Paper
Banerjee, Abhijit, Angus Deaton, and Esther Duflo (2004). American Economic Review (Papers & Proceedings), 94(2): 326–330
Read the original paper → Opens at the publisher; use the DU library / JSTOR login if it asks for access.
In one paragraph
What determines health and well-being in a very poor rural area, and does access to health care matter independently of income? Rather than rely on existing data, the authors ran their own interlocking surveys in 100 hamlets of Udaipur district, Rajasthan, in 2002–03 — a village survey, a facility survey of the public and private providers villagers actually use, and a household survey with health measurements — with the NGO Seva Mandir and Vidhya Bhawan. The paper reports what those surveys found about health status, the use of public, private and traditional providers, and the condition of government facilities.
Setting and question
Udaipur district: a large tribal population, very low female literacy, and a three-tier public health system that exists on paper. The question is descriptive but foundational — what are the facts a theory of health in poor countries has to explain?
Data and identification
There is no treatment here; the contribution is the design of the sample (stratified by distance to a road, hamlets selected with probability proportional to size, ten households per hamlet) and of the instruments (facility visits, anthropometrics, self-reported and measured health). It is the paper to read for how a survey gets built and why each choice was made.
Main results
Poor measured health alongside heavy spending on care; widespread use of private and traditional providers; public facilities frequently closed or without staff. The details are the point — read the tables.
Why it is on the reading list
The foundation for the later Udaipur experiments (including Banerjee, Glennerster & Duflo 2008). It shows that credible empirical work often starts with careful measurement rather than with an identification strategy.
No interactive explainer for this paper yet. If you build one, see Contribute to get it added.