

Notes

1. An income-targeted system must test means (proxy means testing), which may result in high Type I and Type II errors (that is, they do not reach many for whom they are intended and do reach some who are not the intended beneficiaries). They may also create poverty traps, such that anybody who increases their income to just above the threshold loses the grant and faces a marginal ‘tax’ rate of up to (and even above) 100 per cent, fostering what economists call moral hazards and immoral hazards. In addition, as Ellis (2012) notes, income-targeted cash transfers may lead to social tensions between included and excluded groups.

2. One important difference between 2 and 3 on the one hand and 1 on the other is that whereas 2 and 3 are initially at a point of tangency, 1 initially appeared at the kink on the original budget line, which is not at a point of tangency.

3. NSS conducts national consumption surveys every year. Every five years, however, they conduct the survey with a larger sample size.

4. We conducted the balancing test on each individual item in ‘PCnonfood’ variable between C1 and C2. All the items in the group are balanced in the baseline except for personal care for men and women.

5. WALD tests for these tables are also presented in the Online Appendix, tables A7–A9.

6. In the Online Appendix, tables A4–A6 we present the impacts on people’s consumption of subgroups of non-cereal food items, for which some quantity information is available. Specifically, we considered the impact of the shift to a cash transfer on quantities consumed (kilograms) for pulses, milk and edible oil. These quantity regressions (similar to those for monetary values) suggest that the CT shift increased consumption of these three important products in the non-cereal food group. Pulses appear significant in OLS and ANCOVA regressions but insignificant in the fixed effect regression for both groups C1 and C2. However, we found no additional impact of opening a bank account; C1 and C2 never differ significantly from each other.

7. The Wald test suggests that there are no differences between C1 and C2, see the Online Appendix.