Behavioral Economics curriculum

Dual-process theories split thinking into two types: Type 1, or System 1, works quickly and automatically and leans on heuristics, while Type 2 is slow and draws on mental resources. System 1's quick answers can be helpful, but they often distort judgments when they clash with normative rules. The split is a model, not a settled fact: Gigerenzer, for one, does not accept the System 1 and System 2 framing.

Type 2, or System 2, thinking is deliberate and analytic, and it can step in to correct a quick intuitive answer. Because it needs time and cognitive resources, experimenters can largely shut it off with tight time limits or a competing mental task. One model holds that this cost is why people often keep their first answer; a rival model says both kinds of thinking run in parallel from the start.

In a study of 63 Chinese undergraduates, those who gave the stereotyped answer on base-rate problems, where a stereotype and the numbers disagree, were slower and less confident than people answering matched problems with no clash. Earlier studies report the same, so the conflict seems noticed even when not resolved. Noticing is not correcting: on the parallel-processing view, the error is a failure to override the intuition.

On the cognitive-miser account, wrong answers on reflection tests come from unwillingness to go beyond the first, effortless response. The evidence is mixed: in one study of 391 people, time spent on the questions was only weakly linked to accuracy, and earlier work found most correct solvers had already answered correctly under time pressure. Some errors reflect weak numeracy or faulty reasoning rather than too little effort.

Reflection-test problems invite an answer that springs to mind at once, and solving them means doubting that answer. Think-aloud studies of a verbal version found that most correct answers involved reflection and most lured answers did not, with exceptions in both directions. The older math version may also track numeracy, so a low score need not mean a lack of reflection.

When a judgment calls for one attribute, such as probability, people may judge by another that is easier to reach, such as resemblance, and the gap between the two becomes a systematic error. In three experiments, first drawing attention to the attribute actually asked about reduced base-rate neglect and ratio bias, but only when people were not under cognitive load.

Heuristics-and-biases research uses mental shortcuts to explain systematic departures from normative answers, and many accounts treat them as giving up some accuracy for less effort. That trade is not universal: tested on many data sets, two simple rules that ignore some information, Tallying and Take-the-Best, predicted new cases better than linear regression, especially when the training sample was small.

Striking or familiar events come to mind easily, so they can carry more weight than their real frequency warrants. A review of climate-risk studies reports that flood insurance purchases rose in areas recently hit by floods and fell back as memories faded. The mechanism is not always at work: one experiment testing personal anti-vaccine stories found no evidence that they persuaded through availability.

Televised images of the aftermath of plane crashes can push people's sense of crash risk above what the statistics support. That differs from overweighting a probability someone states: here the error comes from what memory supplies. Availability-driven fear can also fade: studies of home prices after floods and wildfires found prices fell, then recovered as the event slipped from memory.

Resemblance can stand in for probability, so a case that fits the prototype seems more likely than one that does not. In a chart review of 3,199 moderately to severely injured patients at nine non-trauma centers, only 24 percent were transferred to a trauma center as guidelines recommend; transfer was more likely when textbook signs of severe trauma were present, but most patients lacked them.

In the classic taxi problem, people typically answer 80 percent, the witness's accuracy, when combining it with how few of the cabs are blue gives about 41 percent. Critics note the setup matters: in a replication of a related problem, the neglect disappeared when people drew the case descriptions blindly from an urn. The normative answer combines both: the base rate sets the starting point and the case details adjust it.

A pair of events can never be more probable than either event by itself, yet people often rate the pair higher. Wording matters: in one study, asking which description is more frequent instead of more probable cut the error from 77 to 27 percent, and many people seem to read the word probability loosely rather than mathematically. Brief training on this and similar problems produced clear gains in one French study.

Early studies by Kahneman and Tversky found people draw conclusions from samples that are too small, and people expect even short runs of coin flips to come out nearly even, a pattern linked to the gambler's fallacy. Some researchers argue this reflects limited, finite experience with random sequences rather than a simple defect. Either way, the trap is treating a short run as the long-run pattern.

People picked because of an unusually high reading, such as high health costs, will likely measure lower next time even without any intervention, an effect of random measurement error combined with how extreme the first reading was. Mistaking that drop for a program effect is common, and evaluators rarely estimate it. An equivalent comparison group measured the same way separates real change from the drift.

On this account one feeling shapes both judgments, so an activity's rated risk and benefit move in opposite directions, though objectively the two should often be unrelated or even rise together. A survey of 602 Swedish adults found the pattern across domains and question formats; a small follow-up of 41 students found it weaker in those higher in cognitive reflection. Earlier work found time pressure strengthened it.

Asked which of two cities is larger, someone who has heard of only one can infer that the familiar city is bigger. The fast-and-frugal school argues that such a shortcut can be accurate when the environment links recognition to the answer. A reanalysis of 16 data sets found people relied on recognition more when they also knew something else about the option, so recognition alone does not decide how much people lean on it.

A first number pulls later estimates toward itself, and the adjustment away from it often stops short. In one experiment with 55 medical residents, a high anchor before a clinical case roughly doubled the odds of diagnosing the target condition compared with a low anchor. In two other experiments, consumers used prices as anchors only when those prices looked plausible against the price they already expected.

Ease can be a fair signal when it lines up with valid cues, but it misleads when it comes from something irrelevant, such as mere repetition. A fast-and-frugal rule built on it picks whichever of two recognized options comes to mind faster. The account has limits: in one study, the increase in believability from repetition was similar for foreign-accented and native speakers, against the fluency prediction.

A 2026 meta-analysis of 182 studies found a small but reliable effect that held for true and false statements alike. Judging whether a statement is true on first encounter was linked to a smaller rise later, though earlier work suggests that protection fades within about two days. A 2021 map of the field found all studies were run at Western universities and most used trivia statements.

Judged after the fact, the probability of the outcome that happened is rated higher than in foresight, without outcome knowledge. One study found the bias is malleable: people asked to judge again under foresight instructions lowered their estimates, and two weeks later their last stated answer, biased or not, rather than the outcome knowledge, determined their judgment.

A 2023 replication with 692 online participants found the same medical decision was rated better when it ended well than when it ended badly, even among people who said outcomes should not count. In a study of 191 Dutch physicians, rated quality of care differed sharply by outcome on three of four case vignettes. Judging the decision itself means asking what was known at the time.

Researchers separate three forms: overestimating one's own ability, placing oneself above others, and overprecision, an excessive belief that one knows the truth. In a study of 92 Austrian entrepreneurs answering general-knowledge questions, about 84 percent were overconfident, though not at every difficulty level. A review of four professions found overconfidence the most recurrent bias in the studies it covered.

One explanation is that planners focus on the details of the task at hand and set aside how long similar tasks actually took. Another is that people misremember past durations, since few time their own tasks. In three experiments, giving people feedback on how long an earlier task actually took did not reduce the bias in their next prediction.

Taking the inside view, a planner reasons from the parts of the current project; taking the outside view, from the record of comparable tasks, which works like a base rate. One account blames the planning fallacy on attention locking onto the case details. Knowledge of a comparable task mattered in three experiments: a second prediction was less biased when the earlier task had been the same task rather than a different one.

People often sense when an intuition clashes with a rule they know, and in think-aloud studies most correct answers on a reflection test involved reflection. Judging a claim's truth on first hearing is also linked to a weaker illusory truth effect. Slowing down is no cure-all: response time barely predicted accuracy on one reflection test, and after brief training most people answered correctly without deliberating.

The theory scores each option as a probability-weighted average of how much its possible outcomes are valued and recommends the option with the highest score. It is read both as a rule for how people should choose and, in classical economics, as a description of how they do choose; as a description, it makes faulty predictions in many real situations. Whether those departures are errors or reasonable preferences is debated.

Kahneman and Tversky built it on observed patterns that expected utility misses: reference dependence, loss aversion, diminishing sensitivity, and probability weighting. Options are first coded as gains or losses from a reference point. The original version could violate stochastic dominance, which the cumulative version fixes, and a study of 28.5 million online trades found its reflection and loss-aversion patterns.

The reference point can be current wealth, an aspiration, or an imagined comparison, so the same result can register as a gain or a loss. In one line of studies, shoppers who had missed a bigger discount used it as their reference, so a smaller later discount felt like a loss and was often passed up. Moving the reference point can change the choice without changing the outcome.

Tversky and Kahneman's 1992 estimate put a loss at 2.25 times an equal gain. A 2024 meta-analysis still found losses weighing more on average, but less: a mean near 2 and a median near 1.7. In one set of four experiments, changing the range of gains and losses people had just seen produced loss aversion, neutrality, or the reverse, and other experiments found none for small stakes.

Going from nothing to a modest gain feels bigger than adding the same amount to a large gain, and losses work the same way. That makes the value curve concave for gains and convex for losses, whereas the utility curve in expected utility theory is typically one or the other, not both. Prospect theory uses this shape to predict playing safe with gains and gambling with losses.

In the classic disease problem, most people chose the sure option when outcomes were described as lives saved, but most chose the gamble when the same outcomes were described as deaths. A 2020 survey of 88,181 people in 47 countries replicated this framing effect, which varied substantially across nations. When even-odds outcomes are learned from experience rather than stated, the pattern can reverse.

Prospect theory links this to a value curve that flattens as gains grow, so a sure gain can feel worth more than a chance at a larger one. The pattern responds to wording, not just payoffs: a 2026 study with a large sample matched to US Census data found a gain-loss framing effect even when both frames described every outcome fully and in matching terms, a case where it had been thought absent.

In prospect theory's estimates, people overweight probabilities below about one third and underweight those above it, and changes near zero and one count for more than expected utility allows. The distortion fades, or for rare outcomes can flip, when people learn odds by sampling. In one experiment with 236 Chinese undergraduates, those told to take a distant, third-person view of each lottery distorted less than a baseline group.

In a hypothetical game of Russian roulette, people will pay more to remove the last bullet than to go from four bullets to three. The Allais choices show the same pull: preferring a sure large prize to a gamble with a bigger upside and a small chance of nothing, yet preferring the bigger prize when both options are uncertain, fits no expected utility function, though it can seem rational.

Moving an outcome from impossible to merely unlikely changes how an option is valued more than expected utility predicts. Prospect theory offers this as one explanation for both buying insurance against rare losses and betting on rare gains. The effect is clearest with stated odds: when people learn about rare outcomes through experience, they tend to underweight them instead.

The classic account treats giving up an owned item as a loss, which looms larger than the matching gain. Other work points elsewhere: in a meta-analysis and three experiments, owning an item did not change beliefs about its quality or market price, and the authors read buyers and sellers as acting sensibly on those beliefs. The effect also vanished among Kanak participants tested in a Kanak setting.

In 2026 survey research on US drug policy with 5,053 respondents, support for keeping familiar legal substances legal was higher than for invented drugs with the same properties, and the reverse held for illegal drugs. Information about harm narrowed the gap for legal substances. In a lab study of health insurance choices, a default pulled choices toward it at first, but the pull faded as people repeated the decision.

In hypothetical vaccine choices, earlier experiments found that many people decline a vaccine even when its side-effect risk is lower than the risk of the disease it prevents. The bias does not show up everywhere: one experiment with 202 women imagining a sick child found no difference in how severe side effects and disease symptoms seemed, or in how well they were recalled.

Past costs cannot be recovered whichever way you decide, yet they still pull people toward continuing what they started. A 2023 replication with 821 US online participants confirmed a weaker effect for time than for money in one test but found it for both in another. Classic scenarios used to measure the effect correlate only weakly with one another, so different vignettes may not measure the same thing.

The classic setup has someone invest, learn the investment is going badly, and then decide whether to put in more. Personal responsibility for the first choice was thought to drive the effect, but a preregistered replication found only partial support for that and failed to replicate a related study. Two experiments found evidence that people who inherited a goal, rather than setting it or being assigned it, escalated less.

A loss frame and a gain frame can describe exactly the same deal, so only the wording differs. Sun-protection researchers often expected gain frames to persuade more, yet meta-analyses of sun-protection studies and physical-activity trials found no overall difference between the two. Because the frames are equivalent, restating a loss-framed pitch as a gain shows whether anything but the wording changed.

The endowment effect would predict that a product already in use feels like something to give up rather than turn down, but that link is a conjecture and the effect itself is debated. The documented risks are plainer: the FTC warns that trials usually need a card number and will probably charge you if you do not cancel on time, and one review counts a missing end-of-trial reminder as a barrier to canceling.

Overweighting small chances of loss can make paying to avoid them attractive. In the Netherlands in 2014, only 11 percent of the insured took a voluntary deductible for a lower premium, though in hindsight the highest one would have paid off for almost half; the reasons are still only proposed. Climate-risk studies report the opposite error: rare disasters are often underweighted until one strikes, and then insurance demand jumps.

A ticket pairs a very small probability with a large prize, and overweighting small stated probabilities makes the bet look better than its odds justify. Tversky and Kahneman offered this as one explanation of gambling. Players also misread chance in other ways: a study of lottery betting found people were less likely to bet on numbers that had recently won, as the gambler's fallacy predicts.

An SEC investor bulletin, drawing on a Library of Congress report, lists it among investing behaviors that can undermine performance, and notes that winners sold often go on to outperform the losers still held. One study linked greater corporate transparency to a weaker disposition effect among the firm's individual shareholders, mostly because they became less likely to sell stocks held at a gain.

When people have not adjusted to a recent loss, a risky bet that could erase it can become attractive, and prospect theory treats that as risk seeking to offset the loss. Real gambling data complicate the picture: in one online game with over 2,500 players, players were less likely to stop and raised their stakes after wins rather than losses, though players did play faster after losing.

Shown two choices one at a time, a sure gain or a gamble and then a sure loss or a gamble, many people end up with the lower-value combination. In a 2025 replication with 896 US participants, among those picking one of the two key pairs, only 5 percent chose the higher-value pair when the choices came separately, against 93 percent when the combined outcomes were shown as one choice.

Seen one at a time, a daily cigarette or doughnut can seem trivial, while the year's total makes the consequences hard to ignore; a classic review argued broad brackets tend to bring choices closer to long-term goals. A 2025 replication supported six of seven classic demonstrations, but widening the bracket further added nothing, and the authors call for a clearer definition of the idea.

One theory holds that people make regret-minimizing choices, which can be cautious or risky depending on the situation. In one experiment with 133 Chinese undergraduates, prompting anticipated regret in a balloon-pumping risk task made choices more cautious and more satisfying. Yet a 2025 preregistered replication failed to reproduce a classic study of how feedback about forgone outcomes shifts risky choice.

Because outcomes are coded as gains or losses from a reference point, often the status quo, and losses loom larger than gains, the option already in place can start with an edge. A 2026 US survey exposed that edge by pairing real drugs with invented ones of the same properties: real legal substances drew more support for legality than their invented matches, and real illegal ones drew less.

Classic decision theory assumes description invariance: if two wordings state the same outcome, preferences should not change. Framing research shows they often do, and in a predictable direction. One review adds a caution: a speaker's choice of wording can carry real information, such as what the speaker expects or prefers, so reacting to a frame is not always irrational.

Beef described as 80 percent lean is also 20 percent fat; the positive label is thought to call up positive associations. One review judges the evidence limited but mostly positive. A meta-analysis of food studies found a large effect on attitudes but almost none on intentions, and letting people actually try a product can shrink or remove the effect.

In the classic disease problem, about 75 percent chose saving 200 of 600 people for sure, yet about 75 percent preferred the gamble when the sure option read as 400 deaths. Drawing on meta-analyses, a large replication project and a 49-country survey of over 100,000 people, one review puts the effect at about 0.6 standard deviations, while noting that about a quarter of classic participants did not follow the pattern.

A health message can stress the benefits of being screened or the costs of skipping screening. Explanations draw on loss aversion, but meta-analyses report very small, mostly nonsignificant effects, and a 2021 review pooling 165 effect sizes from about 50,000 people found almost no average difference in persuasiveness, with likely results in a new study ranging slightly either way.

When 640 New York area sports fans were surveyed before a Giants home game, tickets described as a percentage remaining seemed less available than the same supply given as a count, most of all for high-demand games. Earlier research finds that both percent-off and money-off promotions raise perceived savings; in one brain-recording study, people low in numeracy judged price differences worse under either form.

The original problem asked whether people would make a 20-minute trip to save 5 dollars on a 15-dollar purchase versus a 125-dollar one, and far more agreed for the cheap one. In a 2025 registered replication with about 500 US online participants, most refused the trip at either price. The authors caution that changed costs since the 1980s, or sample and design differences, may explain the failure, so it does not prove the effect is gone.

In one online study, people said they would walk or drive a greater distance to save the same amount of money when they answered in miles rather than meters, which the authors attribute to the smaller number that miles produce. Mental accounting research similarly notes that, when bills come monthly, small monthly income affects spending more than a larger sum received once a year.

In a 2025 registered study that tested this prediction of Thaler's, a yearly membership described as merely 27 cents a day was rated more attractive than one priced at 100 dollars a year, a nearly identical cost. The gap shrank but held when both versions were shown side by side. Thaler's proposed reason is that people tend to ignore small, routine expenses.

In the classic theater problem, people who lost a 10-dollar ticket were less willing to buy another than people who lost a 10-dollar bill, and a 2025 registered replication confirmed it with a smaller gap. The likely reason is that the replacement feels charged to a ticket account already spent, while lost cash sits in no account. Mental accounts are also thought to simplify decisions and aid self-control.

In Thaler's account, prior gains encourage risk seeking. A 2025 registered replication found no such increase after an imagined 30-dollar win, while a meta-analysis of 36 studies of spending and risk-taking after unexpected income found a low-to-moderate effect with high variation: larger in questionnaires and among students, and weak and nonsignificant in field experiments.

In one study described in a 2025 meta-analysis, students told that a tuition refund was a bonus spent more of it than students told it was a rebate. Across 36 studies, unexpected income raised spending and risk-taking on average, but the effect was nonsignificant in field experiments, some studies found the reverse, and the range of likely future results ran from a decrease to a large increase.

Heath and Soll found people less willing to buy a theater ticket after spending on a related sports ticket than after an unrelated dinner. A 2025 registered replication confirmed the dinner comparison, but an unrelated flu shot behaved like the sports ticket, so whether relatedness is the trigger is unclear. Mental budgets are thought to cut the effort of spending decisions, aid self-control and prevent overspending.

Researchers argue that a credit card pushes the bill weeks past the purchase, so its cost is less salient at checkout than handing over cash. In experiments with students choosing among three options, paying by card weakened the usual pull toward the middle option, which the authors traced to less pain of paying. Wine bought as futures to drink years later is likewise seen more as investment than spending.

Research reviewed in a 2025 paper holds that cash produces the most payment pain and that credit cards raise willingness to spend. In a brain-scanning shopping study using people's own cards and cash, card purchases were tied to reward activity, while the insula, a region earlier linked to payment pain, did not separate the two. A phone-payment study found payment pain mattered only at higher prices.

Shafir and Thaler posed questions to wine collectors about bottles bought long before drinking, and a 2025 registered replication with a general online sample confirmed key results: people did not judge a bottle's cost by its current market price, and an advance wine purchase was seen more as an investment. The proposed reason is that such an expense may be booked as an investment, avoiding the feeling of spending.

In a scenario replicated in 2025, people would pay more for the same soda from a fancy resort hotel than from a run-down grocery store, which Thaler explained as the setting raising the reference price; the effect was small. Deal feelings respond to framing too: in an online experiment with 342 people, a free-gift label kept mismatched bundles feeling like a smart buy where a discount label did not.

The Federal Trade Commission's pricing guides say a former price is a legitimate comparison when the product was genuinely offered at it, on a regular basis, for a reasonably substantial period, and that inflating a price briefly to stage a big cut makes the bargain false. Lab studies also find that placing the regular price to the left of the sale price raises people's estimates of the reference price.

Some consumer studies find partitioned prices can raise purchases, while a 2026 brain-imaging study in which 40 people judged shoes at a constant total price found a clear preference for free shipping over a lower price plus a shipping fee. Under the Federal Trade Commission's fee rule for tickets and lodging, sellers may itemize mandatory fees as long as the total price stays the most prominent figure.

A 2022 Federal Trade Commission staff report says drip pricing hampers price comparison, and it cites a study in which ticket buyers not shown fees up front spent about 20 percent more and were 14 percent more likely to complete the purchase. Since May 2025, a commission rule has required the total price up front for live-event tickets and short-term lodging.

The Federal Trade Commission's Rule on Unfair or Deceptive Fees, in effect since May 2025, requires sellers of live-event tickets and short-term lodging to show the total price up front and more prominently than most other price figures. Taxes, shipping and truly optional add-ons may come later but must be disclosed before payment. The commission expects upfront totals to make comparison shopping easier.

In a 2022 survey of 2,527 US primary care staff, the smallest incentive said to motivate extra effort averaged about 2,155 dollars as a bonus but 1,185 dollars as a penalty, roughly half as much. These were stated preferences, not observed behavior. A card surcharge and a cash discount can be the same price gap described from opposite sides.

Earlier studies offered self-control reasons, such as wanting to keep track of money, along with perceptual fluency and even the sense that small notes are dirty. Later work complicates the picture: across seven experiments with about 4,000 people, including two field studies, people tended to pay with the denomination that matched the product's price, whether large or small.

Mental accounts break fungibility, the principle that a dollar is a dollar wherever it sits, though they are also thought to help with self-control. In five experiments with 2,325 people, reminders of what else the money could buy reduced willingness to purchase about equally for low- and high-income participants, which suggests both groups tend to neglect opportunity costs.

Studies that described a sure option in both terms, such as 200 saved and 400 not saved, found the classic framing effect disappeared, though other wording changes can increase or even reverse it. In a 2026 US replication, a framing effect persisted among 477 people choosing between completely described, matched options. Argentine medical trainees showed smaller framing effects after statistics courses, in a before-and-after comparison.

In an online experiment with 801 UK adults, framing comfort care as a refusal of treatment cut the share choosing it from 84 to 65 percent. A review of side-effect warnings, such as 70 percent will not get a headache versus 30 percent will, found positive framing promising for reducing nocebo effects, but only six studies had tested it, and key questions remain open.

In Tversky and Kahneman's demonstration, people first said whether the share of African countries in the United Nations was above or below a number from a wheel of fortune, and their guesses tracked that random number. One paper's literature review calls the effect robust for numeric estimates such as general knowledge, but inconsistent when people state what goods are worth to them.

An influential study reported strong anchoring on what people would pay for goods like wine and small electronics, but later studies found weaker effects or none. A meta-analysis of 53 studies found a moderate effect, and in a set of three experiments, arbitrary anchors shifted willingness to pay substantially only when answers were hypothetical, not when real money was at stake.

Tversky and Kahneman proposed that people start at the anchor and adjust, halting at the edge of the plausible range, so a low anchor leaves estimates low and a high one leaves them high. A rival account holds that an anchor makes matching information easier to bring to mind. The two are not mutually exclusive, and the claim that adjusting requires mental effort is contested.

Negotiation research reports that first offers shape counteroffers and settlement prices, and buyers are often advised to open. Newer studies find limits: in online scenarios, anxious buyers with weak alternatives made first offers less favorable than going second would likely have yielded, and a replication attempt found that cold cues did not reduce anchoring. Earlier work found that weighing the other side's alternatives shifted outcomes.

The Federal Trade Commission's pricing guides describe a pen normally sold at 7.50 dollars, offered at 10 dollars for a few days, then advertised as cut back to 7.50, and call that bargain false. Effects on buyers vary: in an online experiment with 271 adults over 40, high versus low advertised hearing aid prices did not change rated purchase likelihood overall, though exploratory analyses found effects among people with poorer hearing.

Adding a larger, pricier option to a set can turn the old top choice into the middle one, and middle options tend to gain share. A very high price might also act as an anchor. But arbitrary price anchors often shrink once real money is at stake, and the showpiece idea extends these findings rather than resting on a direct test, so it is plausible rather than established.

In sequential rating tasks, judgments are biased by what came just before: contrast pushes a rating away from it, and assimilation pulls it closer. One study found both in face and ringtone ratings, and feedback about the previous stimulus's value removed the contrast. In a dating-style task, a face was more often rated attractive when the face before it was attractive.

Lab studies since the early 1980s have shown it, but concerted replication attempts found it only in stylized lab tasks with numeric attributes. In 3.6 million UK grocery wine purchases, dominated decoys did make their target wine more likely to be chosen, but by roughly 1 percent. Small per shopper, a shift that size can still add up to large sums for a seller.

If car A beats car D on price and fuel economy while car B beats D only on price, adding D often raises A's share, breaking the rule that a new option cannot help an existing one. Accounts differ: one says D makes people weight A's strong attribute more, another that A just wins more comparisons. Familiarity with the product domain weakens the effect.

One account holds that when people are unsure which attribute matters most, the middle option is easiest to justify. In a study of 241 students in Toronto and Paris making hypothetical health coverage choices, more detailed coverage information shifted choices toward the middle tier, away from both the basic and the top plans. The authors left open whether that shift was good or bad for the people choosing.

An extreme option is clearly worse on one attribute, so it can seem riskier than a middle option that loses out on none. In a survey of pork buyers in one Chinese city, the pull toward the middle was weaker when people chose first without the extra options and saw them only afterward, which suggests a choice made before seeing the full range is less swayed by position.

Three tiers invite the compromise and decoy effects: adding a top tier turns the old top tier into the middle option, and an inferior option can raise its neighbor's share, as decoys did slightly in grocery wine sales. In a study of hypothetical health plan choices among Basic, Enhanced and Superior tiers, more coverage detail moved choices toward the middle tier.

Simple scalable choice models, a foundation of normative economic theory, imply that adding an option cannot reverse the ranking of the others or raise any one's share. The attraction, compromise and similarity effects all break those rules, which is why they matter for theory. Measuring them is tricky: a reanalysis of five published studies found that a more robust test changed the conclusions in at least a quarter of cases.

In the classic version, many people choose a bet with a high chance of a modest prize over one with a small chance of a large prize, yet assign the long shot a higher money value. The pattern has puzzled researchers for more than half a century. A 2024 test of one explanation, that pricing a bet in isolation skips comparison with its rival, found no support.

In the famous jam study, a 24-flavor display drew more shoppers than a 6-flavor one, but 3 percent of those who stopped at the large display bought, against 30 percent at the small one. One meta-analysis put the average overload effect at virtually zero with wide variation, another tied it to specific conditions, and a preregistered test of 2 to 80 charity options found no effect on giving.

A later meta-analysis found overload reliably appears when any of four factors is high: set complexity, task difficulty, preference uncertainty, or the decision goal. A reanalysis found that when those factors are low, more options can even help on some measures. In a test of 2 to 80 charity options, overload was absent whether or not people had settled preferences.

One review notes that insurance purchases jump the year after a catastrophe and then decline as the event fades from attention. It also notes that if shoppers attend only to a base price and ignore hidden add-ons, sellers with transparent prices can lose out. For tickets and lodging, the Federal Trade Commission's fee rule requires the all-in total to be the most prominent price.

On number lines, adults and children from about age 8 place numbers like 698 too far left of 701. A review cites a US study in which used cars just under 80,000 miles sold for about 200 dollars more than cars just over it. Elsewhere the bias is inconsistent: a large online pricing experiment found none, and Japanese cardiac-arrest records showed none at age thresholds.

One paper's literature review notes field studies reporting revenue gains after companies adopted just-below prices, and lab work in which such prices made a discount be remembered as larger. Yet in its own online experiment, 266 people made 4,788 purchase decisions on identical lottery tickets with no left-digit effect, and the authors conclude the effect may be weaker than suggested.

The perceptual fluency account holds that a round price such as 2.00 dollars is easier to process and so feels more attractive, and earlier studies found people lean toward round amounts when they set their own prices. That clashes with the left-digit account of 1.99 dollars. An online experiment that pitted the two against each other with otherwise identical lottery tickets supported neither.

In a lab task with 60 volunteers, a fictitious drug that had no effect was judged more effective when described as expensive than when described as cheap. Price can cut both ways while shopping: one experiment found a high price improved attitudes when adding a product to a cart, through perceived quality, but lowered them at the order stage, through felt sacrifice.

In a grocery experiment with 250 US shoppers, people under mental load were more likely to choose a food bundle even when it carried no discount, suggesting bundles appeal partly because they save effort. In a lab study, a bundle with a free tie-in product was chosen more often than one with a practically identical total split across both products, so in both cases the bundle won without a price advantage.

The Federal Trade Commission's example is a resort charging a 199-dollar nightly rate plus a mandatory 39-dollar daily resort fee; under its fee rule the fee must be included in the total price shown. Sellers may still itemize, but the total must be the most prominent figure. Research finds that splitting a price can change how an offer is judged even when the total is unchanged.

In New Hampshire primaries and general elections between 2000 and 2016, candidates gained from being listed first in 86 percent of 84 tests, about 1.5 to 1.7 points for the 2016 major presidential candidates. As theory predicted, the gain was larger in primaries, in less publicized races and in contests without an incumbent.

In an online supplier-rating experiment, asking people for two reasons a past score was too high weakened a high anchor's effect on some ratings but not a low anchor's, which responded to a different technique. Anchoring has also looked weaker for familiar goods, when real money is at stake, and, in one exploratory analysis, among buyers who rated their knowledge higher.

Classic economic theory says a delayed outcome should lose the same proportion of its value for each added period of delay, which is called exponential discounting. Health outcomes break the pattern for some people: in most studies, roughly 3 to 10 percent prefer to bring an illness forward or put off an improvement, the reverse of usual discounting.

People discount rewards in the near future more steeply than distant ones: delaying an immediate reward by one day cuts its value proportionally more than delaying a distant reward by a day. Hyperbolic curves fit such cross-sectional choices better than exponential ones. Studies with money that follow the same people over time, though, have not consistently found the preference reversals hyperbolic curves predict.

The quasi-hyperbolic model keeps a steady discount rate but multiplies every future period by an extra factor, called beta, so the present stands apart from all later moments. In a representative survey of Hungarian adults, present bias was linked to saving decisions and financial difficulties. The survey was not incentivized and shows an association, not that present bias causes money problems.

Asked a week ahead, people in one study chose a healthy snack more often than when the same choice was offered for eating right away. Such reversals fit hyperbolic discounting, but the evidence is mixed: in a weeks-long study with real money, reversals toward the sooner amount were not significantly more common than reversals the other way, and many health-related reversals follow shifts in motivation or surroundings.

People often make far-sighted plans when outcomes are distant, then reverse in favor of short-term rewards once the moment arrives, one account of why intentions go unfulfilled. Economists call this myopic preference reversal. Binding the future self in advance requires awareness that future preferences may not match current goals, so only someone who foresees the conflict can plan against it.

Behavioral accounts frame procrastination as preferring an immediate competing activity over a delayed task. In one study of New York University psychology students tracked through a semester-long research requirement, steeper discounting of future rewards correlated with more procrastination. The link showed up only with a fine-grained measure of work progress, not with completion day or a survey.

Thaler and Shefrin described temptation as two coexisting, inconsistent sets of preferences, one concerned with the long run and one with the short run. Some researchers have looked for its basis in competing brain systems, but one review cites substantial evidence against the idea that systems for impatience and patience compete during discounting tasks.

Commitment devices impose costs for falling short, add no reward for success, and are entered voluntarily, such as depositing money that comes back only if a diet improves. Taking one up is not the same as changing. In a field experiment, over a third of middle school students opted into a device tied to better school conduct, yet enrolling did not change their behavior.

Federal consumer guidance calls recurring automatic transfers from checking to savings one of the easiest ways to save consistently. Stricter is not always better: in a Ugandan school program, children saved more in a loosely restricted account than in one locked to school expenses. In a small app trial, fewer people took up deposit contracts than rewards, and deposits were not significantly more effective.

Thaler and Benartzi designed the plan to work around hyperbolic discounting: retirement contributions rise automatically with future raises rather than coming out of current pay. A later paper on saving describes it as effective at getting employees to join pension plans and raise their saving rates. Because it depends on raises arriving, one review argues that stagnant wages blunted its effect.

The pairing joins something pleasant now with a behavior that is not enjoyed in the moment, such as exercising while listening to a favorite podcast. An oral-health toolkit offers it as a way to raise adherence, with phone time as the reward that comes only after brushing teeth. A review of exercise maintenance gives a similar case: watching television solely while on a gym treadmill.

Friction raises the time or effort an action takes. In a three-week trial with 70 iPhone users, an app showing a full-screen prompt to quit or continue once a session passed a self-set limit cut self-reported daily time on each person's most problematic app by about 29 minutes versus controls. Problematic social media use scores did not drop significantly, and the authors note that app blocking can provoke resistance.

Loewenstein proposed that people ignore or underestimate visceral drives they are not currently feeling, and in one study smokers who were not craving underestimated how much they would later value a cigarette. In a brain-imaging study, people paid more to avoid eating unpleasant foods when the choice was real than when it was hypothetical, consistent with disgust being underrated in the abstract.

Projection bias is wrongly carrying a current state into the future when preferences change. In an auction, hungry participants bid a higher premium for cheese than sated ones, even for cheese due when they would be full. In United States sales data, buyers grew likelier to pick an Energy Star air conditioner as the prior week's temperature moved away from 20 to 22 degrees Celsius, which the authors tie to projection bias and salience.

People usually forecast whether a future feeling will be good or bad fairly well, but misjudge its intensity and duration. Accuracy varies: in two studies of real decisions, ranking residency programs and voting, people forecast emotional intensity more accurately than frequency or duration. Forecasts also miss for routine experiences, as when people predict how happy their meals will make them.

Brickman and Campbell likened emotional adaptation to sensory adaptation, implying that lasting gains in happiness are out of reach. Revisions proposed by Diener and colleagues hold that most set points are positive, that they differ by person, and that people differ in how fully they adapt. In over 3,500 older Europeans, well-being recovered after short-term disability but adapted much less after lasting disability.

In patients having painful procedures, remembered pain tracked the worst moment and the final minutes, and long procedures were not remembered as especially bad. In 64 women, a horror clip that ended at its scariest point left more anxiety in hindsight than a longer version with a milder ending. For a complex virtual reality experience, average feelings predicted memory better than peak and end.

In a foundational study, people held a hand in cold water for 60 seconds and, on another trial, for 90 seconds with the last 30 slightly warmer; most chose to repeat the longer one. Duration is not lost from memory but carries little weight. In a small vocabulary-app study, doubling task length did not significantly change how pleasant it was remembered, which the authors call consistent with duration neglect but not a strong test.

In pain studies, the experiencing self goes through more total pain in a longer episode, while the remembering self may judge it less painful because it ended better. In a study of backcountry skiers, facial expressions showed less happiness while skiing than during breaks, and moment-by-moment and self-reported emotions were largely unrelated.

In a negative option, a business bills automatically until the customer cancels, as when a free box turns into paid monthly shipments. The Federal Trade Commission (FTC) says trouble arises when billing terms are unclear or canceling is hard, and it has alleged that one learning site promised easy cancellation while routing users through a long, confusing path.

Trials usually collect a card so the seller can bill a customer who misses the cancel deadline. The Federal Trade Commission lists further traps: shipping fees on a supposedly free offer, pre-checked boxes that can authorize charges after the trial, and online ads written by affiliate marketers who are paid for every click, some with exaggerated claims.

A lower monthly payment often comes from stretching the loan, and the Consumer Financial Protection Bureau notes that a longer term means paying more interest over the life of the loan. The Federal Trade Commission adds that low-payment car loans often carry longer terms and higher rates, and that add-ons can slip into a deal. Both agencies point to the total cost, not just the payment.

Ariely and Wertenbroch studied deadlines as a form of self-control by precommitment. A replication of their second study, with adults at a large United States public university, did not reproduce the original results: changing the deadlines had a negligible effect on the performance measures, and evenly spaced deadlines set by the experimenters did not stand out at reducing procrastination.

The goal-gradient hypothesis holds that motivation grows as the distance to a goal shrinks, as when rats run faster the closer they get to food. In lab studies and a television game show, people put in more effort near the end of chance-based games, and their reports suggested the change was unintentional. Under some conditions, effort instead slows as a goal nears.

Choices made ahead of time can differ from choices made in the moment, as when people in one study picked healthier snacks for next week than for right now. Commitment devices build on this: people may adopt them in self-aware moments to resist later temptation. The calm state has a blind spot, since people tend to underestimate cravings they are not feeling, and in one school experiment signing up did not change behavior.

Descriptive norms can steer behavior through social proof because they signal what has worked for others. In five German field experiments, a stick-figure sign stating what most people do and approve of cut paper towel use and raised mask wearing, but it had no effect on plastic coffee lids, so the same nudge can fail in a new setting.

Cialdini and colleagues' focus theory separates descriptive norms, the typical behavior in a group, from injunctive norms about what is approved. Several meta-analyses find both kinds of appeal promote sustainable behavior when a majority already does it. In energy field experiments, social norm interventions mostly produced small reductions, and in one online giving game descriptive norms had no visible effect.

Injunctive norms signal what a group approves or expects, so they are thought to work through social pressure to conform. In an online dictator game, a suggested donation raised the chance of giving at least that amount, while a statement of typical giving did not. For behavior only a minority performs, a review of 54 studies found injunctive and growing-trend appeals less prone to backfire than static minority figures.

When a target behavior is rare, a descriptive norm message can produce the opposite of what was intended, because it tells people their undesirable behavior is typical. Stressing that a behavior is spreading, or that most people approve of it, is proposed as a way around this. Yet a secondary analysis of four alcohol studies found no sign of such a boomerang among students drinking below the reported norm.

In evolutionary models, following the majority often beats other social learning strategies, but formal analyses suggest consensus loses reliability when people's judgments are not independent. The Securities and Exchange Commission warns that jumping on the bandwagon with heavily promoted stocks may mean significant losses, and that fraudsters can spread false claims to create a buying frenzy.

Bikhchandani and colleagues showed that once early choices line up, later deciders can do best by following them, so a cascade can lock in an option even after most private evidence points the other way. Lab versions find many decisions follow this logic, yet participants tend to overweight their own private information compared with the rational benchmark.

A mania or bubble is a rapid price rise reflecting collective enthusiasm, usually followed by a sharp fall when wide-scale selling begins. Momentum investors bet that recent trends will continue, and noise traders buy or sell without fundamental data. The Securities and Exchange Commission cites a Library of Congress report that lists such behaviors among those that can undermine investment performance.

In a replication at the University of Bern, participants matched the group's wrong line-length answer on 33 percent of critical trials, and 25 percent when correct answers were paid. Researchers separate normative influence, yielding to social pressure, from informational influence, yielding under uncertainty. Most answers stayed independent, a side of conformity research that only 15 percent of 30 surveyed psychology textbooks reported.

Pooling 21 of Milgram's conditions with 740 participants, a synthesis found that 43.6 percent continued to the maximum shock, with the experimenter's directiveness, legitimacy and consistency among eight factors that influenced obedience. In a medical version of a sequential choice task, people weighted a prior decision more when it came from someone of higher rank, most of all when it contradicted their own information.

Social exchange theory holds that an unsolicited favor creates a felt obligation to repay it. A Cochrane review cited in one trial found up-front payments raised the odds of a postal survey response by more than half compared with delayed payments. That trial itself, a telephone survey of 720 primary care patients in Hong Kong, found no difference by timing, so the effect may not hold in every setting.

In Washington, DC in 2013, Medicare prescribers who received drug-company gifts wrote 2.3 more claims per patient, at $50 more per claim, with 7.8 percent more brand-name drugs than non-recipients. The study could not establish causation, and companies may target those who already prescribe costlier drugs. In a field experiment, food vendors served larger portions to customers who complimented or tipped them.

Commitment theory predicts that a public, low-cost act of support makes consistent follow-up more likely, an effect found in several domains but not replicated in every study. The rival slacktivism account predicts that the token gesture replaces real action. Pooling three blood-donation studies, one team saw an unconfirmed tendency for an initial act to raise later support when people were first prompted to think about helping others.

Commitment theory predicts the effect, and one review judged that it appears more often than chance would allow. Conditions matter, including feeling free to refuse, having no outside reason to comply, the cost of each request and the delay between them. Accounts of the literature also describe many failures to replicate and occasional reversals, where the small request lowered later compliance.

A review of energy field experiments notes that public commitments appear more effective than private ones, and that people may also be more willing to commit in public. A weight-loss review found commitment devices appeared most successful when made publicly, though on limited evidence. The same energy review found commitment interventions' effects essentially zero for the most part.

Research on advice treats perceived similarity and expertise as key predictors of a source's impact, partly because similar people are assumed to share the receiver's needs. In two online experiments, each added shared trait raised ratings of a new virtual partner's likability, with music taste, religion and ethical views as consistent predictors. Those studies measured liking, not compliance.

Unity rests on a sense of we-ness: Cialdini argues people trust, favor and copy those they count as their own group. A research agenda on lobbying notes that advocates often approach officials who share their nationality or language. In an experiment with 240 Arab citizens of Israel, messages stressing benefits to the group raised stated intentions to keep social distance, while self-focused messages lowered them.

Marketers signal that a product is scarce because of high demand or limited supply, and both kinds of message have been found to lift product evaluations. In three experiments, compared with demand-based messages, supply-based scarcity raised purchase intention for consumers with a strong need for uniqueness but lowered it for those with a weak one. The FTC lists fake countdown timers among misleading dark patterns.

The original definition pairs a pervasive worry that others are having rewarding experiences one is absent from with a wish to stay continually connected to them; some definitions extend it to missing a profitable investment. A 2024 scoping review of 106 articles found agreement on that core but a fragmented literature, with debate over its boundaries and a stated need for causal research.

On a United States physician-rating site, prospective patients leaned on expert ratings when patient ratings were few, but on patient ratings once they grew numerous; a lab study replicated this. The FTC rule targets fake or AI-generated reviews, pay conditioned on positive or negative sentiment, undisclosed insider reviews, threats used to remove negative reviews, and fake followers or views.

The FTC counts payment, free or discounted products, and family or employment ties as material connections, because knowing them lets followers weigh an endorsement. A disclosure belongs with the message itself, not buried in hashtags or a profile. In one experiment with 406 adolescents, younger teens needed fuller disclosures than older ones to recognize the persuasion, and purchase intention was unaffected for all.

In negotiation experiments, people routinely turn down final offers that are better than nothing. When brain and facial muscle activity were recorded, unfair offers drew frowns, but frowning did not predict rejection. Rejection dropped when refusing would not cost the proposer anything, even though reported anger rose, which the authors read as rejecting to restore equity rather than to vent.

A proposer offers a division of a sum, and a rejection wipes out both shares. Proposals are more equal than in the dictator game, where responders must accept, suggesting proposers expect rejections. In a modified version with 1,370 participants, 53 percent chose not to learn whether the offer they received was unfair; they rejected 6 percent of offers, while those who checked rejected 39 percent of unfair ones.

Costly punishment, harming someone at a cost to oneself, occurs even in one-shot interactions and is often credited with sustaining cooperation. A review notes that most experiments showing punishment helps cooperation prevent targets from retaliating; when allowed, punishment often provokes retaliation. In a foraging society, every mechanism tested raised cooperation, but only social-image incentives significantly raised average earnings.

Pure altruism is the reward that cannot be traced to impure motives such as warm glow, the status of being seen giving, or expected reciprocity. One account holds that people help partly because they anticipate that good feeling. It can be undercut: in a series of experiments, learning about children who could not be helped reduced the warm glow from aiding those who could.

The Federal Trade Commission has described comparison sites that claim neutrality but rank companies by what they are paid, review sites a seller secretly controls, and fake followers and views for sale. The Securities and Exchange Commission describes pump-and-dump schemes that spread false claims to start a buying frenzy, then sell at the inflated price. Each dresses an interested party up as independent evidence.

Choice architecture covers how a decision is laid out: the number of options, whether a program is opt-in or opt-out, how alternatives are described and where they sit. A cafeteria that puts healthy food at eye level is a standard example. The term names a tool, not a purpose: the Federal Trade Commission lists defaults set to benefit the company rather than the user among dark patterns.

Thaler and Sunstein argued that every arrangement of options pushes in some direction, so the designer cannot avoid deciding on one. On that view, even food set out at random still sways what diners pick. That leaves a question philosophers still raise: whether a deliberate layout respects autonomy less than an accidental one with exactly the same effect.

Thaler and Sunstein's definition separates a nudge from rules that forbid options and from taxes or payments that change the stakes. Defaults, reminders and eye-level placement of healthy food are standard examples. The edges are blurry: some writers set nudges apart from merely informing people, while Sunstein counts warnings and disclosures as nudges too.

Sunstein and Thaler called it libertarian because no option is removed or made harder, and paternalistic because it aims at the chooser's good as the chooser sees it. One philosophical analysis notes that nothing in the view interferes with liberty, so benevolence would describe it as well as paternalism. Others object to steering people by exploiting the biases the view cites.

Defaults stick partly because people tend not to act to change the status quo. When a state enrolled child care providers in an informational text-message program by default, fewer than 3% of those reached opted out. The same stickiness can serve a seller: the Federal Trade Commission lists a default that is good for the company but not the user among dark patterns.

Under opt-in, people join only if they act; under opt-out, they are enrolled unless they act. The Labor Department says automatic enrollment raises retirement plan participation. A widely cited comparison found opt-out countries had about six times higher organ donor registration, though a higher registration rate does not guarantee more actual donations.

In a US automatic-enrollment 401(k), a set percentage leaves each paycheck and goes into a default investment unless the worker acts, and the Labor Department says the design raises participation. Some plans must raise the rate by 1 point a year for several years, and some let new enrollees withdraw within 30 to 90 days. The rate and the investment are choices the plan makes for anyone who never chooses.

At one military base, a cover sheet that both explained the federal savings plan and prompted an active choice was followed by twice as many enrollments as at comparison bases, so the prompt's own share is unclear. A Treasury pilot that encouraged new hires to choose a higher contribution found no detectable effect. Neither test used random assignment, and one critique of nudging holds that active choice is needed to form one's preferences.

When Virginia let rental-aid applicants in qualifying zip codes attest to their income instead of submitting income documents, approvals rose by more than 10 percentage points and processing sped up; Kentucky saw about 10 points, with no detected rise in fraud. Yet in Virginia new applications did not rise significantly, so easing one step did not by itself draw in more applicants.

One gambling-research taxonomy defines sludge as harmful friction, such as obstacles to withdrawing money from a gambling site. Federal Trade Commission staff called an alleged six-to-nine-screen cancellation path a prime example of what one workshop panelist described as friction that tires people until they give up. Usage varies, though: another framework counts an 'Are you sure?' warning before an impulse buy as beneficial sludge.

One English study checked family doctors' practices for sludge by reviewing their websites and phoning 85 of them as mystery shoppers: 54% refused to register someone without photo ID or proof of address, against national guidance. US federal evaluators sort such burdens into learning costs, like working out eligibility, and compliance costs, like applying. Some checks do serve a purpose, such as confirming eligibility before payments go out.

Feedback turns hidden outcomes, such as a household's energy use or a clinic's vaccination rate, into numbers people can see. A review of energy field experiments found norm-based messages mostly produce small reductions. Adding a bar chart to a text letter did not change English general practices' antibiotic prescribing, and star-rated immunization report cards sent to school leaders did not raise compliance.

Letters telling high-volume Medicare prescribers of the antipsychotic quetiapine that their prescribing was extremely high next to in-state peers, and under review, cut prescribing with no detected harm, though the comparison came with the review notice. In England, norm letters on broad-spectrum antibiotics had no detectable effect, although earlier letters to high prescribers had cut antibiotic use.

In a megastudy with millions of pharmacy patients, text reminders raised 30-day COVID-19 booster uptake by about 1 percentage point, a 21% rise. The best-performing federal student aid emails lifted renewal applications from about 4.3% to 7.8% within 20 days. But automated texts sent before passport appointments did not change the no-show rate, so a reminder is not guaranteed to work.

In a pharmacy megastudy, a text that urged patients to make a plan and suggested the same day, time and store as their last shot raised vaccination by about 1.2 percentage points. It was one of three messages that could not be ruled out as the top performer. Because it paired the plan with a suggested slot, it does not show that planning alone did the work.

Salience can also be used in reverse: the Federal Trade Commission describes highlighting a subtotal in a bright box while listing mandatory fees below in plain text. Seeing is not the same as acting, though. After a large fast-food chain put calories on its menus, customers in four New England cities were more likely to notice calorie information, yet calories purchased did not change compared with chains that did not label.

The Federal Trade Commission judges a website by the net impression of its whole design, not by whether accurate words appear somewhere; it has alleged fees hidden behind tooltips and below the fold. A federally required mortgage form, the Loan Estimate, instead uses a standard layout in clear language. And when an ad is built to look like news, the FTC says a disclaimer is unlikely to undo the misleading impression.

Lenders must use the same Loan Estimate form for most US mortgages, which the Consumer Financial Protection Bureau says makes loans easier to compare. A 2024 Federal Trade Commission rule requires the total price, fees included, up front for live-event tickets and short-term lodging, partly to ease comparison shopping. Standard forms leave gaps: lenders itemize origination charges differently, so the bureau says the total is what matters.

COVID-19 vaccine incentives show the range. Offering Swedes about 24 dollars for getting vaccinated raised vaccination, while a California Medicaid plan that randomized members to different incentive structures found no change. State lotteries that rewarded only a first dose lifted first doses in most states but much less so completed series, which suggests that what exactly is rewarded matters.

When 20 US states ran COVID-19 vaccine lotteries in 2021, the behavioral-economics rationale was that people give long-shot chances extra weight. Results were mixed: one analysis linked the lotteries to higher vaccination overall and in some states but not others, and another found first-dose gains in most states it could assess but null or negative effects on completed vaccination in over half.

A large meta-analysis of over 200 studies reported an average effect of d = 0.43, but a reanalysis correcting for publication bias found no clear evidence of an average effect. Trials run by two large US nudge units, where every trial was counted, averaged about 1.4% against 8.7% in academic papers. Small is not zero: the nudge-unit average was still significant, and effects vary widely.

Common objections hold that nudges often work without people noticing, that they exploit faulty reasoning rather than engage it, and that some cross into manipulation. Others argue that a focus on individual choice can pull attention away from structural causes. Not every nudge draws these objections: a cigarette warning is openly deliberate and removes no option.

Across 27 studies, information-only outreach raised health insurance enrollment by about 0.6 percentage points and hands-on enrollment help by about 2, and federal evaluators concluded that groups with low enrollment need other approaches to see a big change. Critics add that nudges can distract from structural causes. Yet in one megastudy, free rides to pharmacies added nothing beyond vaccine reminders.

Sunstein argues that choice architecture, especially when officials design it, should be transparent and open to public scrutiny. Openness need not erase the effect: in a small federal survey test, explaining why yes was preselected raised acceptance by about 15 points, while explaining a texting default did not detectably change opt-outs. Purpose matters too: a default can serve the chooser or, like an add-on preselected at checkout, the seller.

Federal Trade Commission consumer guidance warns that a free trial will probably turn into a charge if it is not canceled in time, and that a box already checked at sign-up may authorize further charges or data sharing. Agency staff also flag privacy settings that collect the most data by default. Because opting out takes more effort than going along, the starting setting often becomes the final one.

Designer Harry Brignull coined the term in 2010, and panelists at a Federal Trade Commission workshop noted that such designs often exploit cognitive biases or delay needed information. They are common: an automated scan of 11,000 shopping sites found 1,818 instances on 1,254 of them. The agency reports that the tactics are often combined and tend to have stronger effects together.

The name comes from insect traps. The Federal Trade Commission's matching category, roadblocks to cancellation, includes letting people sign up online but making them cancel another way, hiding the cancellation phone number, understaffing the line or adding a sales pitch. Researchers describe this gap between the effort to get in and the effort to get out as bordering on manipulation.

In one example, a shopper who picks a one-time purchase instead of a subscription is shown a decline option worded as not wanting to save money. The Federal Trade Commission lists the tactic among dark patterns identified by its workshop panelists and in academic research. It removes no option; it relies on shame, framing the other choice as a bad decision.

In one study presented at a Federal Trade Commission workshop, more than twice as many people accepted a free trial when the monthly charge appeared only in small gray print. A 2010 US law requires online sellers of such plans to disclose terms before taking billing details, get express consent and offer a simple way to stop charges. The law sets conditions rather than banning renewal, and agency guidance notes auto-renewal can be convenient.

Drip pricing shows only part of the price until late in checkout, after the buyer has invested time. One study discussed by the Federal Trade Commission found ticket buyers not shown fees upfront spent about 20% more and were 14% more likely to complete a purchase. A 2024 rule from the agency requires the full price, mandatory fees included, to be shown up front for live-event tickets and short-term lodging.

The Federal Trade Commission lists preselection as a dark pattern when a default is good for the company but not the user, as with trip insurance added to a purchase automatically or a tracking-cookie box already checked. In one case the agency alleged that a TV maker's vaguely named default setting let it collect and share viewing data. The tactic relies on people not noticing.

The Federal Trade Commission's list includes 'only 1 left' messages when stock is not actually low, false claims about how many others are viewing an item, and countdown clocks that vanish or reset when they run out. Each variant is defined by being false, so a genuinely limited supply is not the problem; a deadline that resets when reached is one sign the urgency is invented.

The Federal Trade Commission describes prompts that offer only 'Yes' or 'Not now', and location-data requests offering only accept or remind me later, which return until the person gives in, likely out of frustration. In some, there is no way to refuse data collection at all. Each prompt looks harmless alone; the pressure comes from repetition with no lasting exit.

In one case, the Federal Trade Commission alleged that a children's learning site made members click through six to nine screens of offers and links that led away from cancellation, despite promising easy cancellation. Agency staff say canceling should be at least as easy as signing up and possible through the same medium, so an online sign-up should not require a phone call to quit.

A 2010 law requires online subscription sellers to disclose terms, get express consent and offer a simple way to stop charges. A 2024 Federal Trade Commission rule requires total prices up front for live-event tickets and short-term lodging. The agency's broader 2024 click-to-cancel rule was later vacated, and in March 2026 it asked the public whether and how to amend its rule on negative option plans.

Behavioral finance adds psychology to financial theory instead of assuming that market participants always choose rationally. One studied bias is the disposition effect, selling winners too quickly and holding losers too long, shown in US stock trading and reported in many markets since. In a classic study of 78,000 brokerage households, frequent traders netted less than infrequent ones, a result read as consistent with overconfidence.

Jegadeesh and Titman found that buying past winners and selling past losers earned significant returns over 3 to 12 months. One explanation is underreaction: investors respond slowly to information that arrives gradually. Other work links momentum to time-varying risk exposure measured through betas, and some finds that momentum eventually reverses, with the timing tied to trading volume.

Without lasting biases, mispricing should be short-lived; researchers expect it to persist when the biases continue and it cannot be arbitraged away. Trading costs run higher in hard-to-arbitrage stocks, such as tiny, thinly traded or very volatile ones, and short-selling constraints can block the selling side of a trade. Persistence is not proof of error, though: another view treats such returns as pay for risk that pricing models miss.

A federal evaluation team works with agencies to test letters, emails and forms. Postcards to people eligible for the Earned Income Tax Credit raised tax filing by 0.6 to 1 percentage points and credit dollars paid, though not the share claiming it, and one email nearly doubled military savings-plan enrollment. Several of the team's tests, such as flu-shot emails to veterans, found no detectable effect.

Johnson and Goldstein reported about six times higher donor registration in opt-out countries, and other studies report gains in actual deceased donations. Yet a 2025 study of countries that switched found no significant rise in deceased donors and a 29% fall in living donors. Chile's 2010 opt-out law was followed by mass opting out and a 15-year low in donors, though donations were already falling.

In a trial with 228,000 Medicare beneficiaries, letters raised flu vaccination from 25.9% to 26.5%, and text reminders in a pharmacy megastudy lifted booster uptake by about 1 point. But emails to veterans in one Veterans Affairs system did not change flu shots, and feedback to care teams at another VA site did not raise vaccination. Correcting for publication bias, one reanalysis found evidence against an average effect of health nudges.

Comparison reports pair a household's own usage with the average and with efficient neighbors. A review of field experiments found that social-norm messages mostly produce small reductions, which it defined as under 5%. In a federal trial in New York City public housing, where the government pays the utility costs, four such letters produced no significant change in energy use.

A large project repeating about 100 psychology studies replicated roughly 40%; the Many Labs projects replicated 29 of 51 effects; and a test of 21 social science studies from Nature and Science replicated about 60%. None of these samples was drawn at random from the literature, so none gives a single replication rate for the field.

An early meta-analysis of published studies found a moderate effect, d = 0.62. Later analyses that allowed for publication bias suggested a much smaller effect or none, and preregistered multilab tests found d = 0.04 and 0.06, not significant, or a significant but weak 0.10. The effect is contested rather than settled in either direction.

A 1996 study reported that people exposed to an old-age stereotype walked more slowly afterward. A 2012 replication with automated timing found no effect; in a follow-up, slower walking appeared only when experimenters expected it, and the authors judged that priming, if real, works differently than assumed. Nine experiments testing whether words like 'professor' improve quiz answers also failed.

The original study reported higher testosterone, lower cortisol and more risk-taking after expansive poses. One p-curve analysis found no evidential value; another, by Cuddy and colleagues, found it for feelings of power but not for hormones or risk-taking. A 2019 study of 82 men found repeated posing did not change hormone levels, and a 2024 analysis argued the p-curve method itself is unreliable.

When journals and authors favor striking results, the published record becomes a biased sample of the studies actually run. A reanalysis of the nudge literature found severe bias and, after correcting for it, no clear average effect; nudge-unit trials, all of which were counted, averaged far smaller effects than published papers. Corrected ego depletion estimates also shrank, which researchers tie to this bias and possibly p-hacking.

Among large heart-disease trials funded by a federal institute, 57% published before 2000 reported a benefit against 8% after, a shift strongly associated with registering outcomes in advance. In 2024 sports science papers, preregistered studies supported their hypotheses less often, 45% versus 56%, a significant gap of negligible size, and most preregistered papers with a hypothesis had not registered it or reported a different one.

Ioannidis argued from a statistical model that findings are less likely to be true when studies are small, effects are small, and designs and analyses are flexible. Replication data agree in part: among 176 student replications, larger original effects and within-participant designs replicated more often. By these measures, a single small study with a flexible analysis is a weak basis for confidence.

Many dark patterns work by manufacturing urgency, like fake countdown timers, or by revealing costs only after a buyer feels committed at checkout. Federal Trade Commission consumer guidance stresses finding a trial's terms, length and cancellation steps before signing up, and calls terms that cannot be found a warning sign. It also advises watching for pre-checked boxes and marking the trial's end date.