More Than Good Intentions

by Dean Karlan

  Soon after, I was invited to give an informal talk to the newly formed Henry E. Niles Foundation on the topic of what we knew about microcredit. Instead, I told them how little we knew, how even the basic workings of microcredit were still not fully understood. They were intrigued by the same puzzle I had been grappling with: How were microentrepreneurs borrowing (successfully!) at these interest rates with no business training at all? Their interest also helped to get the FINCA Peru project started.

  FINCA Peru’s borrowers were required to make weekly or monthly deposits corresponding to the size of their loans, and were also encouraged to hold voluntary savings in interest-bearing accounts. So there was some training, at least in the sense of behavioral conditioning: Clients learned about timely repayment and savings because they were required to do it. But there was nothing in the way of skills development, money management, or business or financial literacy education.

  Would expanding the training program help borrowers do better? To answer this question, Martín and I identified over two hundred existing FINCA Peru village banks in Lima and Ayacucho (a university town in the Andes) and randomly assigned half to receive thirty-minute business training sessions during their weekly meetings. This lasted for one to two years for each group. They learned basic business lessons, including orderly record keeping, understanding their markets, diversifying their inventories, and keeping business funds separate from personal funds. The other half of the groups continued to have meetings focused strictly on the loans, as usual. They were monitored for control.

  The results were not thrilling, but there were some bright spots. Clients in the treatment groups did adopt some of the strategies they had learned in the trainings. And their business revenues increased, particularly in bad months, though not by a lot. Microentrepreneurs everywhere face seasonal fluctuations in their business revenues, from changes in both supply and demand. Sometimes the stars align for a good month, when inventory is cheap and there always seems to be a line of customers around the block. But things can go badly too. There is the month when nobody is buying anything, when school fees or annual rent payments are owed, when the wholesaler raises his prices, when the flu gets passed around. As a result, this month the kids have to skip school to help out at the shop. Or this month they miss meals because income is down.

  But if you received the training and changed some business practices, maybe you manage to avoid these bad outcomes. Clients who received training implemented strategies to guard against seasonal swings. Of the few differences between treatment and control, this was the largest, suggesting that the training’s impact, while mediocre across the board, was concentrated where it was needed most.

  FINCA Peru was glad to see its clients doing better, and it might have continued offering business training for that reason alone. Happily, looking at its own bottom line only made the decision easier. Even after accounting for the price of running the training sessions, the program amounted to a net gain for the bank, because clients receiving training were more likely to repay their loans on time, and less likely to drop out of the lending program. FINCA Peru took this good news and ran with it. Soon after the study ended they mandated training for all their clients.

  But could it have been better? IPA works constantly to refine and replicate, testing ideas in different settings and with different people. This is essential if we hope to break through from particular findings to general lessons about what works. In a second study, led by Antoinette Schoar of MIT Sloan School of Management (and now the managing director of IPA’s Small and Medium Enterprise Initiative), Greg Fischer of London School of Economics (and a board member of IPA), and Alejandro Drexler of the University of Texas-Austin, a microfinance program in the Dominican Republic tried to put a finer point on the impacts of business training for its clients.

  Rather than just testing one training module, they tested two, and compared each to a control group. They found that standard accounting training did not work so well. But specific “rule of thumb” training, which gave microentrepreneurs simple heuristics for keeping track of money, did. In fact, the effect of the “rule of thumb” training was identical to what we had seen in Peru: Individuals found ways to smooth out their income so that bad months weren’t so bad anymore. They didn’t suffer so much when times were lean.

  But neither of these studies found training recipients making the leap from microbusiness into small and medium enterprises. No transformations. No picture-perfect stories that jibe with the publicity we hear about microcredit. So Miriam Bruhn of the World Bank, Antoinette Schoar, and I joined forces on a project in Mexico that focused on individually tailored training, more like business consulting, for small and medium enterprises in this study. Here, individual mentors were assigned to small and medium enterprises not to teach basic skills, but to get to know the businesses and entrepreneurs closely and advise them on how to improve. The program, funded by the state government, primarily sought to increase employment. It did not achieve that goal. But profits more than doubled for the firms in the program, rising by 110 percent!

  What is the lesson? Between the Peruvian, Dominican, and Mexican examples, it appeared that training featuring immediately practicable, concrete lessons, and more intensive, personalized consulting-style mentoring worked better than generalized skills training. Naturally, more intense training is also more expensive, but in the Mexican study the increase in profits outweighed the higher price. Overall it was even more cost-effective than FINCA Peru’s training program.

  These studies show that microentrepreneurs can improve their businesses through training, but the bigger takeaway was the evidence that they had something to learn in the first place. As I said before, the fact that the poor are not all innately top-flight entrepreneurs should not come as a big surprise—but, judging from Muhammad Yunus’s words, it’s something many microcredit advocates need to hear.

  Not everyone (in developing countries, or anywhere else, for that matter) is cut out to run a business—or to take on entrepreneurial debt. For some it’s because they lack expertise or aptitude, but for most the answer is probably simpler. They are not great entrepreneurs because being great entrepreneurs isn’t their main goal in life. People pursue happiness in other ways: doing work they enjoy more, spending time with their families, watching movies with their girlfriends in the afternoon.

  What happens when the commonsense fact of people’s varying abilities and priorities collides with the worldwide enthusiasm for entrepreneurial microcredit? You end up trying to fit some square pegs into some round holes, lending to people who aren’t going to succeed. You make loans and watch hopefully, waiting for the businesses to sprout up like so many blades of new grass; but what ends up growing is not a uniformly lush, verdant carpet of green. There are some bare spots.

  The Pursuit of Rice Cookers

  When you get down and poke around in that bare dirt, you notice that some of the seed you thought you had scattered there never made it into the ground. Some microcredit clients don’t even appear to have tried to build enterprises with their loans. Lenders—and donors—often get irritated when they hand out money for one reason, only to see it used for something else entirely.

  Jake can personally attest to this. While he was living in Ghana, he befriended a man named Philip, who had a way of getting himself into tight spots.

  One day as they were walking to lunch, Philip said he needed help. He had rented a room he couldn’t afford, with the intention of staying just a couple weeks while he found cheaper lodging. But during that time he had run up a rent bill that already exceeded his savings, and the owner of the room would not let him check out for fear that he would disappear. That meant Philip was digging himself further and further into debt with each passing night.

  “Jake,” he said, “the way it has come down, I have to ask you for support. If you can help me, then by all means I will settle the balance on my bill at the guesthouse, and I will pay you back at the earliest wh
en I take my salary.” Jake was doubtful. He’d lent Philip money in the past and not been repaid, and he wasn’t keen on getting burned again.

  The tension grew in the afternoon, when two uniformed police officers appeared in the office where both Jake and Philip worked. They stood in front of Philip’s desk and asked him to step outside. Philip went quietly and returned about twenty minutes later. He came straight to Jake’s desk. “See, Jake?” he said. “It’s serious.” Jake gave him the money the next morning.

  About two weeks later, around the time when paychecks went out, Jake asked how things were going. Philip seemed upbeat. “I am out of that guesthouse. I wouldn’t let them catch me again,” he said, shaking his head and pantomiming grabbing an animal by the scruff of its neck.

  “And did you settle the whole bill? You don’t owe anything more to them?”

  “Well, there is some small balance left, but for that they wouldn’t chase me.”

  “A balance?” Jake had lent Philip enough to repay his whole debt. Where had it gone, if not to the owner of the guesthouse?

  “Well,” Philip said, looking away, “I also bought a rice cooker. Now I can cook on my own.”

  This kind of scenario infuriates donors. We dig deep into our own pockets to help a guy like Philip with his rent and he turns around and buys a kitchen appliance. Jake dressed Philip down.

  A picture of composure, Philip held his ground. He smiled and heaved a tired sigh: “I knew you would be upset with me. But you don’t understand how it is with this guesthouse man. Once I gave him something, I knew he would not make trouble again for some few weeks. I can pay him the rest from my salary check.”

  Philip had gone back on his word. He really had said he would use the money to pay his rent.

  But then his way of dealing with the world did work in its way. Neither the police nor the guesthouse proprietor was seen or heard from again, and Philip enjoyed many heaping bowls of perfectly cooked white rice in his new room.

  Slippery Money

  Money is fungible, as economists say. It is slippery. It moves like mercury on a tabletop, sliding effortlessly from place to place and leaving no residue. If Jake had cut a check directly to the guesthouse proprietor, things might have been different. But cash—unlike, say, a coupon—is not tied to any particular person, product, or store. As Philip so ably demonstrated, it can be spent on anything. Short of tracking the serial numbers on bills—or physically tailing somebody around to watch him spend the money, like the researchers in the Sri Lanka study we saw last chapter—it is virtually impossible to follow any particular wad of it as it passes from hand to hand. (Besides, as we’ll see later in this chapter, tracking the particular wads of cash we lend out doesn’t necessarily tell us what we need to know.) So when we make rules or restrictions about the use of a cash loan or donation, we typically have no more than the recipient’s word that they will be followed.

  The question of right and wrong here is complicated. Organizations who act like Jake did, stipulating that aid money be spent in a particular way, often do so with good intentions. I have seen some microlenders require, for example, that clients bring in receipts for investments that match their businesses. Still, recipients often know more about their immediate and changing needs than anyone else. Philip did. So did the borrowers we saw last chapter in the Credit Indemnity study, who made themselves better off with unrestricted loans.

  There’s a larger point here than who’s right and wrong. In fact, there are two larger points—we’ll come to the second one later. The first is this: When we insist that microloans be spent on microenterprises, and then we ask people how they spent their last loan and how they plan to spend the next one, we shouldn’t be surprised to find a lot of lying. If people were always straightforward about their intentions, many would never qualify for assistance. Many potential donors (Jake included) would hesitate to lend money to Philip if we knew how he actually planned to use it, or even if we believed he could decide later how he would spend it. In cases like these, where a person’s eligibility depends on their willingness to commit to behaviors we cannot (or will not) monitor or enforce, it’s hard to see the merit in forcing the issue. Aren’t we just angling for false promises?

  If we are, then we’re ultimately hurting ourselves. When people cannot—or choose not to—be honest about how they put resources to use, we get the wrong ideas about how these programs actually function. That was exactly the upshot of the evaluations we saw last chapter, which began to tease out the true impacts of microcredit. They suggested that our picture is incomplete. If we really want to make microcredit work for the poor, we cannot delude ourselves into thinking that all microloans go toward enterprise investments. For how can we hope to fix a machine when its inner workings look nothing like our schematic?

  Getting the Truth

  This is a critical point in our approach to addressing the problem of world poverty. If we want to act with more than good intentions, we must have an accurate picture of the process of development, and of the specific ways it improves—or fails to improve!—lives. Thanks to the combined tools of behavioral economics and rigorous evaluation, this is actually within reach.

  Regarding the use of borrowed money, there is a clever way to find out how people are really spending their loans without forcing them to fess up directly. The trick is realizing that people are willing to reveal sensitive truths as long as they can hide them in a cloud of mundane ones. So rather than asking a touchy question point-blank, we can imbed it in an innocuous list.

  It works as follows: Imagine you wanted to find out whether people had been stealing Milky Way bars from the corner store. You could just ask—but you probably wouldn’t be surprised if everyone said no. And you would be right to doubt their answers. Instead, make two lists of statements, hand one (randomly chosen) to each customer, and ask: “How many—not which ones, but just how many—of the following statements are true?”:

  List 2 gives Milky Way thieves the cover they need to admit their actions without fear of being found out. Suppose a customer (who knows himself to be a thief) is presented with the second list. He says he agrees with two of the four statements. You can’t peg him for stealing because he might lie and say he was only agreeing, for instance, with statements (2) and (3). But the information is in there; we just need to extract it.

  Randomization is precisely the tool for the job. Because the customers were randomly assigned to lists, there should be no systematic differences between those who got List 1 and those who got List 2. In particular, they should not differ (on average) in their agreement with statements (1) through (3), which are common to both lists. That means the average agreement with all of List 1 is the same as the average agreement with just statements (1) through (3) of List 2. Subtracting that from the average agreement with all of List 2 gives us what we want. The quantity we care about—average agreement with statement (4), or the portion of customers who are Milky Way thieves—is exactly what’s left over when we take away agreement with (1) through (3).

  This technique reveals very clearly what the group is doing without exposing the behavior of any individuals. And it can do more than help solve the case of the missing Milky Ways. Pia Raffler, former Uganda country director for IPA and now a Yale political science Ph.D. student, Julian Jamison, an economist at the Federal Reserve Bank of Boston (and someone who considers a marathon a “training run”), and I used it in Uganda, where we were evaluating a Grameen Foundation and Google program that provided answers to health questions by text message. We wanted to know about people’s sexual behavior—specifically on the sensitive topic of infidelity—but of course we knew folks might not tell the truth when asked directly. When we did ask directly, 13.3 percent of respondents admitted to having been unfaithful in the past three months. But when the infidelity question was embedded in a list, we found that 17.4 percent of respondents—about a third more—had been sneaking around.

  Jonathan Zinman and I used the same tech
nique in a project with the Peruvian microfinance organization Arariwa to learn what clients were actually doing with their loans.

  According to Arariwa’s rules, loan funds could only be used to invest in a business. If a borrower admitted to spending her loan on food, medicine, school fees, or any other kind of consumption (as opposed to investment), she would likely be barred from borrowing in the future. Still, the bank wanted to know where all the loan money was going. So it asked.

  To hear the borrowers tell it, almost everything was—predictably—aboveboard. When questioned directly about what needs their loans had covered, 8 percent said they had spent some of their loan on household goods. Another 7 percent admitted to spending on their children’s education. And a measly 2 percent claimed to have spent on health care. Everybody else, apparently, had followed the bank’s instructions to the letter and eschewed consumption entirely.

  Which would have been great, except that it was not true.

  Using the list randomization approach yielded drastically different answers. Once the sensitive questions (like “Did you spend part of your microloan money on household goods?”) were surrounded by easy ones (like “Did you spend part of your microloan money on business supplies?”), we started to see the real picture. Now it looked like 32 percent of borrowers had used some loan monies for household goods, 33 percent for children’s education, and 23 percent for health care.