Misbehaving

I want to tell you two stories about my friends Amos Tversky and Daniel Kahneman before we begin. The stories set the tone. Amos was diagnosed with metastatic melanoma in early 1996, a prognosis he shared with only a handful of close friends — those of us in the inner circle were not allowed to tell anyone except our spouses. He did not want to devote his last months to playing the part of a dying man. There was work to do.

About six weeks before Amos died I visited him under the thin disguise of finishing a paper we had been working on. We spent some time on that paper, then watched an NBA playoff game. That was Amos: wise in nearly every aspect of his life, including how to face dying. His son Oren read from a note Amos had written days before the end: “I feel that in the last few days we have been exchanging anecdotes and stories with the intention that they will be remembered.”

Early in my teaching career I gave a midterm exam designed to distinguish stars from the middle pack from those in genuine trouble. The exam worked. Scores spread out nicely. But the average score was only 72 points out of 100, and students were furious — even though the grading curve put everyone where they’d have been on any other exam. Their grades hadn’t changed. Only the number on the top of the page had.

I solved the problem by making the next exam out of 137 points. Average scores jumped to 96 — students were delighted. Nothing about their actual grades changed. Everything about how they felt changed. That is the book in miniature: factors economists consider irrelevant — reference points, labels, the number printed at the top of a page — turn out to matter enormously to actual humans. I call them Supposedly Irrelevant Factors, or SIFs.

The problem is not that people are stupid. The problem is with the model economists use. Economic theory posits that people are Econs — fully rational, self-controlled, with preferences unaffected by irrelevant frames. Real people are Humans. Homo sapiens, not homo economicus. We do not have infinite cognitive capacity, perfect willpower, or complete indifference to how choices are presented. The field I helped build takes those departures seriously.

I began to have deviant thoughts about economic theory while I was a graduate student at the University of Rochester. My thesis advisor Sherwin Rosen gave the New York Times this assessment of my career as a graduate student: “We did not expect much of him.” He was not wrong to be uncertain. I was hardly a star. But I noticed something that gnawed at me: the price at which people were willing to sell something they owned was reliably higher than what they’d pay to acquire the same thing.

Standard theory said this discrepancy shouldn’t exist. If you own a bottle of wine and would not sell it for less than $100, you shouldn’t be willing to pay more than $100 to acquire it. But people clearly would. A simple thought experiment: you are given a Cornell coffee mug. How much would you accept to give it up? Now imagine you don’t have the mug. How much would you pay to get one? The answers are systematically different. Giving something up registers as a loss; acquiring registers as a gain. Losses sting more.

I called this the endowment effect: we value things more once we own them. It violates the Coase theorem, a cornerstone of law and economics that says rational parties will trade their way to efficient outcomes regardless of who starts with what. If I own the mug but you’d value it more, we should trade. But the endowment effect means the owner’s minimum selling price consistently exceeds the non-owner’s maximum willingness to pay — and trades that should happen don’t.

The discrepancy between buying and selling prices got my mind wandering. Once I started paying attention, anomalies cropped up everywhere and I started writing them on the blackboard in my office. Jeffrey and I got free tickets to a blizzard-threatened basketball game in Buffalo. We went. Would we have driven to buy those same tickets? Absolutely not. The tickets were free — so losing them to the weather felt like losing nothing. But sunk costs aren’t supposed to matter.

Another item on the List: a colleague refused to sell a bottle of wine he’d bought for $5 and that now fetched $100 at auction, but also wouldn’t pay $35 for a similar bottle at a store. He was neither willing to sell nor willing to buy at the market price — a flat contradiction. My colleague was also a professor of economics. The List was not about the naivety of ordinary people. It was about something systematically wrong with the standard model.

An Econ would not care whether money spent yesterday can be recovered. A Human does. Vince paid $1,000 for an indoor tennis membership, then developed tennis elbow two months in. He kept playing in pain because he did not want to waste the membership fee. This is the sunk-cost fallacy — letting bygones govern current decisions. Economists know it’s irrational. Humans do it constantly, across jobs, marriages, and research projects, because the loss of an unrecovered cost feels real.

In 1976 a colleague mentioned in passing that I might enjoy reading some work by two Israeli psychologists. I was back in my office in Rochester and walked to a part of the library I had never visited. I picked up their summary paper published in Science: “Judgment Under Uncertainty: Heuristics and Biases.” The thesis was simple and elegant: humans use mental shortcuts, heuristics, that cause predictable errors. As I read, my hands were shaking.

The insight that changed everything: Kahneman and Tversky showed that the errors weren’t random. Ask people to estimate whether there are more gun deaths from homicide or suicide, and most say homicide — there are actually nearly twice as many by suicide. The point is not that individuals err; it’s that the errors are systematic and predictable. Errors that predictable cannot average out across a population. They have consequences at scale.

I spent the next year at Stanford contriving any excuse to be around Kahneman and Tversky before they arrived. Victor Fuchs at the NBER took pity on me and put me on his grant. Danny and I started taking long walks in the hills near campus. We were equally ignorant about each other’s fields, and that symmetry made the conversations something I had never experienced before. He a psychologist, I an economist — each trying to teach the other the norms of his discipline.

I would sometimes walk up the hill to find Danny and Amos in the final stages of writing prospect theory. They argued over every word. Their conversations mixed Hebrew and English with no acknowledgment of the switch. What struck me was their process: they would try a thought experiment on themselves first. If they agreed, they provisionally assumed others would respond similarly. Then they’d test it. The paper took months to polish — Amos’s goal, always, was to get it right.

Prospect theory proposed what I consider the most important pair of ideas in the book. First: people evaluate outcomes as gains and losses relative to a reference point, not as absolute levels of wealth. Second: losses hurt roughly twice as much as equivalent gains feel good. This asymmetry — loss aversion — is not a quirk. It is a stable feature of human psychology that would explain the endowment effect, the equity risk premium, and much else I had not yet imagined.

My friend Maya was shopping for a quilt. Regular prices: $300 for king, $250 for queen, $200 for double. This week, all sizes were $150. Maya couldn’t resist. She bought the king. She didn’t need a king. But the deal was too good to pass up. This is transaction utility — the pleasure of feeling you got a good deal, independent of how much you actually wanted the thing. Econs do not experience transaction utility. Humans live by it.

I proposed a two-part theory of consumer utility. Acquisition utility is standard economics: the value of the thing minus the price paid. Transaction utility depends on whether the price feels fair relative to a reference price. If you pay more than the reference price, negative transaction utility — a rip-off. Pay less, positive transaction utility — a bargain. That’s why paying $7 for a beer at a resort is annoying but expected, while paying $7 at a corner bodega is an outrage. Same beer.

Transaction utility shapes entire industries. Retailers who have trained customers to expect frequent sales cannot easily switch to everyday low pricing — customers feel robbed of the deal, even if the everyday price is competitive. J.C. Penney discovered this the hard way. Walmart and Costco succeeded precisely because they’ve convinced shoppers the store itself is permanently the good deal. They haven’t eliminated transaction utility — they’ve converted their brand into a permanent bargain signal.

In the late 1970s I spent time interviewing lower-middle-class households about how they managed money. Many kept envelopes or mason jars, one for rent, one for food, one for utilities. Economic theory says money is fungible — a dollar left in the utilities envelope is identical to a dollar in the food envelope. But for these households, they were not interchangeable. Moving money between mental accounts felt like breaking a rule.

Mental accounts exist for good reasons: they are a rudimentary self-control technology. If you know the rent money lives in one envelope, you will not spend it on dinner. But they also produce absurdities. A company with money remaining in a year-end equipment budget will spend it on equipment it doesn’t need rather than let the budget “go to waste.” A family whose gas prices fall will spend the windfall on premium gasoline rather than on other things — because it “belongs” in the gas budget.

Wine collectors reveal the strangeness of mental accounting beautifully. When asked how they feel about drinking a bottle from the cellar that they bought for $20 and now retails for $75, most say it feels like drinking for free. When asked about acquiring expensive wine to lay down for ten years, they describe it as an investment, not a purchase. But if it’s an investment, drinking it is liquidating the investment — hardly free. People can’t keep their own mental frameworks straight.

Rational economic man is not interested in unrecoverable past expenditures. He looks forward, not backward. But Humans look backward constantly — and this produces predictable waste. Vince’s tennis elbow is the archetype. Students who received small discounts on theater season tickets attended fewer performances than those who paid full price: they had less “in the account” demanding to be recouped. The sunk-cost fallacy gets people to do things that pain them rather than accept the psychological loss.

At the poker table at Cornell, I noticed something the standard sunk-cost logic did not cover. Players who were ahead seemed not to treat their winnings as real money — they were playing with “the house’s money” and took risks they’d never take with their own stake. Players who were behind took desperate gambles to break even. Two asymmetries, same mental account: gains feel unreal until locked in; losses feel urgent to erase. Both push behavior away from the rational benchmark in opposite directions.

I came across a quote from social scientist Donald McIntosh that I found profoundly clarifying: “The idea of self-control is paradoxical unless it is assumed that the psyche contains more than one energy system, and that these energy systems have some degree of independence from each other.” This from an obscure book I cannot now explain how I found. But it was obviously true. Self-control is about conflict. You can’t have conflict with only one protagonist.

Hersh Shefrin and I proposed what we called the planner-doer model. At any moment, the individual contains two selves. There is the forward-looking planner who has good intentions and cares about the future, and the devil-may-care doer who lives for the present. The planner can try to influence the doer through rules, rewards, and penalties — but the doer will always do the consuming. Odysseus tying himself to the mast is the canonical example of a planner constraining a future doer in advance.

The practical implication: willpower alone is not a reliable tool. The planner cannot simply instruct the doer to behave. The best strategies are structural. Remove the temptation from the environment. Precommit before the aroused state arrives. Make the unwanted option harder to access. Businesses and individuals who understand this use rules and defaults as substitutes for continuous willpower — and they work better than relying on people to resist in the moment.

With Daniel Kahneman and Jack Knetsch, I started studying what people consider fair. The key finding: people will punish unfairness at a cost to themselves. In the Ultimatum Game, a Proposer divides a sum and the Responder either accepts or rejects — a rejected offer means both parties get nothing. Game theory predicts the Proposer offers the smallest positive amount and the Responder accepts. In practice, Proposers offer close to half, and Responders reject offers below 20%. Humans will burn money to punish greed.

Firms already know much of this intuitively, even if they don’t call it behavioral economics. When a blizzard hits and hardware stores raise snow-shovel prices, customers remember and retaliate by shifting their business — even months later. Raising prices because demand spiked feels exploitative. Raising prices because costs increased feels legitimate. The underlying economics are identical. The sense of fairness is not about efficiency; it is about whether the reference transaction has been violated.

Fairness also explains a major macroeconomic puzzle: why wages don’t fall in recessions. In a land of Econs, when demand for labor drops, wages should fall until the market clears and everyone is employed. In the real world, firms lay workers off rather than cut wages. Survey evidence confirms why: wage cuts feel like violations of an implicit deal. Layoffs feel like misfortune. The reference wage — what people were earning — becomes a floor that firms can’t breach without destroying morale and trust.

Every time I presented my research at a conference in those years, the same battery of objections appeared. The first was the “as-if” argument: even if people don’t literally optimize, they act as if they do, so the models work. The second: “high stakes” — surely at real money, the biases vanish. The third: “learning” — in markets with repetition and feedback, people learn the right behavior. Each objection demanded a different rebuttal, and I spent years developing them.

The learning critique is the one I found most interesting to dismantle. For learning to fix decision-making biases, you need two ingredients: frequent practice and immediate feedback. Riding a bicycle offers both. But the most consequential decisions people face — choosing a career, buying a house, picking a retirement portfolio — offer neither. You rarely repeat them, and feedback arrives years later in forms too diffuse to assign to a prior decision. Learning arguments don’t survive the Groundhog Day test.

The 1985 Chicago conference brought things to a head. Amos, Danny, and I were to face the Chicago economists on home turf. Arrow gave the opening keynote and quietly defected: rationality, he said, is neither necessary nor sufficient to do good economic theory. That was the most useful endorsement we could have received. It did not end the debate, but it made clear that even Chicago’s own intellectual heroes were not uniformly committed to the regnant model.

Finance economists were certain their markets were different. Arbitrageurs — the smart money — would quickly correct any mispricings caused by irrational investors. If noise traders drove prices away from value, sharp traders would buy low and sell high, restoring efficiency and pocketing the difference. The argument was theoretically coherent and empirically comfortable. Only one thing was missing: evidence that it actually happened consistently and at scale.

Werner De Bondt was my first graduate student committed to combining psychology and finance. We formed a simple hypothesis from representativeness: investors extrapolate recent past returns too far into the future, making recent winners overvalued and recent losers undervalued. If true, buying a portfolio of three-to-five-year Losers and shorting Winners would earn abnormal returns as the market corrected its own overreaction. We ran the analysis. Losers outperformed Winners dramatically.

The response from finance economists was not quiet concession. Gene Fama and colleagues proposed that Loser stocks must be riskier — higher returns merely compensate for risk. But when we checked, Losers had lower betas than Winners: they were less risky by the standard measure. The joint-hypothesis problem meant the EMH could always retreat behind the claim that risk is being measured incorrectly. It was a useful fortress but it was not an empirical argument.

The equity premium puzzle asks why stocks have historically earned about 6% per year more than short-term government bonds, when standard theory predicts a much smaller gap. Shlomo Benartzi and I proposed an answer that combined loss aversion with mental accounting. We called it myopic loss aversion: investors evaluate their portfolios too frequently, seeing too many short-run losses, and demand extra return to hold the painful assets.

The logic: if you look at your stock portfolio every day, you will see losses nearly as often as gains — even if stocks are steadily appreciating over years. Each loss registers twice as painfully as an equal gain registers pleasantly. The result is that investors with daily or monthly evaluation periods must be compensated with very high expected returns to hold stocks willingly. Look at portfolios less often — annually — and the equity premium nearly disappears as a puzzle.

Closed-end funds offered a related anomaly: these funds own a portfolio of publicly traded stocks, so their “fair” price is simply the underlying net asset value. Yet closed-end funds routinely trade at discounts of 10–20% to NAV. The EMH says this can’t happen. It happens constantly, persistently, and in the predictable direction that emotion drives: discounts widen when investor sentiment is most pessimistic and narrow when it is most optimistic. Sentiment matters to prices.

By the time I returned from a year in Vancouver in 1986, I had been working on behavioral economics for eight years. The problem was that aside from Amos and Danny, I was mostly talking to myself. The Russell Sage Foundation changed that. Eric Wanner, the foundation’s president, formed the Behavioral Economics Roundtable in 1992 — Akerlof, Blinder, Camerer, Kahneman, Loewenstein, Schelling, Shiller, Tversky, and me — and backed a summer camp for graduate students that became the field’s nursery.

The anomalies column I wrote for the Journal of Economic Perspectives beginning in 1987 served a different purpose: it showed working economists that the list of empirical violations of standard theory was long, varied, and reproducible. Not one anomaly. Not a curiosity. A systematic catalogue that kept growing. The column was widely read — by the AEA’s own surveys, more economists read it than any other feature in the journal. Data accumulates, and accumulated data changes paradigms.

The experiment that finally convinced the experimental economists who remained skeptical of the endowment effect was simple. We brought undergraduates into a lab, gave half of them a Cornell coffee mug, and ran a market. Economic theory predicted eleven trades. We ran four rounds. The trades were four, one, two, and two. Sellers’ median reservation price was $5.25. Buyers’ median willingness to pay was $2.25. The gap was not noise. It was the endowment effect, reproducible, quantifiable, and robust.

One of the privileges of being a professor at a research university is the freedom to study what you find interesting and call it work. I found the NFL draft interesting. Teams use a draft chart — an informal but nearly universal guide to how many picks you should trade for which — and the chart drastically overvalues early picks. The New Orleans Saints once traded their entire draft for the rights to pick Ricky Williams. That trade eventually looked insane.

Five psychological mechanisms explain why early picks are systematically overpriced: overconfidence in scouting ability, overly extreme predictions, the winner’s curse, false consensus, and plain loss aversion from making a public high-profile choice. Massey and I found that the surplus value of early picks — their actual performance relative to what they cost in salary — was negative. Teams were paying more for picks that produced less return than later picks. The market was efficient in price terms; it was irrational in value terms.

The advice was simple: trade down. Accept a bundle of later picks in exchange for a top pick. Use the extra picks to reduce single-player risk and increase the number of chances at finding a productive player. Almost no team consistently followed this advice. Why? Agency problems: the general manager who trades down and the team finishes 7-9 gets fired. The agent’s incentives don’t align with the principal’s. Knowing the rational strategy and executing it are two different problems.

Critics always argued that high stakes would discipline behavior. Show people real money — not hypothetical gambles — and the biases vanish. Deal or No Deal offered the highest-stakes natural experiment available outside of financial markets. Contestants faced choices between a certain cash offer and a continued gamble over prizes that could reach hundreds of thousands of euros. The choices showed loss aversion, reference point effects, and house money behavior — at stakes that dwarfed any lab experiment.

One contestant refused a certain offer of $125,000 at a moment when the expected value of continuing was also exactly $125,000. At these stakes — a normal year’s household income — one might expect rational behavior. She said no deal. Another, Frank, turned down offers sequentially escalating toward $100,000 on a show where the top prize was $100,000. He left with $10. These are not anomalies of the poor or unsophisticated. They are features of how humans relate to reference points and to regret.

For years I thought the most important application of behavioral economics to savings was obvious: automatic enrollment in 401(k) plans. Default matters enormously. When plans switch from opt-in to opt-out, participation rates jump from 30–40% to over 90%. Economics predicts defaults should be irrelevant — rational actors evaluate their choices and enroll when it’s in their interest. But the default is treated as a recommendation, and inertia does the rest.

Automatic enrollment fixed the entry problem but not the savings rate problem. Most auto-enrolled participants saved at the default rate of 3%, invested in a money market fund — settings the employer chose to minimize legal liability, not to maximize retirement security. With Shlomo Benartzi I designed Save More Tomorrow: people commit now to raise their saving rate when their next raise arrives. Contribution increases happen automatically at every subsequent raise until they opt out or hit the cap.

SMarT, as we called it, used behavioral insights against themselves. Inertia, which keeps people undersaving, now keeps them saving more. Loss aversion, which makes people reluctant to cut take-home pay, is bypassed by tying increases to raises — the contribution goes up before the employee ever sees the increase as theirs. Present bias, which makes future commitments feel cheap, is exploited by asking people to decide now about behavior that starts later. The first company to try it quadrupled its average saving rate.

The pivot to policy began with a simple question Cass Sunstein and I kept asking each other: if we accept that people make predictable mistakes, what is the appropriate role for government? The libertarian answer is: nothing. The traditional paternalist answer: mandate or ban. We wanted a third option — policies that improve outcomes for Humans without requiring anyone to do anything. We called this libertarian paternalism. A colleague suggested the term was an oxymoron. I wanted to post a three-word reply: “No It’s Not.”

The best illustration is a fly etched into the urinal at Amsterdam’s Schiphol Airport. Aim improves dramatically. No rule required. No fine imposed. Just a nudge — a feature of the environment that attracts attention and guides behavior without restricting choice. A nudge is any factor that influences choices without changing the feasibility or the economic incentives. Choice architecture is the design of the context in which choices are made. All choices happen in some architecture; the question is whether it is designed intentionally or not.

Organ donation offers the starkest example of default power. In opt-in countries, donation rates hover around 15%. In opt-out countries — presumed consent — they approach 100%. Same operation, same legal right to decline. The only change is what requires a form. We ultimately recommended a third path: mandated choice, where renewing a driver’s license requires selecting a preference on the spot. Prompted choice removes the default entirely and forces a decision. Many states adopted it.

In 2008, before Nudge was published in the UK, a Conservative Party advisor named Rohan somehow found a pile of advance copies and stacked them on his desk. David Cameron — the future prime minister — walked past, picked one up, put it on his reading list. Two years later, when Cameron formed his coalition government, one of the first institutional innovations was a Behavioral Insights Team: seven people in a borrowed corner of Admiralty Arch, tasked with applying behavioral science to UK government policy.

Two mantras became the team’s operating principles. First: if you want to encourage something, make it easy. Remove friction before trying to change preferences. Second: run randomized controlled trials. Don’t assume an intervention works. Test it against a control. Many of the BIT’s biggest successes came not from clever psychology but from eliminating unnecessary steps — simplifying a form, sending a timely reminder, making the desired action the path of least resistance.

The BIT’s tax compliance experiment became a canonical case. The UK’s equivalent of the IRS sent letters to late payers. One version was the standard legal notice. Another added a single line: most people in your area have already paid their taxes. Tax collection from the nudge group improved significantly. The intervention cost nearly nothing. Similar logic — social norms as a nudge — now operates in energy conservation programs across dozens of countries.

It has now been more than forty years since I started writing the List on my blackboard in Rochester. Behavioral economics is no longer a fringe operation. When this book was published in 2015, I was serving as president of the American Economic Association; Robert Shiller, a fellow traveler, was my successor. As I wrote then: the lunatics are running the asylum. I am slowly adapting to the idea of going mainstream. Sigh.

The core lesson I take from this story is the importance of observation. Behavioral economics began with simple noticing: people eat too many nuts when the bowl is left out. People treat different dollars differently. People make mistakes — lots of them — in ways that are predictable and consistent. The theory came later. Before any theory, there was a willingness to take the phenomena seriously rather than dismiss them as irrelevant to the model.

The equally core lesson is data. Stories are how we remember things — and I have told many in this book — but a single anecdote is only an illustration. Conviction requires systematic evidence. My Chicago colleague Linda Ginzel always tells her students: “If you don’t write it down, it doesn’t exist.” The only protection against overconfidence is to document your track record of predictions, including the wrong ones. The field was built on data that the incumbent paradigm could not explain. And the data won.