DEEP READING · READ 937
Poor Economics · Abhijit Banerjee & Esther Duflo · 2011
Every apparently irrational choice a poor household makes — not buying a two-dollar bed net, skipping a free vaccination, paying for a wedding while short of grain, borrowing at annual rates in the tens of percent — hides a ledger an outsider cannot see; and the real difference between them and you is not judgement but the fact that your good decisions were made for you long ago by institutions, while every one of theirs has to be got right by hand. The method is to stop arguing about whether aid works and start running small experiments that can actually be proved wrong.
Banerjee and Duflo are development economists at MIT, married to each other, and founders of the Abdul Latif Jameel Poverty Action Lab (J-PAL) — the institution that carried the randomized trial out of medicine and into anti-poverty policy. Eight years after this book they shared the 2019 Nobel with Michael Kremer, precisely "for their experimental approach to alleviating global poverty." Poor Economics (2011) is the first systematic showcase of that approach and also a statement of posture: swap the grand narrative for questions small enough to be tested.
Settle the method first or everything after it floats. A randomized controlled trial is how new drugs are tested: split people at random into a treatment group and a control group, and because a coin decided who went where, the two groups differ on everything else — wealth, health, conscientiousness — only by random noise. Whatever difference appears afterwards can therefore be attributed to the drug.
Transplant that to poverty and the "drug" becomes a programme: free bed nets, fertilizer delivered to the door, a bag of lentils for a family that turns up for a shot. What it solves is the most stubborn disease in development economics: self-selection. Discovering that "households using bed nets have healthier children" tells you almost nothing — the households that bothered to collect a net were already more careful, better off, and closer to the clinic. The coin toss cuts that confound out at the root.
The real ambition here is a change of posture. Twentieth-century development argument was dominated by two figures: Jeffrey Sachs, who held that poor countries are stuck in a poverty trap and need a large one-off push to escape it, and William Easterly, who held that aid breeds dependence and feeds corruption, so the only way out runs through markets and local incentives. They argued for twenty years without moving each other, because the word "aid" contains several thousand entirely different things. Banerjee and Duflo's answer is that this fight cannot be won and need not be fought — cut the question down until a single experiment can answer it, then go and run the experiment.
The authors worked through household surveys across eighteen countries, and the first finding is already subversive: when an extremely poor household gets a little more money, it does not convert all of it into calories. A meaningful share goes to tea, tobacco, sugar, a television, a funeral, a wedding. Even within food, people often buy better-tasting rather than more filling — in parts of India, calorie intake per head has barely risen with income.
The condescending reading is short-sightedness. The authors' reading is cooler and much harder to argue with. First, nobody lives merely in order to remain alive. A family in a South Asian village told researchers that what they most needed was a television — because life is boring, and boredom is a real form of suffering. A respectable wedding buys dignity and social standing, and in a place with no formal protection of any kind, social standing is your insurance. Second, calories stopped being the binding constraint some time ago. The cheapest food in the world is now cheap enough; the real shortage tends to sit in micronutrients — iron, iodine, vitamin A, needed in tiny quantities and impossible to substitute. Research in Indonesia found that iron supplementation measurably raised the earnings of self-employed men, because anaemia takes physical strength straight off the top; iodine deficiency damages brain development before birth.
The conclusion is not that the poor buy the wrong things. It is that breaking poverty by getting people to eat one more bowl of rice pays far less than you would think, while micronutrient interventions are absurdly cheap, enormously profitable, and almost nobody does them.
The Sachs–Easterly argument can be drawn as two curves, and that drawing is the book's analytical spine. The horizontal axis is what you have today, the vertical what you have tomorrow, and the dashed diagonal is standing still.
Schematic, not real data. Left: below the threshold the curve sits under the diagonal, so you slide down — meaning one push past the threshold changes the outcome permanently. Right: the curve stays above the diagonal — poor is just poor, help helps, but there is no magic door.
If the world looks like the left panel, Sachs is right and a big push does permanent work. If it looks like the right panel, Easterly has a point: money relieves but does not redirect. The book's most important methodological stance is that this is not a matter of position but of measurement — and the answer differs by domain. Among the genuinely undernourished the trap is real (too little food, too weak to work, still less food). On calorie intake across most of the developing world today, it is not. Ask first whether this particular domain has that threshold, and only then decide whether a one-off push is the right instrument.
The most glaring fact in development is that many life-saving things are extremely cheap — bed nets, oral rehydration salts (a few cents a sachet, and the standard treatment for dehydration from diarrhoea), chlorine tablets, free vaccines — and uptake stays dismally low. The two standard explanations are ignorance (so: campaign) and poverty (so: subsidize). The authors show the real culprit is often a third thing: small hassles, compounded by ordinary procrastination.
The prettiest evidence comes from the immunization experiment in Udaipur, Rajasthan, where full childhood immunization ran at single-digit percentages. Arm one: make the mobile immunization camps reliable — fixed days, someone always there. Coverage roughly doubled, and remained low. Arm two: the same reliable camps, plus a kilo of lentils for each shot and a set of plates for completing the course. Coverage jumped to something over a third.
The counterintuitive part is the cost: the arm with the "bribe" had a lower total cost per completed immunization. The nurse and the cold chain are fixed costs sitting there regardless; the more people come, the less each one carries — and the lentils cost less than the idle waiting they eliminated.
Why should a kilo of lentils move a decision about a child's life? Because the two sides of the decision are asymmetric. The benefit is distant, abstract and probabilistic (your child might not get measles some day); the cost is immediate, concrete and certain (a two-kilometre walk, a queue, a crying child, a day's wages). Humans are subject to time inconsistency — we always intend to do the right thing next week, so we never do it. The lentils are not a purchase of consent; they put a small immediate benefit on the other side of the scale so that today becomes the day. The same mechanism explains Duflo, Kremer and Robinson's fertilizer work in Kenya: farmers know fertilizer pays, but by planting time the money is gone. Offer a small discount and free delivery in the few days after harvest, when the cash is in hand, and usage rises sharply. A small nudge at the right moment beats a large subsidy at the wrong one.
This is the case worth remembering on its own, because it demonstrates evidence closing a long-running argument. Should bed nets be given away or sold for a token price? The case for charging sounds entirely reasonable: (1) people do not value what they get for nothing; (2) a price screens for who really needs it; (3) free distribution destroys the local retail market and breeds dependence.
Cohen and Dupas tested all three in Kenya. Raise the price from zero by even a little and take-up collapses. People who received nets free were no less likely to actually hang and use them than people who paid — so much for not valuing the free thing. And most decisively, households that had once received a net free were more likely to pay for their next one, having used one and learned what it does. That last result guts the central worry: free is not the beginning of dependence, it is the entry point to learning.
The rise in enrolment across the developing world over recent decades is a real achievement, but the authors point at a broken link in the middle: the children arrived in the classroom and the learning did not follow. India's ASER survey — a household-level reading and arithmetic test run by the education organization Pratham — finds year after year that a large share of children in grade five cannot read a grade-two text.
The cause is a set of mutually reinforcing beliefs. One: parents treat education as a lottery ticket. They believe the returns sit at the very end — only going all the way through and landing a government job counts as winning, and stopping halfway is worth nothing. So the moment a child looks like a poor bet, the investment is withdrawn and concentrated on the brightest sibling. Two: teachers and curricula are on the same side. Textbooks are written for the strongest students and teaching is pitched at the front row, so the children who are behind are quietly written off in week one. Three: the belief comes true. The abandoned child does indeed fail to learn, confirming that they were never the studious type.
The remedy is almost insultingly plain: teaching at the right level. Pratham's remedial tutors pull the furthest-behind children out for short periods and teach them not by grade but by the level they are actually at, and per dollar the effect dwarfs nearly every other education intervention. The insight is not that tutoring works; it is that the system's failure point is not its inputs but the fact that it delivers the lesson to the wrong people.
Poor households do not spend little on health; the problem is where it goes. The pattern the authors describe recurs across countries: public clinics are free but the nurse is frequently absent — surprise audits in several countries find absence rates around forty per cent — while the "doctor" in the private clinic often has no licence and is enthusiastic about injections and antibiotics, because patients feel that only a needle counts as treatment.
Meanwhile clean water, nets, handwashing and vaccines — nearly free, with enormous returns — cannot be given away. A brutal cognitive mechanism is at work: most illnesses resolve on their own. You take the injection, you take the course of antibiotics, and you get better — so "expensive treatment works" is confirmed by your own experience over and over. Prevention's success, by contrast, looks like nothing happening at all, and no experience ever registers it. That is why the poor have faith in costly cures and none in cheap prevention: one of them pays you feedback daily, the other never pays you any.
When this book appeared, microcredit was at the summit of its myth — Yunus and Grameen had just taken a Nobel Peace Prize. The authors ran one of the first serious randomized evaluations, in Hyderabad, and the result flatters nobody: households with loans did invest more in small businesses and buy more durables, but over the short run there was no detectable improvement in health, children's education, women's standing, or average consumption.
Why? The answer hides inside the machinery that makes microcredit work at all: small loans, repaid in frequent fixed instalments beginning a week or two after disbursement, under group liability. That discipline is exactly what keeps default low enough to cover costs — and it also locks the money into uses that are stable, low-risk and immediately cash-generating, while the investments that actually make a business grow (a machine, a bulk purchase, a second stall) pay back late and carry risk. The thing that makes microcredit function is the same thing that caps what it can become.
What the poor often lack more urgently is somewhere safe to put money. With no account, cash kept at home is borrowed by relatives and spent on impulse; so people join rotating savings clubs, or build a house one brick at a time, or even pay interest to borrow simply to impose an obligation on themselves. Which explains something that looks absurd on its face: repayment becomes a savings discipline — what they are buying is not money but self-control.
"The poor are natural entrepreneurs" is the most romantic line in development talk. The data support its first half: the first few dollars put into a market stall earn a startlingly high marginal return. The authors immediately supply the second half: that return curve flattens almost at once. Ten times the capital in a vegetable stall buys one more stall on the same street — demand is finite, and skill, management and market radius cap the ceiling.
So they offer the book's least flattering and most honest sentence: the poor are entrepreneurs by necessity, not by nature. They run stalls because nothing else is available, and when you ask what they want for their children, the answer is overwhelmingly not "inherit the business" but a steady job with the government or a large firm. A steady job is not merely income; it is predictability — and only with predictability do you dare keep a child in school to the end, borrow to build a house, or make any plan that assumes you will still be here next year. What looks like entrepreneurship everywhere is the absence of jobs everywhere.
The other great development argument is about institutions: unless property rights, rule of law and democracy are fixed at the root, nothing else matters. The authors do not deny that big institutions matter, but they note a practical fact: the big reform usually cannot be had, while marginal changes frequently work — and can be measured.
Three examples. First, India's village councils randomly reserve the leadership for women in some villages; Chattopadhyay and Duflo found that under a female head the mix of public goods shifted noticeably toward things women cared more about, such as drinking water. Better still, villagers initially rated female leaders poorly — but after actually living under one, the bias fell substantially. Prejudice was cured by experience, not by exhortation. Second, in some Indian schools a date-stamping camera plus pay tied to attendance cut teacher absence by close to half — the cost of monitoring can be as low as one camera. Third, printing candidates' criminal records or the audit results of public projects and putting them in voters' hands changes how people vote: voters were not indifferent, they were uninformed.
Stack every case in the book on top of each other and one conclusion rises out of the pile — the book's real moral core. Rich people get so many decisions right not mostly because they are wiser, but because the decisions were already made for them: the water was chlorinated before it left the plant, the pension comes out of the payslip automatically, school attendance is compulsory, vaccination is written into the maternity protocol, the iodine is already in the salt. You can go a whole year without thinking about any of it and remain perfectly fine.
The poor have none of those defaults. Every day they must decide personally: boil the water or not, buy the chlorine or not, walk two kilometres for a shot or not, save against a drought that may not come, trust which unlicensed doctor. The less money you have, the more decisions must be got right, and the worse the cost of getting one wrong — the most underrated expense of being poor. Follow that conclusion and the shape of good policy is obvious: not teaching poor people to choose better, but supplying them the defaults — making the right thing happen without anyone having to resolve to do it.
The book runs in a straight line from method to conclusion:
1. The poor are no less clever than anyone else; fill in the constraints and almost every apparently absurd choice adds up — what needs fixing is the environment, not their reasoning.
2. "Does aid work" has no answer; "did this programme, in this place, move this number, and by how much" does — and a coin toss cuts causality out of the confounds.
3. Spare money does not all become calories: people also buy relief from boredom, dignity, and standing. The neglected shortage is in micronutrients like iron and iodine — absurdly cheap, enormously profitable.
4. A poverty trap is a measurement question, not a position: find out whether this domain has the threshold before deciding whether a one-off push is the right tool.
5. What blocks a good thing is usually neither ignorance nor poverty but a small hassle plus ordinary procrastination — a kilo of lentils multiplied Udaipur's child immunization rate, and lowered the cost per shot while doing it.
6. The free bed net experiment ended an ideological fight: charge even a token and take-up collapses, free recipients use them just as much, and they are more likely to buy the next one themselves — free is not the start of dependence but the entry to learning.
7. Enrolment is not learning. Parents treat school as a lottery where only finishing counts, teachers pitch at the front row, and the written-off child duly fails. The fix is teaching at the child's actual level rather than their grade.
8. Cheap prevention will not sell and costly cures always do, because most illness resolves on its own: treatment pays you feedback every time, while prevention's success looks exactly like nothing happening.
9. Microcredit is a useful financial product, not a miracle: the frequent-repayment discipline that makes it function is also what stops it growing anything — and what the poor often lack more is a safe place to save; sometimes what they borrow is not money but self-control.
10. The line to keep: the rich had their good decisions made for them by institutions, while the poor must personally serve as their own water utility, pension fund, doctor and insurer. The less money you have, the more you must get right — so good policy does not teach better choosing, it turns good outcomes into defaults.