This paper revisits a part of the analysis by Banerjee et al. (2020), in which they examine the consequences of the nation-wide scale up of reforms to the funds management system (e‑FMS) in India’s national workfare programme, using a two-way fixed effects specification. They report a substantial 19 percent reduction in labour expenditures. We exploit the recent literature that highlights the limitations of the TWFE estimator in the presence of staggered roll out and effect a Goodman-Bacon decomposition of the TWFE coefficient, to pinpoint sources of identifying variation. We undertake a detailed examination of subsamples of six constituent and valid DiDs based on timing of treatment that are averaged into the TWFE coefficient to identify heterogeneity in treatment effects. This disaggregated subsample analysis does not support the conclusion of any reductions in MGNREGS labour expenditures, suggesting that the TWFE coefficient based on the full sample is indeed biased.

Authors:

• Deepti Goel
• J.V. Meenakshi
• Zaeen De Souza