The financial crisis blew a hole in big-think economics, raising the profile of a new breed of skeptical empiricists committed to assiduous testing and tangible results, no matter how tiny. Even lentils can lead to little miracles
Capitalists have devised countless ways to separate people from their money. Big-box retailers employ "markdown optimization" software. Chain stores break out the new seasonal collections right when the IRS sends out its first refund checks. Apparel brands release special "payday" promotional codes to online shopping sites on alternate Fridays. The behavior of the flush consumer is tracked far and wide.
Until Esther Duflo and a network of like-minded professors came along, no one spent much time tracking the spending habits of, say, Western Kenyan maize farmers.
Where buying power is minimal and retail sales figures don't exist to be analyzed, consumer behavior would seem a fruitless area of study. Duflo, 38, is an economist at Massachusetts Institute of Technology's Abdul Latif Jameel Poverty Action Lab (J-PAL). Last year she received a MacArthur "Genius" grant, and in April she won the 2010 John Bates Clark Award, considered a stepping stone to a Nobel. Yet her work is so minutely focused that its importance is not easily grasped at first glance. In 2000, Duflo and colleague Michael Kremer embarked on an extended inquiry into what was essentially a sales and marketing mystery: Why didn't the maize farmers of Kenya buy more fertilizer? Calcium ammonium nitrate was widely known to be effective on maize, the Kenyan Ministry of Agriculture officially endorsed its use, and market surveys conducted by a Dutch nongovernmental organization suggested that the product was held in high regard among farmers. Most expressed their intention to use it in the future, yet only 29 percent had done so the season before. Fertilizer represented the most important, logical, life-improving purchase these poor farmers could make. So why were so few of them buying it?
Working with the Dutch NGO International Child Support, Duflo, Kremer, and their then-assistant Jonathan Robinson (now an assistant economics professor at the University of California at Santa Cruz) ran a three-year experiment comparing the output of fields where the fertilizer was used to fields where it wasn't. On average, the fertilizer proved effective, but the cost-benefit analysis varied widely among plots. The official treatment recommended by the Ministry of Agriculture was unprofitable for most farmers, whereas halving the prescribed quantity and skipping the recommended hybrid seeds yielded a steady 70 percent annualized return on investment.
The first trial yielded a convincing advertisement for the virtues of fertilizer; 97 percent of the farmers told researchers later that they planned to use it next season. Only 38 percent followed through, and the team from MIT had an epiphany: When the farmers said they were too poor to buy fertilizer, they actually meant that they were temporarily broke. The money that came in after the harvest had been spent on other things before the next planting season arrived. An economist would describe the farmers as "stochastically present-biased."
Duflo had an idea: What if they got farmers to commit to buying fertilizer right after harvest, when they were still flush, by offering free delivery on their fertilizer shipments? The catch was that they had to commit their funds then and there. To gauge the program's effectiveness, they also monitored a control group of farmers who received no special offers. Finally, later in the season, they offered another group of farmers a straight-up subsidy, a 50 percent discount.
The results were striking: About a third of control-group farmers used fertilizer that season, vs. 45 percent of the farmers offered free delivery, and 46 percent of the discount voucher recipients. There was virtually no difference between the demand for full-price and half-price fertilizer if the full-price fertilizer was sold at the right moment. So between two effective policies with identical outcomes, one might cost three times as much as the other, all because of an irrational—and utterly commonplace—quirk of consumer behavior: the failure to plan ahead.
Closing the Last Mile
The case of fertilizer and the Kenyan maize farmer is, Duflo points out, a classic "last mile challenge," a term often used to describe the unwieldy work of getting broadband cables and DSL service to every last cul de sac in every last Zip Code. The situation is the same in Duflo's world, where numerous technologies have failed to reach needy end-users. From fertilizer to vaccines to mosquito nets and hand sanitizer, scores of low-cost innovations persistently fail to save or alter tens of millions of lives, either because they never reach the people who need them or the people who need them don't use them.
Closing the last mile is a matter of perennial debate in development economics. On one side are the macro-interventionists like Columbia University's Jeffrey Sachs, who frame the alleviation of poverty in Africa as a straightforward matter of diverting such an infinitesimal fraction of Western wealth to the continent that it is fundamentally myopic (if not immoral) to count shillings. On the other are libertarian-leaning economists like Bill Easterly and Dambisa Moyo who favor profit-motivated investments, like those increasingly made in recent years by Chinese sovereign wealth funds, as the best way to help the developing world. The interventionist's approach to Kenyan fertilizer is simple: subsidize it, generously. The libertarians—with whom the World Bank has officially sided on the issue since the 1980s—regard subsidies as the opiates of economic development. Between these two warring camps, there's Duflo and J-PAL, miniaturists committed to neither ideology, out there asking West Kenyan farmers how much money they have from one week to the next.
The Birth of the Randomistas
MIT is the undisputed capital of the "Lab." There's an AgeLab and Air-Sea Interaction Lab, a Humans & Automation Lab and a Distributed Robotics Lab, a Complex Systems Research Lab and the SENSEable City Lab. The Poverty Action Lab joined this lineup with little fanfare in 2003, when Duflo and two colleagues—Abhijit Banerjee, her former faculty adviser, and Sendhil Mullainathan, a 2002 MacArthur grant recipient—procured a $300,000 university grant to launch a research network dedicated to promoting and running "randomized control trials" like the fertilizer experiments.
Since then the lab has grown to comprise 46 "affiliated" professors from about a dozen schools who have together embarked upon at least 200 randomized trials in 33 countries, mapping the irrational judgments and hidden inefficiencies that sabotage efforts to help the poor. The mission is to test out a range of possible solutions and determine the one that works best for the money.
The appeal of J-PAL's experiments is that something almost always wins, and the relative differences are often surprising. While the boilerplate sex education lecture never had any meaningful impact on the sexual behavior of Kenyan teenage girls, a more specific lesson on the risks of being seduced by "Sugar Daddy" types within the 25-35 age bracket resulted in a considerable drop in teenage pregnancy and sexual activity across all age groups. When an Indian NGO organized mobile immunization camps in 30 villages to compensate for India's chronically short-staffed clinics, a survey showed that kids from those villages were nearly three times as likely as kids from control group villages to have gotten all their shots. But in 30 other villages, in which the NGO organized camps and offered moms a free kilogram of lentils for every vaccination administered, kids were more than six times as likely to be up-to-date on their shots—and the kids in neighboring villages all had compliance rates more than twice the control villages. Good policy, it turns out, promotes itself.
It's quite possible that no previous MIT Lab has ever taken its favored brand of scientific methodology as seriously as J-PAL takes the randomized trial. Such is the intensity of their faith in it that they're easy to caricature and have inspired a fair number of detractors. What will these tiny policy adjustments add up to? Impoverished children with all their shots will give birth to a new generation of impoverished children if rural India's policymakers fail to nurture new industries. But optimizing industrial development is not a matter of choosing between a few simple incentive programs you can compare in a randomized trial. So the "randomistas," as Duflo and her cohorts are sometimes derided by traditional economists, ignore the subject. "J-PAL economists can be disturbingly monolithic about their commitment to randomized trials as the source of intellectual discovery," says Harvard University political economy professor Dani Rodrik.
If J-PAL economists can be slaves to their methodology, even critics concede such focus is an understandable reaction to an era in which all economics seems ideological. Facts—the duller and more incontrovertible the better—are a critical weapon in driving policy forward. Plenty of economists hail J-PAL for leading a data-driven credibility revolution in a profession in which, as the labor economists Joshua Angrist and Jörn-Steffen Pischke recently observed in a paper, "Hardly anyone takes anyone's data seriously."
That the policy implications of a multiyear, hyperintensive randomized trial can often seem comically minor doesn't bother the randomistas, many of whom have spent their careers watching other economists make bold, sweeping, but ultimately ineffectual policy prescriptions. Duflo describes the J-PAL philosophy as "come on, we're trying to do one thing here, let's just do it. There are worse ways and there are better ways, let's just compare and evaluate them and pick the best one. Ideology doesn't really matter so much when the objective is getting kids to show up for school or immunizing children."
J-PAL practices economics for economists who have lost their faith in economics. Even Rodrik concedes that J-PAL's "low-impact" methodology of studying the economy is a reasonable reaction to a profession characterized by "a series of snake oil salesmen."
Duflo was raised in the Paris suburbs; her father was a math professor and her mother a doctor who traveled regularly to Africa to volunteer in clinics. "I never experienced poverty—I grew up French middle class," she says. But from a young age she was made very aware of what poverty meant. "My mother would come back from Africa and show us slide shows of where she had been and what she had seen. She had this general sensitivity about it, the house was filled with books about poverty and public health, so that was kind of always in the background, and in the background I was always thinking, maybe I should try and do something about it, but not really having an idea how...."
Duflo was inspired to become an economist while working on an undergrad history thesis in Russia on Stalin's first Five Year Plan. It was 1993, and as she watched tanks storm the Parliament building to quell the rising dissent toward Boris Yeltsin's economic reforms, she had difficulty staying interested in Stalin and history in general.
At the time she was also working as a researcher for Jeffrey Sachs, then a celebrated Harvard economist on a USAID mission to advise the Russian government on managing the sudden transition from socialism to capitalism. Sachs had made his name administering a controversial brand of economic tough love dubbed "shock treatment" in Poland and Bolivia, but by 1993 he had concluded the Russian economy badly needed life support in the form of massive international aid. His pleas to wealthier nations fell on deaf ears, and as the economy worsened and chaos loomed, Sachs was increasingly drawn into vicious turf battles with his Russian-born Harvard colleague Andrei Shleifer. That fall, Boris Yeltsin sent tanks to storm the Parliament and quell dissent; by January, Sachs resigned and left Shleifer to helm the project.
"When the army attacked the Parliament, that was a really dark phase," remembers Duflo. "But at the same time, a lot of it was about economics and trying to get it right, and it collapsed with politics in a way that was interesting. Jeff Sachs had been making the point that much, much, much more money needed to be spent in Russia to avert a political crisis, and he lost that argument, but he may have been right. And I started thinking that economics as a discipline was a good way to understand the world and how all these things would play out. But I was also interested in a more extreme form, using what I learn to help Africa."
Duflo applied to economics graduate programs and enrolled at MIT the following fall. She knew she had made the right decision 10 minutes into her first class with Abhijit Banerjee, who had spent most of the previous decade formulating new theories for examining poverty to better account for behavior, psychology, negative feedback loops, and the quirks of the financial markets. "I was really in the right place at the right time," she says. "From the theory angle, a vast field was open for empirical work."
Banerjee had arrived in Cambridge as a 22-year-old Harvard PhD student in 1983, the same year as Sachs and Lawrence H. Summers, now director of the White House's National Economic Council, two of the youngest economists to receive tenure at the university. (Both were 28.) The laissez-faire Chicago School ideas popularized by Milton Friedman still dominated the field, but the clique of ambitious Harvard economists was determined to best them—and one another. Banerjee avoided the fray, though the Chicago School efficient markets dogma struck him, too, as preposterous. Markets, he believed, routinely failed for a variety of reasons, and he studied how extreme wealth and income disparities could lead to poverty traps that stifled the creation of a middle class. Corporations, he maintains, routinely undermine free-market competition. One paper Banerjee co-wrote with Summers investigated how even seemingly innocuous practices like frequent-flier programs exemplify systemic anticompetitive behavior.
Banerjee's work endeared him to Summers' protege Shleifer, a fellow skeptic of the efficient markets hypothesis. It was Shleifer who in 1991, freshly tenured and busy remaking the long-neglected development economics department, asked Banerjee if he'd try his hand at overhauling the curriculum. Shleifer would later be implicated in a scandal emanating from the consulting he did in Russia, precisely the kind of grandiose economics work that J-PAL would define itself in opposition to. "I suppose you could say we all share a sensibility formed in reaction to spending years in a field that is known to embody a certain, well...." Banerjee trails off. "I don't know if 'smug' is too cruel a word...."
The Next Generation
These days, J-PAL is nurturing a markedly different breed of economist, like Angela Kilby, 25, a research analyst who was planning to major in biology before she took Duflo's class in her sophomore year.
Kilby has been involved with one of J-PAL's most headline-grabbing innovations, X out TB, which distributes urinalysis strips among tuberculosis patients. When doused with urine containing tuberculosis-fighting drugs, the strips reveal hidden codes. Patients can then text the codes to their cell-phone providers and receive free minutes as reward for taking the medication.
Encouraging drug adherence is a sticky last mile challenge almost anywhere you go; some American clinics have taken to handing out computerized pillboxes that enter patients into a lottery in which they can win money when they remember to shake the box (and presumably, take a pill) once a day. TB is uniquely formidable because it requires a powerful cocktail of antibiotics that take at least six months to kill the disease—long after symptoms recede. As with any bacterial infection, a partial course of drugs is worse than no drugs at all, because it promotes antibiotic resistance, a big part of the reason TB now kills between 1.3 million and 2 million people every year.
J-PAL now attracts some of the brightest young minds in economics and is making its influence felt among the dominant institutions of the aid world, including the World Bank, which in 2006 asked Banerjee to serve on a panel that reviewed all the research it published from 1998-2005. Working with a team that included six other J-PAL economists, Banerjee encouraged researchers to be more transparent, noting wryly that "trade-offs tend to be eschewed in favor of ubiquitous 'win-win' scenarios, so that, for example, growth and environmental improvement are never seen as in conflict."
Sachs, too, has reached out to J-PAL. He has spent much of the past decade consumed with administering shock charity in Africa via the Millennium Villages project, in which 78 villages have received massive aid packages that effectively double the local economies overnight. Unfortunately for economists, Sachs didn't think about how the efficacy of the project could be measured until after all the villages had been blanketed with aid money. When he asked J-PAL last year for help evaluating the project, Duflo told Sachs that because the design of the project had not accommodated randomized testing, there was nothing she could do.