Behavioral Economics: Why Humans Are Predictably Irrational and What to Do About It

In 1974, two psychologists spun a wheel that had been rigged to stop on either 10 or 65. Their subjects then estimated what percentage of United Nations member countries were in Africa. People who saw 10 gave a median guess of about 25 percent, and people who saw 65 gave a median of about 45 percent. The wheel was obviously random, and it still moved their answers. The researchers were Amos Tversky and Daniel Kahneman, and the effect, called anchoring, was one of dozens of systematic errors they documented. This article explains the main ones, what the evidence behind them looks like, and where the field's claims are contested.

Educational only, not financial advice. Investing involves risk of loss; past performance does not guarantee future results. Consult a licensed financial advisor before making investment decisions.

Their work founded behavioral economics, the study of how real people, with limited attention and predictable habits, make choices about money. The field has produced Nobel Prizes, reshaped retirement plan design, and spawned government "nudge units." It has also drawn serious criticism over replication and effect sizes, which this article covers alongside the successes.

Key takeaways

• Humans use shortcuts (heuristics) that work well most of the time but produce predictable errors. The errors have direction, which is what makes them useful to study.
• Losses weigh more than equal gains. In the 1992 version of prospect theory, the loss aversion coefficient was set at 2.25, meaning a loss counts about twice as much as a gain of the same size. Later research shows the ratio varies with context and person.
• Defaults and framing often decide outcomes. Automatic enrollment changed 401(k) participation far more than most information campaigns have.
• Nudges are real but their average size is disputed. A widely cited meta-analysis found a moderate effect; a re-analysis that corrected for publication bias found close to zero.
• Knowing about a bias helps only partly. Training studies find reductions of roughly 20 to 30 percent in measured bias, not elimination.

Where behavioral economics came from

For most of the twentieth century, mainstream economic models assumed that people weigh costs and benefits consistently, hold stable preferences, and respond only to relevant information. Herbert Simon had argued in the 1950s that real decision makers have "bounded rationality," meaning limited time and information, and settle for good-enough answers. His critique mostly stayed at the edges of economics.

Kahneman and Tversky, working as psychologists, gave the critique experimental teeth. Their 1974 paper in Science described three heuristics (representativeness, availability, and anchoring) and the biases they produce. Their 1979 paper in Econometrica proposed prospect theory as a descriptive alternative to expected utility theory. Kahneman shared the 2002 Nobel Memorial Prize in Economic Sciences for this work; Tversky had died in 1996, and the prize is not awarded posthumously. Richard Thaler later extended the program into practical economics and won the 2017 prize.

Year	Milestone	Why it mattered
1974	Tversky and Kahneman, heuristics and biases (Science)	Showed that judgment errors are systematic, not random
1979	Prospect theory (Econometrica)	Gave a formal, testable account of choices under risk
1990	Kahneman, Knetsch, and Thaler, endowment effect experiments	Brought the field's methods into mainstream economics journals
1998 to 2000	Odean, and Barber and Odean, on investor behavior	Tied biases to real brokerage records
2001	Madrian and Shea, automatic 401(k) enrollment	Showed a default could shift savings behavior at scale
2008	Thaler and Sunstein, Nudge	Popularized choice architecture as policy
2011	Kahneman, Thinking, Fast and Slow	Brought the program to a general audience

Two systems of thinking, and why the metaphor is only a metaphor

In Thinking, Fast and Slow (2011), Kahneman organized much of this research around two modes of thought. System 1 is fast, automatic, and intuitive: reading a face, completing "bread and ...", flinching at a sudden sound. System 2 is slow, effortful, and deliberate: multiplying 17 by 24, comparing two insurance policies line by line.

Picture System 1 as a seasoned driver on a familiar commute, handling the road on autopilot, and System 2 as the same driver after a surprise detour, suddenly concentrating on every turn. The autopilot is not stupid. It is efficient, and it is usually right. It fails in situations that look familiar but are not, such as a financial product whose terms are different from anything encountered in daily life.

The framing has limits worth knowing. System 1 and System 2 are labels for clusters of processes, not two physical regions of the brain, and some researchers argue the split is too tidy. Used as a mnemonic, it helps explain why a fast intuition often arrives first and a slower reasoning process then justifies it. Used as a literal model, it overstates what the evidence shows.

For a broader tour of the biases that follow from these shortcuts, see our explainer on cognitive biases and the longer treatment of how people actually decide.

Prospect theory: the core of loss aversion

Standard expected-utility models assume people judge outcomes by their final wealth. A $1,000 gain and a $1,000 loss are mirror images, equal in size and opposite in sign. Prospect theory replaced that picture with three observations.

Reference dependence. People judge outcomes as gains or losses relative to a reference point, often the status quo, rather than as final totals. Being up $500 from what you expected feels different from being down $500 from what you expected, even if both leave you with the same bank balance.

Loss aversion. Losses loom larger than gains. In the cumulative version of prospect theory (Tversky and Kahneman, 1992), the loss aversion parameter was 2.25, so a loss is weighted a little more than twice a gain. Reviews describe typical estimates in the range of about 2 to 2.5, with substantial variation.

Diminishing sensitivity. The difference between $0 and $100 feels bigger than the difference between $1,100 and $1,200. This gives the value function an S-shape, steeper near the reference point and flatter far from it, and steeper for losses than for gains.

The reflection effect: risk-averse for gains, risk-seeking for losses

These ingredients predict a distinctive pattern, which a stylized example makes concrete. Offered a sure $500 or a 50 percent chance of $1,000, many people take the sure thing, which is risk-averse behavior in the domain of gains. Offered a sure loss of $500 or a 50 percent chance of losing $1,000, many people take the gamble, which is risk-seeking behavior in the domain of losses. The two choices have the same expected values, mirrored, and the reflection of preferences around zero is what gives the effect its name.

The pattern is a tendency, not a law. Choices depend on probabilities, stakes, and framing. At very small probabilities people often do the opposite, buying lottery tickets while also buying insurance, which prospect theory handles with a separate probability-weighting function. For a deeper look at the loss side, see why losses feel stronger than gains.

Is loss aversion real? The debate

Loss aversion is among the most cited findings in the field, and also one of the most debated. In 2018, Gal and Rucker argued in the Journal of Consumer Psychology that evidence does not support the claim that losses in general are more impactful than gains, and that the concept has been overgeneralized. A reply by Mrkva, Johnson, Gächter, and Herrmann (2020) argued that loss aversion has moderators, such as the person and the context, but that it has not disappeared.

A fair reading is that loss aversion is real in many settings, that its size depends on context and on the person, and that the single number "2" should be read as a rough central estimate, not a constant of nature. The 2.25 figure belongs to one model fitted to one set of choices.

Anchoring: the pull of the first number

The 1974 wheel experiment is the standard illustration. Participants first said whether the true percentage was higher or lower than the wheel's number, and then gave their best estimate. Their estimates stayed close to the number they had just seen. The researchers explained this as insufficient adjustment: people start from a given value and move away from it, but stop too soon. Payoffs for accuracy did not remove the effect in later work, and the pattern has appeared across a wide range of estimation tasks.

Everyday examples are easy to construct. A price tag reading "was $200, now $89" gives $200 as a starting point, and $89 feels like a bargain against it whether or not the item was ever sold for $200. In a negotiation, whichever figure is spoken first becomes a reference for the figures that follow. We explore these mechanisms and the evidence in more detail in how the anchoring effect changes decisions.

Framing works in a related way. The same facts, described as a gain or a loss, can produce opposite choices, as in the classic "400 deaths versus 200 survivors" experiment covered in the Tversky-Kahneman framing experiment. For how that plays out with risk, see how framing changes risk perception.

The endowment effect and status quo bias

In a well-known set of experiments by Kahneman, Knetsch, and Thaler (1990), some participants were randomly given coffee mugs and others were not. Those given a mug asked for much more to sell it than those without one would pay to buy it. Summaries of the study report median selling prices around $5.25 against median buying prices of roughly $2.75, a ratio of about two. Since the mugs were handed out at random, the gap cannot be explained by mug owners being people who value mugs more. Ownership itself seemed to raise the price.

The endowment effect is related to status quo bias, a preference for the current state of affairs. If the status quo is the reference point, then any change involves giving up something (a loss) in exchange for something new (a gain), and loss aversion tilts the comparison toward staying put. The endowment effect has been studied with many goods, and its size varies a great deal, especially for items people expect to trade, such as money or tokens. It is a tendency that shows up under particular conditions, not a universal law.

Status quo bias is why defaults matter so much, which leads directly to the policy side of the field.

Mental accounting and the disposition effect

Thaler also described mental accounting: the habit of sorting money into separate mental categories, such as rent, vacation, or windfall, and treating each differently, even though a dollar is a dollar. A person who would never withdraw savings for a purchase may happily spend an unexpected bonus on the same item.

A close relative in investing is the disposition effect, the tendency to sell assets that have risen in price while holding on to assets that have fallen. Terrance Odean analyzed trading records from 10,000 accounts at a large discount brokerage and found a strong preference for realizing winners rather than losers. The pattern was not explained by portfolio rebalancing or by trading costs, and subsequent returns did not justify it. For taxable accounts it is suboptimal and lowers after-tax returns, he reported, with tax-motivated selling most visible in December (Odean, 1998). Prospect theory offers one explanation: selling a loser forces a loss to be recognized, and the reflection effect makes people reluctant to close the book on it.

The disposition effect is a good example of why behavioral findings matter in practice. Nothing about it requires a lack of intelligence. It requires only an emotional asymmetry that standard models treated as irrelevant.

Nudges: using the biases on purpose

In 2008, Richard Thaler and Cass Sunstein published Nudge, which argued that choices can be steered toward better outcomes by redesigning the environment in which they are made, without banning options or changing prices. They called the approach "libertarian paternalism." The central tool is the default, since people tend to keep whatever option is pre-selected.

Automatic enrollment in retirement plans

The best-known case is the 401(k) retirement plan in the United States. Madrian and Shea studied a large corporation that switched from opt-in enrollment to automatic enrollment, with no change to the plan's economics. Participation rose significantly, and many employees hired under the new rules stayed with both the default contribution rate and the default investment fund, behavior the authors attributed to inertia and to treating the default as implicit advice from the employer (NBER Working Paper 7682). The authors reported the largest participation gains among younger, lower-paid, and minority workers.

The study also revealed the downside. Many employees stuck with the default contribution rate of 3 percent, which is lower than they might have chosen themselves, so a default can anchor people at a level that is not ideal for them.

Thaler and Shlomo Benartzi addressed that with Save More Tomorrow, a program in which employees commit in advance to raise their savings rate with future pay raises. At one manufacturer, participants' savings rose from 3.5 percent of income to 13.6 percent by the fourth pay raise after joining. Because participants chose to join, the causal evidence in that study is weaker than in a randomized trial, so the number shows what is possible rather than what the program reliably produces (Thaler and Benartzi, 2004).

Other nudges and what the data show

Setting	Nudge	What studies report	Source
Retirement saving	Automatic enrollment	Large rise in participation among new hires; many keep default contribution and fund	Madrian and Shea, 2001
Organ donation	Opt-out consent default	Much higher effective consent in opt-out countries; critics note consent is not the same as actual donation	Johnson and Goldstein, 2003
Home energy	Reports comparing a household with neighbors	Average reduction of about 2 percent in a randomized trial with about 80,000 households	Allcott, 2011
Tax collection	Letters stating that most people pay on time	Payment rates up by 1.3 to 5.1 percent within 23 days, depending on the message	Hallsworth et al., 2017

The tax letters are a clear case of a small, cheap intervention. In two natural field experiments run with the UK tax authority, covering more than 200,000 people, a letter telling recipients that nine in ten people pay on time raised payment rates by 1.3 percent. Naming the recipient's own country raised it by 2.1 percent. A message saying the recipient was in a very small minority of non-payers raised it by 5.1 percent, the largest effect (NBER Digest, 2014). Descriptive norms, which say what others do, outperformed injunctive norms, which say what people should do. The energy-report results were of a similar scale, about a two percent reduction on average, which sounds small but applies across millions of households (MIT CEEPR). Social comparison also features in our explainer on social proof.

Organ donation shows how a striking headline can need care. Johnson and Goldstein reported in Science in 2003 that "effective consent" was far higher in countries with opt-out systems than opt-in ones. Critics point out that effective consent is not the same thing as actual donation, since families are often involved and many other factors differ across countries. Consent policy is only one of many factors that differ between countries.

How big are nudge effects really?

This is where the field's credibility is most contested. A 2022 meta-analysis by Mertens and colleagues in PNAS pooled a large body of studies and reported an unadjusted average effect of d = 0.43, a moderate size, concluding that choice architecture is an effective behavior-change tool. A reply by Maier and colleagues argued that the literature has severe publication bias, meaning studies with small or null results were less likely to be published. After adjusting for it, their pooled estimate was near zero (0.04, with an interval running from 0.00 to 0.14) (Maier et al., 2022). They also noted that effects vary across categories, so some nudges may still work even if the average is small.

The sensible conclusion is not that nudging fails. Defaults in retirement plans and some norm messages show clear effects in large field trials. The more careful conclusion is that average effects reported in the literature are probably inflated, that effects differ by setting, and that a nudge should be tested in its own context rather than assumed to work.

Behavioral finance: what the biases do to portfolios

Behavioral finance applies these ideas to investing. Three findings stand out, and each is tied to data rather than anecdote.

Overconfidence and overtrading. Barber and Odean examined 66,465 households at a large discount broker from 1991 to 1996. The average household earned 16.4 percent a year while the market returned 17.9 percent, and the households that traded most earned 11.4 percent. The authors attributed excessive trading partly to overconfidence (Barber and Odean, 2000). One critique holds that the gap mostly reflects transaction costs rather than poor stock selection, a reminder that even well-known results get debated.

Overconfidence is easy to find outside investing. In a 2023 replication of Svenson's classic driving study, published in Meta-Psychology, the majority of participants rated their own driving skill and safety as above average, which cannot be true of a majority. Such "better than average" beliefs are one route to the confidence that drives frequent trading.

Home bias. Investors in many countries hold far more of their home market than its share of global market capitalization would suggest. Familiarity feels like safety, but concentration in one market can raise risk.

Disposition effect. As above, investors tend to sell winners and hold losers, which can raise tax bills without raising returns.

Bias	Plain definition	Classic evidence	Typical money consequence
Anchoring	Estimates stick to an initial number	Tversky and Kahneman, 1974	Overpaying relative to a "list" price
Loss aversion	Losses weigh more than equal gains	Tversky and Kahneman, 1979 and 1992	Holding losers; avoiding reasonable risks
Endowment effect	Owning raises perceived value	Kahneman, Knetsch, and Thaler, 1990	Reluctance to sell or switch
Disposition effect	Sell winners, hold losers	Odean, 1998	Higher taxes, weaker returns
Overconfidence	Overrating one's skill or knowledge	Barber and Odean, 2000	Too much trading, higher costs
Home bias	Overweighting familiar markets	Home-bias literature	Less diversification
Status quo and defaults	Sticking with the pre-selected option	Madrian and Shea, 2001	Low or unchanged savings rates

For the mechanics behind diversification, which several of these biases undermine, see why diversification reduces risk, and for how markets behave when many investors react to losses at once, how stock market crashes happen.

The limits of self-awareness

If biases are predictable, one might expect that learning about them would make people immune. The evidence is more modest. Kahneman has said publicly that decades of studying these errors did not make him immune to them. In controlled studies, Morewedge and colleagues found that a single training session cut measured bias, but only partly: playing a training game reduced bias by at least 31.94 percent immediately and at least 23.57 percent two or more months later, while watching an instructional video reduced it by at least 18.60 percent immediately and at least 19.20 percent later (Morewedge et al., 2015). The games, which gave personalized feedback and practice, outperformed videos in the first test.

Those are meaningful gains, and they are also far from elimination. The research points toward a design approach, not a willpower approach. People who know that a bias exists often do better when the decision itself is restructured. Examples include automating contributions so a choice does not have to be remade each month, choosing broad diversified holdings so there are fewer stock-picking decisions, or setting rules in advance for what to do during a market decline. These are descriptions of how structure can reduce error, not recommendations for any individual's situation. Related reading: index funds versus active funds, dollar-cost averaging, and portfolio rebalancing.

When behavioral insights are used against you

The same findings that help design a better retirement plan can be used to extract money. The U.S. Federal Trade Commission's September 2022 staff report, Bringing Dark Patterns to Light, described several families of design practices that trick or manipulate consumers, among them disguised ads, fake countdown timers, hard-to-cancel subscriptions, and buried fees. Pre-checked boxes and hard-to-find cancellation steps exploit status quo bias in the same way a good default does, with the opposite aim.

Thaler and Sunstein themselves recognized the risk and argued for transparency, a test sometimes summarized as: would the designer be comfortable if the nudge were disclosed to the person being nudged? A related question is who benefits. A default that serves the chooser is a nudge. A default that serves the seller at the chooser's expense is usually called sludge or a dark pattern.

Common misunderstandings

"Behavioral economics says people are stupid." It says people are efficient under limits. The same shortcuts that cause errors in a brokerage account help us cross a street. The errors are predictable, which means they can be studied and designed around.

"Loss aversion is exactly two to one." The 2.25 value comes from one model fitted to one set of choices. Estimates vary by person, stakes, and context, and the concept itself is debated.

"Nudges always work." Some do, in large field trials, and the average effect reported in the literature is probably overstated. Context matters.

"If I know the biases, I am safe." Training helps by roughly a fifth to a third in the studies cited above, and it fades. Structure helps more than vigilance.

"It is just common sense." Some findings look obvious in hindsight, which is itself a bias. Many were contrary to what standard models predicted before the experiments, and the quantitative patterns, such as the asymmetry between gains and losses, are not what intuition alone would give.

The takeaway

Behavioral economics replaced the question "are people rational?" with a more useful one: "in which direction do people err, and under what conditions?" The answers, from anchoring to loss aversion to the pull of defaults, have changed how retirement plans are built and how governments communicate. The field's best-supported results come from large field studies and brokerage records, and its most debated claims concern how big the effects are and how well they transfer. The practical lesson is to treat decisions as designs: where a choice matters, the setting in which it is made, including defaults, timing, and automation, often counts for more than intention.

FAQ

Q: What is behavioral economics in simple terms? A: Behavioral economics studies how people actually make decisions about money, using experiments from psychology. It documents consistent departures from the rational-actor model, such as anchoring on irrelevant numbers, weighting losses more than gains, and sticking with defaults. It then uses those patterns to predict behavior and to design better choice environments.

Q: What is loss aversion, and how strong is it? A: Loss aversion is the tendency for a loss to feel more significant than a gain of the same size. The 1992 version of prospect theory used a coefficient of 2.25. Later studies find that the ratio varies by person and context, and some researchers argue the effect has been overgeneralized, so it is best read as a rough tendency.

Q: Did the 1974 wheel-of-fortune experiment really show anchoring? A: Yes. Participants who saw a rigged wheel stop at 10 estimated about 25 percent of UN members were African, and those who saw 65 estimated about 45 percent. The wheel was visibly random, and the effect appeared even when accuracy was rewarded in later work.

Q: Does automatic enrollment actually increase retirement saving? A: In the Madrian and Shea study of one large employer, automatic enrollment raised participation significantly, and many employees kept the default contribution rate and fund. Participation is not the same as total savings, because a low default rate can anchor people at a lower level than they might have chosen otherwise.

Q: Do nudges work? A: Sometimes. Defaults and some social-norm messages show clear effects in large field trials, such as the tax letters that raised payment rates by 1.3 to 5.1 percent. A 2022 meta-analysis reported an average effect of d = 0.43, while a bias-adjusted re-analysis found close to zero, so average effects are uncertain and context matters.

Q: Can people train themselves out of cognitive biases? A: Partly. In one set of experiments, a training game reduced measured bias by at least 23.57 percent two or more months later, and a video by at least 19.20 percent. The effects are meaningful but incomplete, which is why researchers often favor changing the decision setting rather than relying on awareness alone.

Q: How can behavioral insights be misused? A: They can be used to design dark patterns, such as pre-checked boxes, buried cancellation options, or fake countdown timers. The FTC's 2022 staff report describes several families of such practices. The distinction usually comes down to whether the design serves the person choosing or the party selling.

Sources

• Tversky A, Kahneman D. Judgment under uncertainty: heuristics and biases. Science 1974;185:1124-1131.
• Kahneman D, Tversky A. Prospect theory: an analysis of decision under risk. Econometrica 1979;47:263-291.
• Tversky A, Kahneman D. Advances in prospect theory: cumulative representation of uncertainty. Journal of Risk and Uncertainty 1992;5:297-323.
• Kahneman D, Knetsch JL, Thaler RH. Experimental tests of the endowment effect and the Coase theorem. Journal of Political Economy 1990;98:1325-1348.
• Odean T. Are investors reluctant to realize their losses? Journal of Finance 1998;53:1775-1798.
• Barber BM, Odean T. Trading is hazardous to your wealth. Journal of Finance 2000;55:773-806.
• Madrian BC, Shea DF. The power of suggestion: inertia in 401(k) participation and savings behavior. NBER Working Paper 7682, 2000; Quarterly Journal of Economics 2001.
• Thaler RH, Benartzi S. Save More Tomorrow. Journal of Political Economy 2004;112:S164-S187.
• Johnson EJ, Goldstein D. Do defaults save lives? Science 2003;302:1338-1339.
• Allcott H. Social norms and energy conservation. Journal of Public Economics 2011;95:1082-1095.
• Hallsworth M, List JA, Metcalfe RD, Vlaev I. The behavioralist as tax collector. Journal of Public Economics 2017;148:14-31. Summary: NBER Digest.
• Mertens S, et al. The effectiveness of nudging: a meta-analysis of choice architecture interventions. PNAS 2022. Maier M, et al. No evidence for nudging after adjusting for publication bias. PNAS 2022.
• Gal D, Rucker DD. The loss of loss aversion: will it loom larger than its gain? Journal of Consumer Psychology 2018. Mrkva K, Johnson EJ, Gächter S, Herrmann A. Moderating loss aversion. 2020.
• Morewedge CK, et al. Debiasing decisions: improved decision making with a single training intervention. Policy Insights from the Behavioral and Brain Sciences 2015;2:129-140.
• U.S. Federal Trade Commission. Bringing Dark Patterns to Light, staff report, September 2022.
• Thaler RH, Sunstein CR. Nudge. Yale University Press, 2008. Kahneman D. Thinking, Fast and Slow. Farrar, Straus and Giroux, 2011.

---

Read next:

• Behavioral Economics: Why Humans Make Bad Financial Decisions
• How Inflation Works: The Economics Explained
• Why Gold Is So Expensive: The Chemistry and Economics Behind the World's Most Coveted Metal