Paying Auditors for Honest Appraisals; An Experiment in India’s Gujarat State Shows a Way to Realign Interests and Incentives

Updated July 17, 2013, 9:21 p.m. ET

Paying Auditors for Honest Appraisals

An Experiment in India’s Gujarat State Shows a Way to Realign Interests and Incentives

DAVID WESSEL

The auditing business rests on what Joshua Ronen, a New York University accounting professor, once called “a structural infirmity.” Auditors are paid by companies they audit, much as rating agencies are paid by companies they rate. This gives auditors an economic incentive to lie on their clients’ behalf, even if that puts their reputations at risk. As an old German proverb puts it: Whose bread I eat, his song I sing. These conflicts of interest figured in the corporate-accounting scandals of the early 2000s and the subprime-mortgage securitization debacle of the late 2000s. Despite attempts to lessen conflicts, resolution has defied easy solution.Now a band of economists, the new breed that dabbles in experiments rather than models, thinks it has evidence that there’s a better way: Pay auditors out of a central fund, randomly double-check their work and link pay to accuracy. In short, change the incentives.

The setting for the two-year experiment was the Indian industrial state of Gujarat, where authorities are battling air and water pollution from textile plants. That is a long way from Wall Street and U.S. corporate boardrooms, but researcher Michael Greenstone of the Massachusetts Institute of Technology says, “Every single third-party audit market shares this characteristic: The auditor is hired and paid by the outfit being audited.”

Confronting the failure of inspection to curb pollution, India’s courts launched outside environmental audits in Gujarat in 1996. It was a sophisticated scheme. No auditor, for instance, could audit a plant more than three years in a row to avoid too cozy a relationship. The Gujarat Pollution Control Board was no pushover: It cut off electricity to plants when auditors reported they were found to have blatantly flouted the rules.

But the system wasn’t working. Some auditors charged half the price of what it cost to do the three-times-a-year tests, suggesting they were submitting phony results. Some were suspected of taking bribes. Some plants installed antipollution gear, but didn’t use it.

Looking at test results of one air pollutant, researchers found hints of corruption: 73% of auditors’ initial readings fell just below the regulatory standard; only 7% of initial readings found violations. Double-checks of the same plants by higher-grade inspectors found only 18% of the same plants had readings that hovered just below the standard; 59% actually violated it.

Mr. Greenstone and colleagues—pioneering experimentalist Esther Duflo of MIT and Harvard’s Rohini Pande and Nicholas Ryan—tracked 473 textile plants in Ahmedabad and Surat.

For half it was business as usual; plants hired auditors at the going rate. For the other half, auditors were assigned randomly, paid from a central pool at a rate high enough to cover costs and warned there was a 1-in-5 chance that an independent inspector would check their work. In this second group, far fewer (39%) of the initial auditor readings came in just below the regulatory standard; far more (42%) were above it. The double-checks were closer to the initial readings, as well.

In the second year, the researchers added a feature to the experimental group: Auditors got a bonus if their results were close to those found by the double-checkers, giving them more reason to be accurate. The result? A sharp increase in the pollution levels reported by the initial audit.

Overall, auditors in the experimental group were 80% less likely to report falsely that a plant was in compliance with air- and water-pollution rules than auditors in the business-as-usual group. Fortunately for the researchers, some auditors worked in both sets of plant—and the data show these same auditors behaved very differently depending on the circumstances.

“We were not very surprised,” says pollution-board administrator Hardik Shah, “but we didn’t have anything concrete to tell auditors and firms: You can do better.”

The Gujarat Pollution Control Board doesn’t care that the research will be published in the prestigious Quarterly Journal of Economics in the U.S. It cares about reducing pollution.

So the economists checked to see what the textile plants were doing. It turns out that plants in the experimental group reduced emissions of polluted water over the two-year period. The researchers surmise that plant managers changed their ways because they realized regulators were getting more accurate information from audits and that would lead to unwelcome punishment. The largest improvements came from the worst polluters, the ones most at risk of being shut by the authorities.

The pollution board is now moving to adopt some features of the experiment, developing software to randomly assign auditors instead of letting plants pick their own, planning to ban bargaining over audit fees and preparing to initiate random double-checking.

Auditing multinational companies is different than monitoring textile plant pollution, of course. Frequent rotation may be inefficient because it can take time for an accountant to understand a firm, though the U.S. Public Accounting Oversight Board has considered requiring it. Big audit firms’ need to maintain reputations can offset conflicts of interest.

But the lesson from Gujarat is clear: There’s a way to pay outside auditors so they have an incentive to tell the truth.