Gujarat Pollution Audit Intervention

In many regulated markets, private, third-party auditors are chosen and paid by the firms that they audit, potentially creating a conflict of interest. In collaboration with the Gujarat Pollution Control Board (GPCB), researchers affiliated with the EGC (Rohini Pande and Nicholas Ryan) and J-PAL (Esther Duflo and Michael Greenstone) conducted a two-year field experiment in the Indian state of Gujarat that sought to curb such a conflict by altering the market structure for environmental audits of industrial plants to incentivize accurate reporting. The researchers designed and evaluated a modified audit system that sought to improve the accuracy of auditor reporting on pollution. The sample consisted of the population of audit-eligible plants in the two largest cities of Gujarat, India.

The researchers obtained from GPCB a list of all red-category (i.e., high pollution potential) small- or medium-scale plants. Just before the 2009 audit season, the researchers selected a provisional sample of audit-eligible plants and then randomly assigned half of the plants within this provisional sample, stratified by region, to the audit treatment group. Treatment plants were formally notified of the changes in the audit regulation that would apply to them by a letter from GPCB. Relative to the status quo, the treatment altered three components of the audit system during year 1: an auditor was randomly assigned to the plant, paid from a central pool at a fixed rate, and its reports were backchecked for accuracy. In year 2 only, direct incentive pay for auditor accuracy was added.

The researchers collected data from several sources to evaluate the intervention, especially on the basis of the accuracy of auditor reporting and the pollution response of plants. Two data sources are used to measure accuracy. First, audit reports were filed with GPCB in 2009 and 2010. These reports cover a mandated set of water pollutants and air pollutants. The second source of data for auditor accuracy is the backchecks, which were conducted in a sample of treatment plants throughout 2009 and 2010. The third source of data, on actual plant pollution emissions, is an endline survey conducted from April through July 2011, approximately six months after the last audit visits in the treatment group. The fourth source of data comes from the GPCB administrative records. These data cover GPCB’s plant inspections for plants in the audit sample between 2008 and 2011. These datasets are available here.

Gujarat Environmental Inspection Trial

High pollution persists in many developing countries despite strict environmental rules. To study how plant emission standards are enforced, the researchers, again in collaboration with GPCB, experimentally doubled the rate of inspection for treatment plants and required that the extra inspections be assigned randomly. The goal of the experiment was to estimate the impact of moving from the status quo, infrequent inspections allocated with discretion, to regular inspections of all plants at prescribed inspection rates. Such a reform would bring the GPCB into compliance with its own prescribed inspection rates and the Central Pollution Control Board’s (CPCB) inspection rules.

To this end, between August 2009 and May 2011 the researchers worked with GPCB to increase inspection frequency for a random subset of highly polluting plants. By CPCB rules, these plants are supposed to be inspected either once per year if they are small scale or once in 3 months if they are medium scale. From this population, the sample of 960 plants was drawn in two batches. The researchers selected all 473 audit-eligible plants in Ahmedabad and Surat and then randomly selected 488 plants from the remaining audit-ineligible population. Inspection treatment assignment was randomized within region by audit-treatment status strata. The treatment was thus cross-randomized and implemented concurrently with the pollution audit reform treatment.

The plants assigned to the inspection treatment were assigned at least one annual initial (routine) inspection and up to four inspections per year. In the first quarter, the plant was assigned one initial inspection, after which it was randomly assigned on a quarterly basis to be inspected again with probability 0.66. After four quarters, this cycle started over. Regional GPCB teams consisting of an environmental engineer and scientist conducted treatment inspections. Each morning in each region, the designated inspection team was randomly assigned a list of plants from the treatment group at which to conduct initial “routine” inspections that day. This mimicked GPCB’s practice of assigning teams to plants, except that the plant assignment was random, rather than being based on an official’s discretion.

For program evaluation, the researchers collected data from two sources: an end-line plant survey and GPCB administrative records. The end-line survey was conducted between April and July 2011 by independent agencies. The survey collected pollution readings, expenditures for abatement equipment investment and maintenance, and data on other aspects of plant operations. The second source of data comprises GPCB documents on its interactions with plants; these documents were categorized by (a) whether they record an action of the regulator or a plant and (b) the type of action they record. These datasets are available here.