More Than Good Intentions

by Dean Karlan

  Let’s talk about basics. People need to eat. And that means sometimes we have to give them food. People need medicine. That means sometimes we have to hand out pills and give immunizations. People need to go to school. That means sometimes we need to get students and teachers into classrooms.

  Addressing world poverty is a dynamic, complex problem. But we won’t solve it if we see it only as that.

  We need to see individuals. Individuals with different capabilities and different needs. Individuals like Vijaya, whom we’ll meet in chapter 7. What she really needs is a way to stop her husband from drinking away the money she earns. Individuals like Elizabeth, whom we’ll meet in chapter 10. What she really needs is better customer service from her local hospital.

  When we think about poverty this way, in concrete terms, we begin to see a path forward. Actually, many paths forward. The possible solutions are as numerous and as varied as the people they serve and the needs they address. To find them we need to think creatively, cast a wide net, and recognize that we are unlikely to find a single answer for everybody. At the same time, we must be methodical and tenacious. If a development program is supposed to help solve a specific, concrete problem, let’s put it to a specific, concrete test. If it passes, great. If not, fix it or try something else. In this way, step by step, we can refine the tools we use and the ways we use them; we can make real progress against poverty.

  3

  TO BUY

  Doubling the Number of Families with a Safety Net

  According to archaeologists, the use of blankets began during the reign of Neanderthal man. Which means that, in some sense, the Snuggie was thirty thousand years in the making.

  Down through the ages, all the great thinkers and creative geniuses of human history muddled through with the same old flat-as-a-pancake bedcovers, leaving the landscape of blanketry more or less as they had found it. Then, in 1998, there was a breakthrough.

  Gary Clegg, a freshman at the University of Maine, was set upon by the harsh New England winter. Even in his own dorm room, he couldn’t do his homework. He was cold just sitting at his desk. Regular blankets helped, but they were cumbersome and restrictive. So he asked his mom to make him a blanket with sleeves. The first version wasn’t perfect, but successive prototypes got better and better. By the spring thaw, the Slanket was born.

  Oblivious, most of the world did not take notice; the sleeved blanket languished in obscurity for a decade. Few realized what they were missing until 2008, when a campy commercial for the Snuggie, a copy of the Slanket, began to air on late-night television. In the space of two minutes, it articulated a thorny and pervasive problem and presented an elegant solution. “You want to keep warm when you’re feeling chilled, but you don’t want to raise your heating bill. Blankets are okay, but they can slip and slide. And when you need to reach for something, your hands are trapped inside. . . . The Snuggie keeps you totally warm and gives you the freedom to move your hands. So now you can work the remote or read a book in total warmth and comfort.” Spoof ads on YouTube also helped spread the word.

  Finally the masses knew. Mankind had been brought to the threshold of a new era. Now it had only to leap.

  And leap it did. Four million Snuggies were sold in the first year. Snuggie fan clubs sprang up by the hundreds. There were Snuggie pub crawls. The cast of Good Morning America did a show wearing Snuggies. As of February 2010, the number of sleeved blanket users worldwide is estimated at twenty million and rising every day. It is nothing short of a revolution.

  Of course, you could be a cynic and go around taking the wind out of everybody’s sails. You could say that the Snuggie isn’t a revolution, but a cheap, thin blanket with two holes cut in it and stovepipe sleeves attached. And you might be right. But who cares? The people have spoken. (The earlier Slanket continues to sell well, but not as well as the Snuggie.)

  You Can Sell Anything

  Advertisers have a saying: There’s no such thing as a bad product, only a bad salesman. We saw in the introduction that, in our donating as much as in our consumption, we respond to the suggestive power of marketing—often to the exclusion of the facts about the thing being sold. When that happens, quality and popularity can diverge. Just because something is good for you doesn’t mean people will buy it (think lima beans); and just because people buy something doesn’t mean it’s good for you (think cigarettes and salty french fries).

  Sophisticated businesses understand this and act accordingly: In the United States alone, companies spent about $412 billion on advertising in 2008.

  But there is a strange disconnect between the way we sell everyday products at home and the way we sell development solutions abroad. Namely, we often don’t think we need to sell development solutions at all, but rather expect them to be adopted on their merits alone. (Note that this approach has not worked very well for lima beans.)

  This is shortsighted. It ignores the fact that development is a two-way street. If we want to help the poor by offering programs and services, two things have to happen: First, we need to make programs and services that work; second, the poor have to choose to sign up for them. Or, in the case of rainfall insurance policies, microloans, and prepaid fertilizer coupons (all of which are among the examples we’ll see later on in the book), they have to buy them.

  In recent years we have begun to make some headway on the first part by coordinating the efforts of researchers and practitioners to rigorously evaluate development programs. But we’re really lagging behind on the second. In some sense, the more we learn about what works, the more we need to get the marketing right—because letting a proven-effective program fail due to lack of interest is a wicked waste.

  A significant portion of advertising money is spent on making good first impressions. This is something development organizations need to think about when they launch brand-new products. If they get the marketing right, they have the potential to generate Snuggie-like levels of excitement.

  The Last-Mile Problem

  The Snuggie is a textbook case of an unknown bursting onto the scene and making a splash. But many of the programs you’ll see in this book are not quite analogous to the Snuggie. They’ve been around, and people know they exist. This is both an advantage and a disadvantage. Familiarity breeds awareness, but it also breeds numbness. To borrow an advertising term, the products don’t pop.

  A prime example is oral rehydration therapy, a dirt-cheap and highly effective treatment for diarrhea. It is a small plastic envelope of salts that, when eaten, allow the body to absorb and retain water. Combined with fluid intake, it effectively neutralizes the threat of mortality from the disease. The salts cost a couple of pennies at most, and in many diarrhea-prone areas of the developing world they are fully subsidized—available for free.

  A cheap and proven cure (with no side effects, by the way) for a deadly disease would seem to sell itself, but the sad fact is that it doesn’t. Nearly two million people, mostly children, die every year from diarrhea. Either they don’t know about the salts, or they don’t want them. Either way, it means we are failing.

  Fortunately, we don’t have to look very hard to find ways to improve on the marketing front. We are constantly bombarded by examples—on the Web, in billboards, in magazines, on television, and on the radio. And in the grocery store, where lima beans still languish more or less unloved, but where there’s a lesson to be learned from the equally humble raisin.

  In biological terms, 1986 was an unremarkable year for the California raisin. It was a dried grape all the way through, from the first day to the last. There were really no developments to speak of. Nor were there any big changes in the availability of California raisins to the American public. They were sold in most grocery stores around the country, as they always had been. There was no scientific discovery that showed raisins to be a miracle food with previously unknown health benefits—they continued to be a reasonably healthy snack option—nor is there any evidence that the nation’s collective palate c
hanged during that short span.

  Nonetheless, 1986 was a turning point. In the words of the raisins’ primary advocacy group, the California Raisin Advisory Board, they were “at best dull and boring” at the beginning of the year; by the end, people were “no longer ashamed to eat [them].” Now, industry groups may be prone to hyperbole, but in this case the proof was in the pudding. And the pudding was full of raisins. For the rest of the decade, sales increased by 10 percent.

  As the Advisory Board’s words implied, the surge in sales had little to do with California raisins themselves, and everything to do with the way the public thought about California raisins. Which, in turn, had everything to do with the California Raisins, a quartet of singing Claymation raisins that burst onto American television screens in 1986, brandishing electric guitars and sporting cool sunglasses, and singing “I Heard It through the Grapevine.” If you remember the California Raisins—and I know many of you do—then you’re living proof: It was a stroke of marketing genius.

  Almost overnight there was a profusion of fan clubs, T-shirts, lunch boxes, and, most important, raisins.

  Sendhil Mullainathan, one of my Ph.D. dissertation advisers at MIT, coauthor of some of the work discussed in this book, and certified MacArthur Foundation “genius,” has thought and written a lot about this issue (though not typically in terms of Motown and dried fruit). He calls it the Last-Mile Problem. It goes as follows: Faced with a stubborn challenge, we employ brilliant minds and vast resources to design a solution. We combine science, engineering, creativity, and careful testing, and often we succeed in solving the technical problem—thus completing 999 miles of a thousand-mile journey. Then, inexplicably, we pack it in. Instead of taking the same rigorous approach to adoption, we just put the solution out there and expect it to speak for itself. All too often—as in the case of oral rehydration salts, an example Sendhil uses in many of his talks on the subject—it doesn’t.

  So let’s put it bluntly: We need to learn from the Snuggie and the California Raisins.

  How Much Is a Photo of a Pretty Woman Worth?

  Part of the problem is that economists aren’t trained to think about the last mile. Take the example of credit: How does a person decide whether to borrow money? In undergraduate and graduate economics courses, students learn models of borrowing that consider the interest rate, the person’s investment opportunities, and the rate at which he values current versus future consumption.

  It all makes analytic sense, but it’s extremely limited. Models are just equations; they can’t see or say anything beyond the variables they comprise. So when a model with these three inputs is used to design a loan product, it spits out a recommendation about those three parameters and nothing more.

  In South Africa, I set out with Jonathan Zinman, a friend and classmate from MIT, to work on a nuts-and-bolts question—exactly the kind of thing our standard economic models are designed to talk about—and ended up learning even more interesting things about the last mile. We wanted to understand how borrowers reacted to different interest rates, so we partnered with a local consumer lender called Credit Indemnity (which has since been bought out by a larger bank) and designed an RCT.

  We were excited to sink our teeth into an important policy issue that has been hotly debated in microcredit circles over the years. (In fact, the dearth of evidence on exactly this issue was one of the main things that had set me on the path toward development economics a decade earlier.) One of the key questions we wanted to answer was whether higher interest rates led to higher default. To find out, we needed a big study in which we offered people different interest rates on loans, and we needed a lot of people to borrow in order to have enough data to answer the question.

  So we had to court a boatload of potential borrowers. We settled on a direct-mail campaign to some fifty-three thousand of Credit Indemnity’s current and former clients. When we sat down with their management to talk about designing the campaign, we asked them what they knew about how to generate the highest response rate to a direct-mail solicitation. As it turns out, they had not done prior tests—so they were full of questions, just as we were. All of a sudden, the study on interest rate sensitivity was a marketing study too.

  Marianne Bertrand, Sendhil Mullainathan, and Eldar Shafir think a lot about exactly these issues of psychology and economics. After the discussion with Credit Indemnity, I was visiting Marianne and Sendhil in Chicago, and we sat down to brainstorm about how we could boost the response rate to direct-mail marketing. As is typically the case with them, five minutes of talking led to ten ideas for potential improvements to the direct-mail solicitations. The ideas were interesting, but we were all struck by the fact that we were just spinning our wheels. There simply was no good data from the “real world” to guide or justify our thinking. (Which is not to say marketing firms don’t do RCTs themselves; they do. In fact, they do tons of them—but they don’t typically share the results with dorky academics like us, nor do they design them in ways that test specific theories of human behavior.)

  This ignorance became the killer insight: What does actually work? And how important are subtle marketing features, relative to the most important factor in our traditional model—the interest rate?

  To pit the marketing tweaks against the interest rate, we had to vary both. So we got the most current flyer from Credit Indemnity and started tinkering.

  In addition to substantive product features, like the interest rate and the application deadline, we varied purely presentational features of the mailer. Should the flyer show a picture of a pretty woman, and if so, what kind of pretty woman? South Africa has a long history of racial issues; would people respond better to a photo of someone of their own race? Would proposing uses for the loans or presenting more example loans (suggestions of how much one could borrow and how long one could borrow for) entice customers? How about displaying the interest rate in different ways, or showing competitors’ rates?

  Putting the variations together, we produced dozens of different flyers and assigned them randomly to the fifty-three thousand names on the mailing list. Months later, when all the application deadlines had passed, we could see which flyers had brought in the most customers.

  The first thing the data showed was that, across the board, customers clearly cared about interest rates. As a standard model would predict, they were significantly more likely to apply for low-rate loans. What was surprising was how much they appeared to care about things besides the price.

  Two marketing features—photographs of pretty women and the number of example loans—proved influential although they had nothing to do with the actual terms of borrowing. From the perspective of classical economic theory, this is strange: Surely no customer would say that his decision to borrow boiled down to the picture in the corner of his pamphlet, but there it was in the data, clear as day. In terms of generating applications, adding a picture of an attractive woman to the flyer had the same effect on men as lowering the loan’s interest rate by 40 percent!

  The response to example loans, a simple table breaking down the monthly payments for a few different loan amounts, was surprising for two reasons. First, flyers with four example loans in the table attracted far fewer applicants than flyers with just one, suggesting that presenting more options actually drove away customers. This directly opposes standard economic theory, which maintains that having more choices is always better for the chooser.

  The second surprising result from the example loans table was just how strong this choice aversion appeared to be. Showing one example loan instead of four attracted as many additional applicants as dropping the interest rate by about a third.

  If I had doubts that marketing could make a difference in the developing world, the South Africa study put them to rest. When simple changes to a promotional mailer (like cutting out three rows on the table of sample loans) generate as much new business as drastic price cuts, you can’t afford to ignore it.

  Now, knowing that market
ing matters is one thing; knowing exactly which changes to make to a promotional mailer is far from straightforward. The hardest part about this study was predicting what would work and what would not. (In fact, before the study began, we guessed at the impact of each marketing tweak—and many of our guesses were wrong.) Race, for instance, has always been a hot-button issue in South Africa; but customers didn’t respond any differently when we varied the race of the person photographed in the flyer. Similarly, many businesses in South Africa ran “cell-phone giveaway” raffles. Assuming that the marketing experts were onto something, we tested it on some flyers. But it didn’t do any good in our test. It actually dampened the response.

  The results that did stand out in South Africa—especially the aversion to more example loans—pointed clearly toward behavioral economics.

  Too Many Choices

  Recent behavioral research has shown traditional economics’ more-choices-are-always-better rule to be far from universal. Sometimes options can paralyze. When they are too numerous or too hard to compare, we often just procrastinate: “This is a lot to think about right now; I’ll get to it tomorrow.”

  People have recognized this tendency in their daily lives for a long time—maybe for as long as they have been making decisions—but nobody put a fine point on it until recently. Behavioral psychologists and economists dubbed it, appropriately, “choice overload,” and set out to measure it.

  In 2002 Sheena Iyengar, a social psychologist at Columbia University (and author of the recent book The Art of Choosing), and Mark Lepper, a psychologist at Stanford University, did a choice experiment at a fancy grocery store in California. They set up a table where shoppers could taste exotic jams. Each person who stopped at the table was allowed to try any number of jam flavors and was given a coupon for a dollar off a jar of their choice. Iyengar and Lepper wanted to see whether choice overload afflicted even casual shoppers, so hour by hour they changed the number of flavors available to sample, from six to twenty-four.