Pollution externalities and health: A study of Indian rivers

India's rivers are heavily polluted. One of the most polluted sites is the city of Kanpur on the banks of the river Ganga, one of India’s longest and holiest rivers. At this site, the Ganga river receives large amounts of toxic waste from the city's domestic and industrial sectors, particularly the tannery industry. We study the impact of a landmark piece of judicially mandated environmental legislation in this city: the 1987 Supreme Court decision in the M.C. Mehta versus Union of India case that mandates, among other things, tanneries in the district of Kanpur to clean up or shut down.

We combine data from demographic surveys, pollution monitors and the geo-spatial coordinates of rivers to identify the impacts of the policy on both pollution and mortality.  Our pollution data, drawn from the Indian Central Pollution Control Board, were originally gathered and used by Greenstone and Hanna (2012).  Our infant health data come from the District-Level Household Survey II (DLHS-2). In the RCH-2 module, mothers report age and survival for all of their children; from these birth histories, we create a panel of district-month infant mortality rates. We start with the raw total of 1,393,431 births from 1967 through 2004 that are reported in RCH-2. We then match each birth to pollution data from the district in which the birth took place and then collapse to district-level means. This yields a panel of 4,394 district-month observation with non-missing infant mortality rate and BOD, spanning 41 districts and 8 states.

Our first approach exploits the novelty and timing of the Supreme Court ruling to estimate its impact on both pollution and mortality. In a difference-in-differences framework, we find evidence that the Supreme Court verdicts produced a significant drop in both river pollution (as measured by Biochemical Oxygen Demand) and health risk (as measured by infant mortality). Our next approach explores the mechanisms of policy impact via a structural model. The model recognizes that the ruling may have affected the outcomes of interest via other channels such as economic impacts due to firm closures, and improved citizen information about both pollution and health. We explore the relative importance of the pollution channel - as opposed to income, behavioral, and other channels - in policy impacts, by constructing two instruments for river pollution. The first is upstream river pollution, which we argue is a valid instrument conditional our controls and fixed effects. The second is the Kanpur policy itself, which we do not argue to be necessarily valid because it affects pollution at the same time as it affects wages and information (among other possible determinants of health). We compare results of two separate instrumental variables (IV) regressions of infant mortality on river pollution: one in which we use only the upstream pollution instrument; and one in which we additionally use the Kanpur policy instrument. Our structural model of health motivates a direct comparison of the coefficients generated by these two IV regressions. Because we have two potential instruments and one endogenous regressor, the Sargan-Hansen test of over-identification restrictions provides a statistical “test of mechanisms”. We find high p-values for the Sargan-Hansen test statistic in most cases, which suggests that the Kanpur policy did indeed reduce infant mortality primarily by reducing river pollution. This result speaks directly to a common uncertainty about whether targeting pollution reduction is valuable when informational campaigns and incentives for avoidance behavior are viable alternatives.