Please click on the arrow bullet to read a quick summary.
I also do mixed-methods work on language endangerment - more here.

Working Papers

Who Lends When it Floods? Banks versus Non-Banks
With Sankalp Mathur. Under review at the Journal of Banking & Finance. [PDF] We examine whether non-banks fill credit gaps following climate shocks. Using the August 2017 floods in two Indian states, Uttar Pradesh and Bihar, as a plausibly exogenous shock, we combine georeferenced flood-extent data with a novel, near-universe panel of Indian retail credit data. A difference-in-differences design, comparing non-banks and banks across flooded and unaffected pincodes, reveals that non-banks expand lending by 5.75% in affected areas relative to banks, alongside broader borrower outreach.

Industrial Policy for Unincorporated Firms: Evidence from India's One District One Product Programme
With Gopal Krishna Roy. R&R at the BE Journal of Economic Analysis and Policy. This paper studies the economic effects of the One District One Product (ODOP) programme, a subnational industrial policy implemented in Uttar Pradesh, India. Using repeated cross-sections of establishment-level data from government surveys of unincorporated enterprises, we construct a novel district–industry panel and implement a difference-in-differences design to estimate the programme’s impact on gross value added, employment, capital, and firm counts. We find little evidence of robust positive effects on these outcomes in the short run.

Policy Work

Language Endangerment and Culture Loss - A Case Study
Social Policy Research Foundation, 2022. [Stub at SPRF website.] [PDF.] [Slides.] [Bib.] This study attempts to analyse the cultural impact of language loss in Tai Khamti speakers, an endangered language community native to North-Eastern India. Primary data is collected to assess cultural markers relating to linguistic maintenance, indigenous knowledge and documentation, linguistic cultural markers, and narrative forms. The analysis of this data provides insight into culture loss through linguistic endangerment in North-Eastern India and the broader intersection of language and culture.

Optimal Lending Allocations for the Priority Sector
For FIDD, Reserve Bank of India. With Sankalp Mathur, Vidhya Soundararajan, and Kritika Chakrabarti.

Work in Progress

Bank Lending and Long-run Flood Adaptation
With Sankalp Mathur.

Land Reform and Solar Expansion: Evidence using Machine Learning and Satellite Imagery
With Ankit Bhatia.

Talks

Using deep learning to classify court hearings.
At xKDR Forum. July 2024. A large proportion of hearings in the Indian court system are non-substantive, in that they do not yield progress in the case. In the era of digitized court data, researchers now have access to millions of court orders, the text of which allows them to understand whether a particular hearing was substantive or otherwise. I provide a study of different AI/ML methods for text classification at scale, involving shallow and deep learning methods as well as a cost comparison to LLMs. I find that labelling data has decreasing marginal returns, allowing for deep learning to theoretically be more efficient than LLMs for judicial data.

Understanding Language Endangerment and its Policy Implications.
Online by the ST & SC Development, Minorities & Backward Classes Welfare Department, Government of Odisha. December 2023. [Event link.] [Slides.] This presentation was delivered to the Department of SC/ST Welfare at the Government of Odisha. It is intended for a generalist audience curious about the linguistic diversity and richness of India, and interested in the challenges and opportunities for language maintenance and revitalization. It is also a useful resource for policymakers, educators, researchers, and activists who are involved or interested in language planning and development.