This teacher was not an exception. We have seencountless examples of this kind of elite bias amongteachers in developing countries. In collaborationwith Pascaline Dupas and Michael Kremer, Estherhelped design a reorganization of Kenyanclassrooms, taking advantage of an extra teacher todivide the class in two. Each class was separated byprior achievement, to help children learn what theydid not know yet.

The curriculumand organization of schools often date back to a co-lonial past, when schools were meant to train a localelite to be the effective allies of the colonial state,and the goal was to maximize the distance betweenthem and the rest of the populace. Despite the in-flux of new learners, teachers still start from thepremise that their mandate remains to prepare thebest students for the difficult exams that, in mostdeveloping countries, act as a gateway either to thelast years of school or to college

In other words, parents see an S—shapewhere there really isn’t one.This belief in the S—shape means that unless par-ents are unwilling to treat their children differentlyfrom one another, it makes sense for them to put alltheir educational eggs in the basket of the child theyperceive to be the most promising, making sure thatshe gets enough education, rather than spreadingthe investment evenly across all their children.

The patternwas even more extreme for girls. In the view of par-ents, each year of primary education was worth al-most nothing for them: 0.4 percent. But each yearof secondary education was perceived to increaseearnings 17 percent.In reality, available estimates show that each yearof education increases earnings more or less pro-portionally. 26 An

He thought that the mostprobable outcome was that after completing sec-ondary school, the boy would get a job in the nearbymall, where his brother was already working. This isa job that he could have had already—but neverthe-less, Pak Sudarno thought it was worthwhile for himto complete secondary school, even if it meant threeyears of forgone salary.

Pratham was running what they called the Balsakhi(meaning “children’s friend”) program. The pro-gram took the twenty children in each classroomwho most needed help and sent them to work withthe balsakhi, a young woman from the community,on their specific areas of weakness.

One study 19 found that anexcellent predictor of the supply of private schoolsin a Pakistani village is whether a secondary girl’sschool had been set up in the area a generationearlier. Educated girls, looking for an opportunity tomake some money without having to leave the vil-lage, were increasingly entering the education busi-ness as teachers

There is a surprising amount of agreement thatprivate schools should play an important role in theprocess of filling the gaps in the education system.India’s Right to Education Act, which was recentlypassed with strong support across the political spec-trum (including the left, which, the world over, hastraditionally opposed the role of the market), is aversion of what is called voucher privatization—thegovernment gives citizens “vouchers” to pay private-school fees

She found that relativeto the older generation, the wages of the youngerone were significantly higher in areas where moreschools were constructed.

It was the classic top-downsupply-driven program: Schools were built based ona prespecified rule that gave strict precedence toareas where the number of unschooled children wasthe highest. If the lack of schools in this area reflec-ted lack of interest in education, this programshould have been a miserable failure.

In The Elusive Quest for Growth, Easterly argues,for example, that the investment in education inAfrican countries has not helped these countries togrow

Unless we can fullyerase differences in income, public supply-side in-tervention that makes education cheaper would benecessary to get close to the socially efficient out-come: making sure that every child gets a chance.

aradoxically, it was one of these replications, inMalawi, that led us to rethink the success ofPROGRESA. The conditionality in PROGRESA isbased on the principle that increased income is notenough and that parents need to be given an incent-ive. Researchers and practitioners started to askwhether an unconditional program could have thesame effect as a conditional transfer. A World Bankstudy found, provocatively, that conditionality doesnot seem to matter at all: The researchers offeredthe families of school-age girls a transfer rangingbetween $5 and $20 USD PPP per month. In onegroup, the transfer was conditional on enrollment.In another, it wasn’t. A third group (the controlgroup) did not receive a transfer.

Thisinspired the design of PROGRESA, a transfer pro-gram “with strings attached.” PROGRESA was thefirst conditional cash transfer (CCT) program: Itoffered money to poor families, but only if theirchildren regularly attended school and the familysought preventive health care.

“Exactly,” says the supply wallah. “This is whysome parents need a push. A civilized society cannotallow a child’s right to a normal childhood and a de-cent education to be held hostage to a parent’swhims or greed.”

In other words, al-though the economic return to education (as meas-ured by the extra earnings of an educated child)clearly matters, lots of other things probably matteras well, things like our hopes about the future, ourexpectations about our children, even how generouswe are feeling toward them

At the core of the demand wallahs’ view is the ideathat education is just another form of investment:People invest in education, as they invest in any-thing else, to make more money—in the form of in-creased earnings in the future. The obvious problemwith thinking of education as an investment is thatparents do the investing and children get the bene-fits, sometimes much later. And though many chil-dren do, in effect, “repay” parents for the invest-ment by taking care of them in old age, many othersdo so only reluctantly, or they simply “default,”abandoning their parents along the way.

For the “demand wallahs,” a set of critics (includingWilliam Easterly) who believe that there is no pointin supplying education unless there is a clear de-mand for it, these results encapsulate everythingthat has been wrong with education policy in thelast few decades. In their view, the quality of educa-tion is low because parents do not care enoughabout it, and they don’t because they know that theactual benefits (what economists call the “returns”to education) are low

Close to 35 percent of children in the seven-to-fourteen age group could not read a simpleparagraph (first-grade level) and almost 60 percentof children could not read a simple story (second-grade level). Only 30 percent could do second-grademathematics (basic division). 6 The math results areparticularly stunning—all over the Third World,little boys and girls who help their parents in theirfamily stall or store do much more complicated

Pratham, an Indian NGO focused oneducation,

In our eighteen-country data set, even among theextremely poor (those who live on less than 99 centsa day), enrollment rates are now above 80 percentin at least half the countries for which we have data.

A large majority of policy makers, at least in in-ternational policy circles, have traditionally takenthe view that the problem is essentially simple: Wehave to find a way to get the children into aclassroom, ideally taught by a well-trained teacher,and the rest will take care of itself. We will call thesepeople, who emphasize the “supply of schooling,”the “supply wallahs,” appropriating the Indian termfor “purveyor of” (as in the western Indian sur-names Lakdawala [wood supplier], Daruwala[booze supplier], and Bandukwala [gun seller])

Those two children had been en-rolled at the government school, but they had bothrun away countless times before their mother aban-doned any hope of being able to make them attend.The ten-year-old boy, who was with his motherwhen we interviewed her, mumbled somethingabout school being boring

Small incentives, like giving dal forvaccines, are another way to nudge people, by givingthem a reason to act today, rather than indefinitelypostpone.

The poor in Delhi spend asmuch on short-duration ailments as the rich, butthe rich spend much more on chronic diseases

In other words, it is important to keepdoing something about your health, even if youknow that you are not doing anything about the bigproblem.

Even if the antibiotics did nothing to cure the ail-ment, it is normal to attribute any improvement tothem. By contrast, it is not natural to attribute caus-al force to inaction: If a person with the flu goes tothe doctor, and the doctor does nothing, and the pa-tient then feels better, the patient will correctly inferthat it was not the doctor who was responsible forthe cure.

Faith, or to use the more secular equivalents, acombination of beliefs and theories, is clearly a veryimportant part of how we all navigate the healthsystem. How else do we know that the medicine thatwe were prescribed will make the rash better andthat we shouldn’t apply leeches instead?

In fact,there is a “psychological sunk cost” effect—peopleare more likely to make use of something they havepaid a lot for. In addition, people may judge qualityby price: Things may be judged to be valueless pre-cisely because they are cheap

The overall sense we get from their study abouthealth care in India, both public and private, isfrightening. Das and Hammer describe it as the3-3-3 rule: The median interaction lasts threeminutes; the provider asks three questions and oc-casionally performs some examinations. The patientis then provided with three medicines (providersusually dispense medicine directly rather than writ-ing prescriptions).

There was also a clear pattern in the errors: Doc-tors tended to underdiagnose and overmedicat

However, this kind of self-awareness is unfortu-nately not universal. In urban Delhi, Jishnu Dasand Jeff Hammer, two World Bank economists, setout to find out what doctors actually know

In the slums of Delhi, astudy found that only 34 percent of the “doctors”had a formal medical degre

More than one-half were to private facilit-ies. The remainder were to bhopas—traditionalhealers who primarily offer exorcism from evilspirits.The poor in Udaipur seem to select the doubly ex-pensive plan: cure, rather than prevention, and curefrom private doctors rather than from the trainednurses and doctors the government provides forfree.

At some point, PSI started distributingsubsidized, but not free, nets in the same clinics.Cohen and Dupas wanted to find out whether theirorganization was still needed. They set up a simpletest: They offered nets at various prices in differentclinics, chosen at random. The price varied fromfree in some places to the (still subsidized) pricecharged by PSI in others.

There is one wrinkle with Sachs’s theory that poorpeople are stuck in a health-based poverty trap andthat money can get them out of it. Some of thesetechnologies are so cheap that everyone, even thevery poor, should be able to afford them

Nevertheless, the conventional wisdom is thattoday, at $20 per household per month, providingpiped water and sanitation is too expensive for thebudget of most developing countries. 13 The experi-ence of Gram Vikas,

Skeptics have been quick to point out that it is notclear whether malaria-infested countries are poorbecause of malaria, as Sachs assumes, or perhapstheir inability to eradicate malaria is an indicator ofthe fact that they are poorly governed. If it is the lat-ter, then the mere eradication of malaria mayachieve very little, as long as governance remainsweak

When a child came to them with diarrhea, all theycould offer the mother was a packet of oral rehydra-tion solution (or ORS, a mixture of salt, sugar, po-tassium chloride, and an antacid to be mixed withwater and drunk by the child). But most mothersdidn’t believe that ORS could do any good. Theywanted what they thought was the right treat-ment—an antibiotic or an intravenous drip. Once amother was sent away from the health center withjust a packet of ORS, the nurses told us, she nevercame back. Every year, they saw scores of childrendie from diarrhea, but they felt utterly powerless.

Government health workers, who are in chargeof delivering basic health-care services in mostcountries, are often blamed for this failure, not en-tirely unfairly, as we will see. They, on the otherhand, insist that plucking these low-hanging fruitsis much harder than it seems

Our reading of the evidence suggests that mostadults, even the very poor, are outside of the nutri-tion poverty trap zone: They can easily eat as muchas they need to be physically productive

The basic human need for a pleasant life mightexplain why food spending has been declining in In-dia

Generally, it is clear that things that make life lessboring are a priority for the poor. This may be atelevision, or a little bit of something special toeat—or just a cup of sugary tea. Even Pak Solhinhad a television, although it was not working whenwe visited him. Festivals may be seen in this light aswell. Where televisions or radios are not available, itis easy to see why the poor often seek out the dis-traction of a special family celebration of some kind,a religious observance, or a daughter’s wedding. Inour eighteen-country data set, it is clear that thepoor spend more on festivals when they are lesslikely to have a radio or a television.

Not surprisingly, both the king of Swaziland andthe South African Council of Churches (SACC) havetried to regulate funeral expenditures. In 2002, theking simply banned lavish funerals 32 and an-nounced that if a family was found to haveslaughtered a cow for their funeral, they would haveto give one cow to the chief’s herd. The SACC, rathermore soberly, called for a regulation of the funeralindustry, which, they felt, was putting pressure onfamilies to spend more than they could afford

Iodine might make your children smarter, butthe difference is not huge (though a number ofsmall differences may add up to something big) andin most cases you will not find out either way formany years.

This does not explain why all pregnant women inIndia aren’t using only iodine-fortified salt, which isnow available for purchase in every village. A pos-sibility is that people may not realize the value offeeding themselves and their children better. Theimportance of micronutrients was not fully under-stood, even by scientists, until relatively recently.Although micronutrients are cheap and can some-times lead to a large increase in lifetime income, itis necessary to know exactly what to eat (or whatpills to take). Not everyone has the information,even in the United States.

The Work and Iron Status Evaluation (WISE)study in Indonesia provided randomly chosen menand women in rural Indonesia with regular ironsupplementation for several months, while the com-parison group received a placebo. 28 The studyfound that the iron supplements made the men ableto work harder, and the resulting increase in theirincome was many times the cost of a yearly supplyof iron-fortified fish sauce. A year’s supply of thefish sauce cost $7 USD PPP, and for a self-employedmale, the yearly gain in earnings was $46 USDPPP—an excellent investment

he British MedicalJournal coined the term “Barker Hypothesis” torefer to Dr. David Barker’s theory that conditions inutero have long-term impact on a child’s life

. But a recent pa-per by Anne Case and Chris Paxson made some pro-gress in nailing down what explains this relation-ship. They show that in the United Kingdom and theUnited States, the effect of height is entirely accoun-ted for by differences in IQ: When we comparepeople who have the same IQ, there is no relation-ship between height and earning. 22 They interprettheir findings as showing that what matters is goodnutrition in early childhood: On average, adults whohave been well nourished as children are both tallerand smarter.

As Amartya Sen hasshown, however, most recent famines have beencaused not by lack of food availability but by institu-tional failures that led to poor distribution of theavailable food, or even hoarding and storage in theface of starvation elsewhere

The logic of the S—shape sug-gests that when resources are tight, it makes “eco-nomic sense” to sacrifice some people, so that therest have enough food to be able to work and earnenough to survive

So it is not becausethey don’t eat enough that most people stay poor

And perhaps they are really less hungry, despiteeating fewer calories. It could be that because of im-provements in water and sanitation, they are leak-ing fewer calories in bouts of diarrhea and other ail-ments. Or maybe they are less hungry because ofthe decline of heavy physical work—with the avail-ability of drinking water in the village, women donot need to carry heavy loads for long distances; im-provements in transportation have reduced theneed to travel on foot

The percentageof people who consider that they do not haveenough food has dropped dramatically over time:from 17 percent in 1983 to 2 percent in 2004. So,perhaps people eat less because they are lesshungry

The change is not driven by declining incomes; byall accounts, real incomes are increasing. Yet,though Indians are richer, they eat so much less ateach level of income that they eat less on averagetoday than they used to.

Households that received subsidies for rice or wheatconsumed less of those two items and ate moreshrimp and meat, even though their staples nowcost less. Remarkably, overall, the caloric intake ofthose who received the subsidy did not increase(and may even have decreased), despite the fact thattheir purchasing power had increased. Neither didthe nutritional content improve in any other sense.The likely explanation is that because the stapleformed such a large part of the household budget,the subsidies had made them richer: If the con-sumption of the staple is associated with being poor(say, because it is cheap but not particularly tasty),feeling richer might actually have made them con-sume less of it. Once again, this suggests that atleast among these very poor urban households, get-ting more calories was not a priority: Getting better-tasting ones was

Equally remarkable, even the money that peoplespend on food is not spent to maximize the intake ofcalories or micronutrients. When very poor peopleget a chance to spend a little bit more on food, theydon’t put everything into getting more calories. In-stead, they buy better-tasting, more expensive cal-ories. For the poorest group in Maharashtra in1983, out of every additional rupee spent on foodwhen income rose, about half went into purchasingmore calories, but the rest went into more expensivecalories.

Even among the verypoor, food expenditures increase much less thanone for one with the budget.

Yet, this is not what we see. Most people livingwith less than 99 cents a day do not seem to act as ifthey are starving. If they were, surely they wouldput every available penny into buying more calories.But they do not. In our eighteen-country data set onthe lives of the poor, food represents from 36 to 79percent of consumption among the rural extremelypoor, and 53 to 74 percent among their urban coun-terparts

As people get richer, they can buy more food.Once the basic metabolic needs of the body aretaken care of, all that extra food goes into buildingstrength, allowing people to produce much morethan they need to eat merely to stay alive.

Pak Solhin explained to himself the fact that hewas unemployed. Although he was evidently willingto work, lack of food made him weak and listless,and depression was sapping his will to dosomething to solve his problem.

The inability of the poor to feed themselvesproperly is also one of the most frequently cited rootcauses of a poverty trap.

The delivery of food aid on a massive scale is a lo-gistical nightmare. In India, it is estimated thatmore than one-half of the wheat and over one-thirdof the rice get “lost” along the way, including a goodfraction that gets eaten by rats

A “poor” personwas essentially defined as someone without enoughto eat

UN’s first Millennium Develop-ment Goal (MDG), which is “to reduce poverty andhunger.”

Finally, understandingwhat is at the root of the stigma associated with women working outside of homemight help design policies to address it: what are husbands trying to signal to othersby acting in opposition to women working outside the home?

Finally, it is important to note that the temporary nature of the job might inter-act with its potential visibility. It is unclear whether the effects would be larger orsmaller if the job was longer, in which case both visibility and monetary payoffswould be greater

when we examinethe degree of misperceptions across different samples and measurement approaches, avery consistent picture emerges: the vast majority underestimate the degree of supportfor WWOH and the average wedge is around 25–28 percentage points

Our results paint a consistent picture. When social norms in the labor market aremisperceived, simple information provision can affect labor supply decisions. In thecase of Saudi Arabia, most people perceive society to be more conservative than itreally is. As a result, too few women work outside of home

Interestingly, we find significant effects for outcomes (applying and interviewingfor jobs outside the home) that are not easily observable (when compared to work-ing outside the home itself). It is possible that the difference in beliefs about othersbetween the two conditions would have been smaller if control participants startedobserving more women working outside the home. Unfortunately, we do not havethe data to test this hypothesis

One might wonder if there could have been informational spillovers betweentreatment and control participants. These participants often reported knowing eachother and could have discussed the information given to them after the main exper-iment. This is particularly related to our hypothesis; if perceived social sanctionsassociated with WWOH have been reduced, some participants would be more opento discussing the topic with their neighbors or friends. While we observe persistentdifferences in beliefs about others in the follow-up survey, we have no way of know-ing if the differences between control and treatment groups would have been evenlarger in the absence of potential spillovers

We find that higherlevels of prior updating lead to significantly higher sign up rates. For example, usingthe specification in column 1, a positive update in prior corresponding to one stan-dard deviation of the wedge in the treatment group causally leads to a 6.9 percentagepoint increase in the job matching service sign up rate. In online Appendix Table B6,we provide suggestive evidence that higher levels of updating also lead to largerchanges in the longer-term outcomes

4.78

isperceived Social Norms.—We start by describing the measured mispercep-tion about social norms relating to WWOH. The average (median) guess is that63 percent (67 percent) of other session participants agree with the pro-WWOHstatement. The average level of agreement across all sessions is 87 percent, a num-ber larger than the guesses of close to 70 percent of participants. Of course, theproper comparison is between a participant’s guess and the share agreeing in hissession.

We then asked respondents to make an incentivized choice between receivinga $5 Amazon gift card and the opportunity to sign up their wives for access to thecompany’s platform and services.

In the second part of the survey, one-half the participants were randomly assignedto a treatment condition. This randomization was conducted at the individual level,based on the last three digits of the respondents’ phone number (provided in thefirst part of the survey). In the treatment condition, participants were given feedbackabout the responses of the other session participants to the two statements aboutfemale labor participation. In particular, we provided treatment participants withcharts embedded in the survey interface showing the proportion of respondents whoreported agreeing and disagreeing with each statement in the first part of the survey.Participants assigned to the control group received no information

In addition to using an anonymous online survey, we attempted to additionallyreduce the scope for social desirability bias (SBS), by avoiding priming effects.The study was framed as a general labor market survey, with filler questions ask-ing opinions on the employment insurance system, privileging Saudi nationals overforeigners into job vacancies, and the minimum wage level. In addition, no Westernnon-Arab was present during the intervention.

The survey itself was administered using the online survey software Qualtrics andwas implemented in two parts. 14 In the first part of the survey, we collected demo-graphic information and elicited participants’ own opinions about a range of topicsas well as their incentivized perceptions of others’ beliefs. In the second part of thesurvey, we randomized participants to our information provision treatment and mea-sure outcomes. At the start of the survey period, a survey link to the first part of thesurvey was provided on a board at the front of the room. Participants were instructedto navigate to the link and take the survey on their personal smartphones.15Since WWOH may be a sensitive issue for respondents, maintaining anonymityof responses is an important focus of our experimental design. While the namesand phone numbers of all participants were collected on the sign-in sheets, thisidentifying information was not collected in the survey itself.

We targetedthis age bracket primarily for logistical reasons in the sense that this populationwould be likely to use smartphones and also the mobile application/service that wasto be offered in the experiment.

We first conducted an experiment inthe field to establish the effect of correcting participants’ perceptions of the beliefsof others on a contemporaneous decision to sign up their wife for a job matching ser-vice. We then administered a follow-up survey three to five months later to collectinformation on longer-term labor supply outcomes.

If married Saudi men believe that a large shareof other men are opposed to WWOH and if they care to a great extent about theirsocial image, they may end up not letting their wives work outside the home, despitethe fact that most would prefer to allow their wives to work if the behavior were notobservable. However, our model predicts that correcting perceptions about the opin-ions of others leads to drastic changes in the share of men willing to let their wiveswork outside the home

On the policy side, our results highlight how simple information provision mightchange perceptions of a society’s opinions on important topics, and how this mighteventually lead to changes in behavior. Conducting opinion polls and diffusing infor-mation about their findings could potentially be used to change important behaviorsin some societies.

The evidence shows that a simple intervention in a natural setting like this canaffect labor supply decisions. Information significantly increase the likelihood thatjob-seekers choose the outside the home option, by 15 percentage points, which is an85 percent increase from the baseline likelihood of 18 percent in the control group

The company embedded an experiment in their efforts to recruit enumer-ators for a one-day surveyor job. Using a sample of around 300 females that hadpreviously indicated interest in working as a surveyor making calls from home, thecompany offered them the choice between that version of the job and a differentversion of the same job, interviewing respondents outside the home, face-to-face,in malls, for a 20 percent higher wage (with transportation costs additionally cov-ered by the company). Before making the decision of what version of the job totake, the survey firm informed a random set of these job-seekers of the findingsfrom our national survey: “In a recent survey of a national sample of about 1,500married Saudi men aged 18–35, 82 percent agreed with the statement, “In my opin-ion, women should be allowed to work outside the home.”

Our results carry some natural policy implications. They suggest that there aretoo few women working outside of the home because the labor market is in anequilibrium where social norms are misperceived—pluralistic ignorance—and thata low-cost intervention correcting perceptions will lead to more women choosingjobs outside the home

Moreover, frequency of discussion is a very strong predictor of the sizeof misperceptions. On average, there are very small misperceptions among thosefrequently discussing the topic, and very large misperceptions among those rarelytalking about it

Out of approximately 700 male respon-dents, 77 percent are in favor of WWOH and among male respondents aged 18–35(the age bracket in our study), the share is 79 percent. The Arab Barometer surveyalso allows us to establish that older men are also supportive of WWOH: amongthose over 35, the share agreeing with that statement is 74 percent. Moreover, thenumbers from the Arab Barometer in 2010–2011

We also conduct a similar-looking, anonymous online survey with a larger,national sample of about 1,500 married Saudi men, aged 18–35. The goal of thisadditional survey is twofold. First, we assess the external validity of the finding thatmost Saudi men privately support WWOH while failing to understand that others doas well. In this broader, more representative sample, 82 percent of men agree withthe same statement on WWOH used in the main experiment. When incentivized toguess the responses of other survey respondents, 92 percent of them underestimatethe true share. These are stronger misperceptions, perhaps because they are no lon-ger asked about their own neighbors’ opinions

To deal with these concerns,three to five months after the original intervention, participants were recontacted byphone and a series of additional outcome questions were collected. 6 We document alonger-term impact on self-reported labor supply outcomes. Wives of treated partic-ipants were significantly more likely to have applied for a job outside the home (upby 10 percentage points from a baseline level of 6 percent) and to have interviewedfor a job outside the home (up by 5 percentage points from a baseline level of 1 per-cent).

In a control group without belief cor-rections, 23 percent of participants chose the job matching service. In the treatmentgroup with feedback on the opinions of other participants, the share went up sig-nificantly, by 9 percentage points (a 36 percent increase).

Next, we evaluate whether correcting these misperceptions matters for a revealedpreference decision, associated with household labor supply. One-half of the par-ticipants were randomly given feedback on the true number of agreements withthe statement in their session

In an anonymous onlinesurvey, 87 percent of the experimental participants agreed with the statement, “Inmy opinion, women should be allowed to work outside the home.” When incentiv-ized to guess how other session participants responded to the same question, aboutthree-quarters of the experimental subjects underestimate the true number. We inter-pret this as evidence of misperception of social norms, even among people from thesame neighborhood who know each other

Finally, using a naturalrecruitment experiment with a local company, we find that misperception correctionleads to a higher share of women taking a higher-paying, outside-the-home tempo-rary survey enumerator job over a similar job to be performed from home.

Three to five months after the main intervention, the wives of men in our originalsample (whose beliefs about acceptability of WWOH were corrected) are more likelyto have applied and interviewed for a job outside the home.

We, therefore, ask: do Saudi men have correct percep-tions of the opinions held by other men regarding women working outside the home(WWOH)? If the social norm is misperceived, then does correcting beliefs leadmore women to work outside the home

“pluralisticignorance” (Katz, Allport, and Jenness 1931). This refers to a situation where mostpeople privately hold an opinion, but they incorrectly believe that most other people

The share of women working outside the home, an action that others caneasily observe, is substantially lower. In a national survey we conducted with youngmarried Saudi males in early 2018, only 4 percent had their wife currently work-ing outside the home.

andomly informing women about actualsupport for WWOH leads them to switch from an at-home tempo-rary enumerator job to a higher-paying, outside-the-home version

Months later, wives of menwhose beliefs were corrected are more likely to have applied andinterviewed for a job outside the home

Correcting these beliefs increases men’s (costly) willing-ness to help their wives search for jobs

We show that the vast majority of young married men in SaudiArabia privately support women working outside the home(WWOH) and substantially underestimate support by other simi-lar men.

To test whether the SEED account balances represent newsavings, or whether they represent shifting of assets betweenaccounts held at the institution, we define a new outcome vari-able: change in balance in all non-SEED savings accounts.

The coefficient on the interaction term is insignificant for allvariables except “active.” This suggests that, within the treat-ment group, the average effect of the treatment assignment isworking fairly uniformly across these other characteristics.

The marketing doesnot appear to have any independent effect

We find thatthose who are time inconsistent (impatient now, but patient forfuture trade-offs) are in fact more likely to take up the SEEDproduct.

On the deposit side, two optional design features were of-fered. First, a locked box (called a “ganansiya” box) was offered toeach client in exchange for a small fee.

Out of the 202 clients who opened accounts, 167opted for this box. This feature can be thought of as a mentalaccount with a small physical barrier, since the box is a smallphysical mechanism that provides individuals with a way to savefor a particular purpose

We conjecture that the amount-based goal is a strongerdevice, since there is an incentive to continue depositing after theinitial deposit

Alternatively, clientscould set a goal amount and only have access to the funds oncethat goal was reached

First, individuals restrictedtheir rights to withdraw funds until they reached a goal

(1) we introducethe product as part of a randomized control experiment in orderto account for unobserved determinants of participation in thesavings program, and (2) we conduct a baseline household surveyin order to understand more about the characteristics of thosewho take up such products; specifically, we link hyperbolic pref-erences to a demand for commitment

This account was a pure commitment savings productthat restricted access to deposits as per the client’s instructionsupon opening the account, but did not compensate the client forthis restriction

We then randomly chose half the clients and offeredthem a new account called a “SEED” (Save, Earn, Enjoy Deposits)account

a savings account with a commitment featurethat restricts their access to their funds but has no further bene-fits

individuals who voluntarily engage in commitment de-vices ex ante may improve their welfare. If individuals withtime-inconsistent preferences are sophisticated enough to realizeit, we should observe them engaging in various forms of commit-ment (much like Odysseus tying himself to the mast to avoid thetempting song of the sirens).

Women whoexhibited a lower discount rate for future relative to current trade-offs, and hencepotentially have a preference for commitment, were indeed significantly morelikely to open the commitment savings account

The savings product wasintended for individuals who want to commit now to restrict access to theirsavings, and who were sophisticated enough to engage in such a mechanism

Perhaps surprisingly, I do not find evidenceof decreased daytime drinking translating into increased labor supply, productivity,or earnings. However, individuals who were randomized to receive sobriety incen-tives took more advantage of a high-return savings opportunity

An alternative interpretation could be that alcohol is a key temp-tation good for this population such that reducing alcohol consumption mitigates theneed for commitment savings

The structure of the experiment allows for an additional test of the hypothesis thatincreasing sobriety mitigates self-control problems. If self-control problems preventindividuals from saving as much as they would like to, and if commitment savingsproducts help sophisticated individuals overcome these problems, then commitmentsavings should have a larger effect for individuals with more severe self-controlproblems. Hence, if alcohol intoxication reduces self-control, then increasing sobri-ety should lower the effect of commitment savings.

ese estimates do not imply that alcohol does not have profoundeffects on labor-market outcomes for at least three reasons

My main measure to assess the impact of incentives on daytime drinkingis the fraction of individuals who arrived sober at the study office among all enrolledparticipants

A remaining concern is that social desirability bias may have also contributedto individuals’ demand for commitment. While it is impossible to rule out sucheffects altogether, several reasons may mitigate concerns regarding potential socialdesirability bias. First, the stakes involved in individuals’ choices were consider-able, which is reassuring given recent evidence suggesting that demand effects areless likely to occur with high stakes

The main outcomes of interest in this study are (i) alcohol consumption andexpenditures, (ii) savings behavior, and (iii) labor-market participation and earn-ings. Each of these outcomes is described below

study

Third, to examine the interaction betweensobriety incentives and commitment savings, a cross-randomized subset of individ-uals was provided with a commitment savings account, i.e., a savings account tha

Second, to measure individuals’ demandfor sobriety incentives and thus to identify self-control problems regarding alcohol,a randomly-selected subset of individuals was given the choice between incentivesfor sobriety and unconditional payments.

First, to create exogenous variation in sobriety, a randomly-selectedsubsample of study participants was offered financial incentives to visit the studyoffice sober, while the remaining individuals were paid for coming to the study officeregardless of their alcohol consumption.

I do not find evidence of significant changes in labor supply,productivity, or earnings, though I cannot reject treatment effects of about 10 to15 percent for these outcomes. In contrast, offering sobriety incentives increasedindividuals’ daily savings at the study office by over 50 percent compared to a con-trol group that received similar average study payments independent of their alcoholconsumption.

Sobriety incentives decreased daytime drinking as mea-sured by a 33 percent (or 13 percentage point) increase in the fraction of individualswho visited the study office sober and equivalent reductions in breathalyzer scoresand self-reported drinking. However, overall alcohol consumption and expendituresremained nearly unchanged. This finding implies that individuals largely shiftedtheir drinking to later times of the day rather than reducing their overall drink-ing as a response to the incentives

This finding provides clear evidence for a desire for sobriety by making futuredrinking more costly, in contrast to the predictions of the Becker and Murphy (1988)rational addiction model, but in line with Gruber and K ̋oszegi (2001).

Individuals’ choices between sobriety incentives and unconditional paymentsreveal substantial willingness to pay and thus demand for commitment to increasetheir sobriety.

To measure theimpact of acute intoxication on intertemporal choices, all subjects were providedwith a high-return savings opportunity. For a cross-randomized subset of study par-ticipants, the savings account was a commitment savings account, i.e., individualscould not withdraw their savings until the end of their participation in the study

However, few real-world examples of successful commitment devices exist, andempirical evidence of positive willingness to pay for such devices is scarce, call-ing into question the underlying models and the efficacy of commitment devices(Laibson 2015, 2018)

andomlyreceiving sobriety incentives significantly reduced daytime drink-ing while leaving overall drinking unchanged.

In a 3-week field experiment, the majority of 229cycle-rickshaw drivers were willing to forgo substantial monetarypayments in order to set incentives for themselves to remain sober,thus exhibiting demand for commitment to sobriety.

Our contention is that for the mostpart, the problem is the opposite: It is too hard tostay motivated when everything you want looks im-possibly far away. Moving the goalposts closer maybe just what the poor need to start running towardthem.

DeanKarlan and Sendhil Mullainathan fully repaid theloans of a random subset of these vendors (in India,and in the Philippines). 13 For a while, many of thevendors managed to stay debt-free: After ten weeks,40 percent were still debt-free in the Philippines.

Those who feel that they have nothing to lose

Self-control may also be more difficult for the poorfor another reason: Decisions about how muchshould be saved are difficult decisions for anyone,rich and poor alike

You’llnever really save enough for that refrigerator, thevoice in your ear insists. Have a cup of tea instead .. .). The result is a vicious circle: Saving is less at-tractive for the poor, because for them the goaltends to be very far away, and they know that there

Al-cohol, in this sense, is a temptation good for manypeople, something that makes immediate claims onus without giving us anticipatory pleasure. In con-trast, a television is probably not a temptation good:Many poor people plan and save for months or evenyears to buy one.

In essence, we seem tohave a vision of how we should act in the future thatis often inconsistent with the way we act today andwill act in the future. One form that this “time in-consistency” takes is to spend now, at the same timeas we plan to save in the future.

This suggests that barriers to savingsare not all externally imposed. Part of the problemcomes from human psychology.

In other words, just asWycliffe Otieno had told us, once they had fertilizer,they didn’t resell it.

We asked him what he did when he had alreadypurchased fertilizer (but not yet used it) andsomeone got sick. Wasn’t he tempted to resell it at aloss? His answer was that he never found the needto resell the fertilizer. Instead, he tended to reevalu-ate the true urgency of any need when there was nomoney lying around. And if something really neededto be paid for, he would kill a chicken or work a bitharder as a bicycle taxi driver (a job he did on theside when he was not too busy with farming).

In surveys conducted over severalyears, Michael Kremer, Jonathan Robinson, and Es-ther found that only about 40 percent of the farmersin the Busia region in western Kenya (not far fromSauri, the village where Jeffrey Sachs and AngelinaJolie met Kennedy, the young farmer who had notbeen using fertilizer before the project gave it tohim) had ever used fertilizer, and just 25 percentused fertilizer in any given yea

Now, saythey go back to drinking their two cups more tea butcontinue to plough the 15.71 rupees they had savedfrom three days of not drinking so much tea backinto the business (that is, borrowing that muchless). That accumulated amount continues to grow(just as the 10 rupees had turned into 10.71 aftertwo days) and after exactly ninety days, they wouldbe completely debt-free. They would save 40 rupeesa day, which is the equivalent of about half a day’swages. All just for the price of six cups of tea!

First, there was the dis-turbing fact that most men did not use their (free)accounts. Many women did not use them either, orused them very little. Forty percent of women didnot make a single deposit, and less than half mademore than on

There is currently an important international effort,led in particular by the Bill & Melinda Gates Found-ation, to increase access to savings accounts for thepoor. Microsaving is poised to become the next mi-crofinance revolution.

Technology can also play arole. In Kenya, M-PESA allows users to depositmoney into an account linked to their cell phonesand then use the cell phone to send money to otherpeople’s accounts and to make payments.

To find out whether thiswas the case, Pascaline Dupas and Jonathan Robin-son paid the opening fees for a savings account at alocal village bank, on behalf of a random sample ofsmall business owners (bicycle taxi drivers, marketvendors, carpenters, and the like) in Bumala. Thebank had an office in the main marketplace whereall these people operated their businesses. The ac-counts didn’t pay any interest. Instead, they chargeda fee for each withdrawal. 3Few men ended up using the accounts that wereoffered to them, but about two-thirds of the womendeposited money at least once.

As a market vendor married to a farmer, JenniferAuma probably lived on much less than $2 a day.Yet she had an array of finely tuned financial instru-ments. We see this kind of financial ingenuity timeand time again

ROSCAs over traditional savings accounts: Theydon’t have fees, she could make small deposits, andon average she got access to the pot much fasterthan it would take her if she saved the same amountevery week.

Each ROSCA had a specific, separate purpose, sheexplained. The small ones were for her rent (thiswas before she built a house), the bigger ones fo

“moneyguards” (acquaintances who take care of small sumsof money for a little fee, or for free), and, as we saw,slowly building a house.

In Africa, the mostpopular instruments are rotating savings and creditassociations (ROSCAs)—more commonly known as“merry-go-rounds” in English-speaking Africa andas tontines in Francophone countries. ROSCAmembers meet at regular intervals, and all depositthe same amount of money into a common pot atevery meeting.

They form savings “clubs” with other savers, inwhich each member is supposed to make sure thatthe others achieve their savings goals. Self-helpgroups (SHGs), popular in parts of India and foundin many other countries as well, are savings clubsthat also give loans to their members out of the ac-cumulated savings of the group.

One of the great virtues of the recent movement,among microcredit enthusiasts and others, to recog-nize the nascent capitalist inside every poor manand woman is that it moves us away from this viewof the poor as either carefree or totally incompetent

The Victorians thought that was just how the poorwere—much too impatient and unable to think farenough ahead. Consequently, they believed that theonly way to keep the poor from sinking into a life ofsloth was to threaten them with extreme misery ifthey ever strayed from the straight and narrow. Sothey had the nightmarish poorhouse (where the in-digent were housed) and the debtors’ prisons thatCharles Dickens wrote about. That view of the pooras essentially different people, whose innate inclina-tion toward shortsighted behavior is what keepsthem poor, has persisted over the years in slightlydifferent forms

If the poor still save brick by brick, it must be be-cause they have no better way to save. Is it becausebanks have not found a way to collect the savings ofthe poor, and there is a “microsaving revolution”waiting to happen?

If you ask owners why they keep an unfinishedhouse, they generally have a simple answer: This ishow they save. The story is familiar. When Abhijit’sgrandfather earned some extra cash, he would add aroom to the house. This is how, one room at a time,more or less, the house where his family still liveswas built. Poorer people cannot afford a whole

As we will see, ideology, ignor-ance, and inertia—the three Is—on the part of theexpert, the aid worker, or the local policy maker, of-ten explain why policies fail and why aid does nothave the effect it should.

J—PAL’s work suggests thatthere are many who share our basic premise—that itis possible to make very significant progress againstthe biggest problem in the world through the accu-mulation of a set of small steps, each well thoughtout, carefully tested, and judiciously implemented

The Inverted L-Shape: No Poverty Trap

The shape of the curve is key: It is very flat at thebeginning, and then rises rapidly, before flatteningout again. We will call it, with some apologies to theEnglish alphabet, the S-shape curve

The implicit argument was that Kennedywas in a poverty trap in which he could not affordfertilizer: The gift of fertilizer freed him. It was theonly way he could escape from the trap

Sachs’s emphasis on one big pus

Dupas’s experiment, individuals were randomly se-lected to receive different levels of subsidy to pur-chase bed nets. By comparing the behavior of ran-domly selected equivalent groups that were offereda net at different prices, she was able to answer allthree of our questions, at least in the context inwhich the experiment was carried out

those thatreceived more aid did not grow faster than the rest

The first flyer raised an average of $1.16 fromeach student. The second flyer, in which the plightof millions became the plight of one, raised $2.83.The students, it seems, were willing to take some re-sponsibility for helping Rokia, but when faced withthe scale of the global problem, they feltdiscouraged.

s universal cash transfer scheme has proven relatively durable in comparison but remainsunder pressure. Transfers were made universally into individual bank accounts beginning in 2011,in which year they accounted for a full 6.5% of GDP and approximately 29% of median householdincome. The government has subsequently allowed the real value of the transfer to fall by morethan half by opting not to adjust it for inflation, however, amid criticism that the transfers reducethe labor supply of the poor (Salehi-Isfahani & Mostafavi-Dehzooei 2017)

o our knowledge, the first significant pilot of UBI in a developing country was conducted inthe Otjivero-Omitara area of Namibia between January 2008 and December 2009. All residentsyounger than 60 and registered as living in the area as of July 2007 received monthly, uncondi-tional transfers.

Third, it may be possible to build small ordeals into the design of programs to create sometargeting by self-selection and thus make programs more progressive.

Second, the kinds of targeting that will make most sense will often be relatively simple, such asgeographic targeting. Designers should guard against creating too much discretion for the front-line staff who implement targeting, including the implicit discretion that is created when a policyis too complicated for beneficiaries to understand it and hold local officials to account. However,these targets may perform extremely poorly in reaching the poor

n such settings, universality has the added potential benefits of reducingadministrative burdens on the bureaucracy (as officials only need to determine who exists, notwhether they also meet eligibility criteria) and, in doing so, also reducing the scope for the abuseof state power

Finally, given the discussion of costs, it is worth asking about the administration or imple-mentation costs of a universal system, i.e., the costs of distributing the transfers themselves. Toavoid double counting, it will be important for developing countries to have a well-functioninguniversal identification system, preferably digital, although this will be a one-time cost.

Hanna & Olken (2018) state that the costs of targeting are quite small in Indonesia and Peru:$42 million every three years, with annual costs of $1.1 million, in Indonesia, and $10.8 million,with annual costs of $1.1 million, in Peru. Kidd et al. (2017) give a good review of targeting anddiscuss some of the costs. For example, the 2009 PMT survey in Pakistan cost $60 million, thePMT survey in Indonesia cost $60 million in 2011 and $100 million in 2015, and Kenya’s HungerSafety Net Program spent approximately $10 million to survey only 380,000 households (4% ofthe population)

Most targeting measures in developing economies focus on household poverty and not indi-vidual poverty, since measuring individual poverty using consumption data is extremely hard andcostly to do reliably. However, focus on the household implies one important assumption: thatthe poorest individuals live in the poorest households

Any method of targeting yields both exclusion errors (poor households that are deemed ineli-gible) and inclusion errors (nonpoor households that are deemed eligible).

inally, targeting may create disincentive effects. As we discuss above, targeting a basic incomerelocates where the disincentive effects may fall (although, of course, they may be of small mag-nitude). Any redistributive scheme can create disincentive effects for some part of the population,depending on who pays and who receives. For example, in a UBI scheme, disincentive effectswould be concentrated among the people who are marginal contributors to the scheme, likely theupper and upper-middle classes.

Yet things are less clearin a world of imperfect markets. As we describe above, in worlds where markets are imperfect,recipients of transfers may use them to relax binding constraints on growth, for example, by pur-chasing assets that they otherwise could not have financed or starting businesses that they did nothave the capital for. In this case, the welfare impacts of the transfers depend on variation in theopportunities and constraints that each person faces, as well as variation in their baseline standardof living (and thus their marginal utilities). It could well be optimal to make transfers to someonea bit better off in a community if this will enable them to make a transformative investment inthat community.

Third, universality could improve the political economy of redistribution. Government capac-ity to implement nuanced targeting schemes is often limited, particularly in the poorest areas,where it is most important to get it right. In cases like these, making eligibility universal may havea modest effect on the realized incidence of benefits while at the same time substantially reducingthe scope for corruption and other abuses of powe

Second, universality could reduce administrative costs.

First, even in a state with strong capacity to target transfers to households with certain char-acteristics, it is not clear whether doing so increases their overall impact. Work quantifying therelationship between impacts and targeted characteristics is limited, as is work on the potentialdisincentive effects of targeting.

We suspect that universality has several underappreciated benefits, and targeting has severalunderappreciated limitations. We review three main issues

ndonesia reduced suicide rates by 18% (Christian et al. 2019). S

UBI could be transformative for people bound by internal constraints. Not having to worryabout making ends meet could free up the mental and emotional bandwidth needed to focus ongetting ahead or reset hopes and beliefs about the future. Whether this is true is, of course, stillto be seen.

his also suggests one potential reason forthe gender gap in the effects of a cash grant—perhaps poor women are more likely to lack theconfidence to succeed as a result of the various societal pressures on them. The seven-year follow-up of the graduation intervention in India finds evidence that the beneficiaries were starting tomove beyond the original businesses that they were helped to set up and into other lines of trade.This seems indicative of growing confidence and optimism

For some, poverty may simply tax the mental and emotional bandwidth needed to thinkthrough important decisions. Mullainathan & Shafir (2013) develop this perspective, drawing onstudies of people who are poor in terms both of money and of time or other resources. For exam-ple, farmers in Tamil Nadu, India exhibit diminished cognitive performance immediately beforeharvest, when they are poorest, relative to after the harvest, when they are flush with cash (Maniet al. 2013). Mechanisms like these can generate poverty traps in which the poor remain poor be-cause they cannot afford lumpy bandwidth-conserving investments that the rich make (Banerjee& Mullainathan 2008)

The second two find evidence that insurance leads to higher investment. The study of Bryan et al.(2014) on migrants in Bangladesh finds that an insured loan, where the repayments are canceledif the migration project fails because of excess rainfall, has positive effects on migration that aresimilar in magnitude to those of a cash grant of the same amount.

What we observe in the data, therefore, is the residual effect of the risk after manyof these adjustments, which may miss a large part of the true cost of risk. Second, the identifi-cation is mostly cross-sectional, which makes it difficult to control for a variety of unobservedfactors. One exception is the work of Cai (2016), who uses a triple-difference estimation to showthat government-subsidized crop insurance in Jiangxi province in China raises the take-up of thehigh-return but high-risk tobacco crop

This sug-gests that UBI could be more effective at alleviating credit constraints if recipients had the optionto turn it into a commitment device by, for example, asking to receive several payments lumpedtogether into one larger tranche

At the same time, a stream of small payments is probably not the best way to structure a cashtransfer if the goal is to finance investment.

When we take all the data into account, it seems likely that credit constraints bind for some andnot for others. Where they do, UBI potentially provides a source of capital to relax the constraint.Indeed, there is some evidence that existing social protection cash transfers have been used tofinance productive investments.

Blattman et al. (2014) report on an experiment in Uganda where young men were given grants ofapproximately $400 per head as a part of a group and find large and durable effects on earningstwo and four years after the intervention.

t appears that the entrepreneurs who chose to set up a business without access to relatively cheapcredit were very different from those who only did so when microcredit was available. The pres-ence or absence of such high-return entrepreneurs in the experimental sample may explain whysome studies find a positive impact, whereas others do not

The gap between lending rates of the banks and their deposit rates,which measures intermediation costs, also tends to be much higher in developing countries thanin the developed world.

Several strands of research suggest that lack of access to capital constrains some more than others

The current trend in economics is to try to connect all interventions to some narrative aboutgrowth, even when it is obvious that it is a stretch. This is unfortunate, both because it blinds us toother priorities and other narratives that may be more compelling and because, for the most part,we know very little about how to make growth happen.