Mines: the local wealth and health effects of mineral mining in developing countries (with Prabhat Barnwal)

Abstract. Do residents of mining communities face health-wealth trade-offs? We conduct the first extensive assessment of this question using micro-data from communities near about 800 mineral mines in 44 developing countries. Households in mining communities enjoy a substantial medium-term gain in asset wealth (0.3 standard deviations), but experience a ten percentage point increase in the incidence of anemia among adult women, and a five percentage point rise in the prevalence of stunting in young children. Prior evidence links both of these health impacts to metal toxicity – and in particular, exposure to high levels of lead. We observe health impacts only near mines of a type where heavy metal pollution is to be expected, and find no systematic evidence that health is affected in ways that are not specific to exposure to such pollutants. Benefits and costs are strongly concentrated in the immediate vicinity (no more than 5km) of a mine. Consistent results emerge from a range of distinct identification strategies. Baseline results come from a cross-sectional fixed effects model, and mine-level and mother-level panels. An instrumental variables approach serves as a robustness check. To demonstrate that the observed health impacts are due to pollution, we develop two difference-in-difference tests tailored to the known association of certain mine types with heavy metal pollution, and to the pathophysiology of lead toxicity. Our results add to the nascent literature on health impacts near industrial operations in developing countries.

Supporting online appendices are available here.

An earlier version was published as a Columbia PER Working Paper.

Under review.


Cost-sharing in environmental health products: evidence from arsenic testing of drinking water wells in Bihar, India. (with Prabhat Barnwal, Alexander van Geen, and Chander Kumar Singh)

Abstract. Groundwater contaminated with arsenic of natural origin threatens the health of tens of millions of villagers across South and Southeast Asia. With a field experiment conducted in Bihar, we assess the scope for cost-shared provision of well-water arsenic tests, and study how households use the information revealed by testing. Demand is substantial, but highly sensitive to price; uptake falls from 69% to 22% of households over our price range (Rs. 10 to Rs. 50 – about equivalent to daily per capita income). Repeating the sales offer after a two-year hiatus raises overall uptake substantially, from 27% to 45%. About one-third of households with unsafe wells switch to a safer water source. Households that bought at higher prices are no more likely to switch, consistent with an absence of sunk cost or screening effects. Finally, we demonstrate that households selectively forget and remove evidence of adverse test outcomes.

Revised and re-submitted,  Journal of Environmental Economics and Management.

Field work is under way for a follow-up project to test whether incentives to voluntarily form well-sharing groups prior to testing can help overcome barriers to sharing.


Geolocation error and the use of DMSP-OLS night lights in high-resolution applications

Abstract. Is night lights data from the Defense Meteorological Satellite Program’s Operational Linescan System (DMSP-OLS) observed precisely enough to measure human activity at high spatial disaggregation? Night lights are routinely used as proxies of ground-based activity at the level of countries, sub-national regions, or metropolitan areas. Due to the data’s resolution (30 arc seconds), they might also be useful in studying processes at much higher geographic disaggregation – for instance, at the level of towns or villages. Yet, DMSP-OLS data are recorded with geolocation error that could interfere with such uses. I use a new data set of 185 calibration sites that are small, bright, and remote, to assess the offset between the actual location of light sources and their recorded location in the most commonly used yearly night lights data product. The error is small enough to be ignored, even in applications where the spatial scales of interest are on the order of a few kilometers. Root mean square error is a mere 0.52km in zonal and 0.67km in meridional direction. I illustrate the potential and limits of very high-resolution applications by benchmarking light data on household asset wealth in all official localities in Mexico. Night lights are a strong proxy measure of cross-sectional wealth differences even within small administrative units, in particular in the poorest, least populous, and most dimly lit regions. The data therefore has potential to improve applications such as poverty maps, although the analysis of changes over time is more subtle.

Revisions requested, Remote Sensing of Environment. Paper available upon request.


Ongoing research

  • Household-level impacts of CGIAR’s modern seed varieties since 1960. (with Prabhat Barnwal, Ram Fishman, Nathan Mueller, and Gordon McCord)
  • Bias when distance is a regressor (with Prabhat Barnwal and Ruinan Liu)
  • The impact of landmine clearance on household welfare – evidence from Cambodia (with Prabhat Barnwal and Ofir Reich)
  • Mining, Income, and Cognitive Performance – Evidence from Chile (with Pablo Egaña del Sol)