The psychology of money: why smart people make predictable financial mistakes

In 1998, the financial economist Terrance Odean published a study of 10,000 individual brokerage accounts spanning seven years of trading activity. The finding was specific and, once you absorbed it, disturbing. Investors were roughly one and a half to two times more likely to sell a stock that had gone up than one that had gone down. They were, in aggregate and with remarkable consistency, selling their winners and clinging to their losers — a pattern that cost them measurably in after-tax returns. Not occasionally. Not in bursts of panic. Systematically, across thousands of accounts, over years.

The study was not obscure. It was published in the Journal of Finance. And the broader phenomenon it documented — loss aversion distorting investment decisions — had been described theoretically two decades earlier by Daniel Kahneman and Amos Tversky, in a paper that would eventually help earn Kahneman a Nobel Prize. By the time Odean’s data confirmed it in real brokerage accounts, the intellectual framework was already becoming famous. In 2002, Kahneman won the Nobel in Economics. In 2011, Kahneman published Thinking, Fast and Slow — a book that would sell millions of copies in the years that followed. In 2017, Richard Thaler won his own Nobel for work on the same territory. Loss aversion, the disposition effect, cognitive bias — these became dinner-table concepts for the educated public.

And none of it made any difference.

The aggregate pattern Odean documented persisted. Investors kept selling winners and holding losers. The knowledge spread. The behaviour didn’t change. That gap — between what we know about our financial cognition and what that knowledge actually does to our financial cognition — is not a curiosity. It is a diagnostic. It tells us something specific about what kind of problem financial failure actually is, and why almost everything we do to address it is aimed at the wrong level.

The architecture

The reason knowledge doesn’t help starts with the mechanism itself, and the mechanism is stranger than the pop version suggests.

Loss aversion, as formulated in Kahneman and Tversky’s prospect theory (1979), is not simply the observation that people dislike losses. It is a precise, asymmetric feature of how the human brain assigns value to outcomes relative to a reference point. The value function is steeper for losses than for gains — empirically, by a factor of roughly two to 2.25. A loss of $100 produces approximately twice the psychological pain that a gain of $100 produces pleasure. The two experiences are not the same event in different directions. They are processed differently, weighted differently, felt differently.

Offer someone a coin flip: heads, you win $150; tails, you lose $100. The expected value of that bet is positive — take it a hundred times and you’ll come out ahead. Most people refuse it. Not because they can’t do the arithmetic. Because the architecture processes the potential loss of $100 as psychologically equivalent to a potential gain of roughly $200, which means the $150 upside isn’t enough. The math says yes. The machinery says no. The machinery wins.

Hersh Shefrin and Meir Statman gave this pattern a name in 1985: the disposition effect. In investment decisions, selling a winning position feels good — you’re locking in a gain, and gains are pleasant. Holding a losing position avoids the alternative, which is selling at a loss, and losses are roughly twice as painful as equivalent gains are pleasant. The result is predictable and was predicted: investors sell their winners too early and hold their losers too long, not because they’ve analysed the fundamentals and concluded the losers will recover, but because selling the losers would mean experiencing a loss, and the architecture is built to avoid that experience at almost any cost.

Odean’s data confirmed this at scale. Ten thousand accounts, seven years, a clean and consistent aggregate pattern. But what makes the finding genuinely troubling — and what the popular account usually leaves out — is what happened when researchers looked at professional traders. Coval and Shumway, in a 2005 study of professional proprietary traders on the Chicago Board of Trade, found that these traders too displayed loss-averse behaviour — assuming above-average afternoon risk to recover from morning losses. Professional training and professional incentives attenuated the pattern. They did not eliminate it. If financial expertise could override the architecture, professionals operating under market discipline would be the population where you’d see it disappear. It didn’t disappear. It got quieter.

The disposition effect is not a mistake people make because they haven’t read enough. It is a structural feature of how the human brain processes value. And the distinction matters, because it determines what kind of intervention could possibly work. If the problem is ignorance, education fixes it. If the problem is architecture, education is irrelevant — which is exactly what the data shows, and exactly what we refuse to accept.

A note on the replication crisis

The scientifically literate reader will want to know whether the findings in this article are casualties of the replication crisis that has battered behavioural economics since the early 2010s. A fair question. A non-trivial portion of the social priming effects cited in Kahneman's own popular work — Thinking, Fast and Slow — have failed to replicate, and Kahneman himself acknowledged this publicly in 2012 in an open letter to priming researchers. But loss aversion, prospect theory, and the disposition effect are not priming effects. They belong to a different empirical tradition — one grounded in actual market data (Odean's brokerage records), controlled economic experiments (Kahneman, Knetsch, and Thaler's endowment studies), and real housing market records (Genesove and Mayer's condominium data). These findings have been replicated across countries, asset classes, and investor populations in the decades since their original publication. The replication crisis is real. The empirical tradition this article draws on is not its primary casualty.

The time problem

Loss aversion explains why we cling to bad investments and flee from good ones. But the architecture fails in a second, distinct way — one that comes from a different body of research entirely, and that governs a different category of financial decision.

Hyperbolic discounting sounds like jargon, and it is, but the phenomenon it describes is genuinely strange. Not “people prefer immediate rewards” — that’s common sense and has been common sense for millennia. The strange part is the inconsistency. Give someone two options: $900 in 30 days, or $1,000 in 31 days. Most people pick the $1,000. The wait is trivially different; the extra hundred is worth it. Now move both options forward so they start today: $900 right now, or $1,000 tomorrow. Same person, same dollar amounts, same one-day gap. And most people flip. They take the $900 now.

Richard Thaler first documented this inconsistency in monetary choices in 1981. David Laibson formalised it in 1997 with what became known as the quasi-hyperbolic discount function: the discount rate applied between the present moment and any future period is disproportionately large compared to the discount rate between any two future periods. Put plainly, the present is special. Not because people are greedy or short-sighted in some moral sense, but because the cognitive architecture applies a different rule to now than it applies to later.

The financial consequence is saving — or rather, not saving. Saving is a transaction between your present self and your future self. You give up money now so that a person you will become later has more of it. And the architecture systematically undervalues that future person. Not slightly. Not by a marginal percentage that careful reflection could correct. The discount is steep and it is automatic, which means the decision to save always feels, at the moment of decision, like a bad trade. The present self is being asked to sacrifice for an agent it is neurologically incapable of adequately valuing.

This is not a willpower problem. It is not a character failing. It is the architecture doing exactly what it does.

The most telling evidence for this comes from the researchers themselves. When Thaler and Shlomo Benartzi designed the Save More Tomorrow programme in 2004 — published in the Journal of Political Economy — they did not try to educate workers about hyperbolic discounting. They did not explain the quasi-hyperbolic discount function and hope for the best. They restructured the decision. Workers committed in advance to allocating a portion of future pay rises — money they didn’t yet have, money that belonged to the future self — to retirement savings. The commitment was made before the money became present-self income. Savings rates among participants rose from 3.5 percent to 13.6 percent over four successive pay raises.

No seminar achieved that. No curriculum. No module on the time value of money. The architects of the research understood what the architecture actually was, and they didn’t trust education to reach it. They rebuilt the decision environment instead.

The ownership distortion

Loss aversion doesn’t only distort how people trade stocks. It distorts how people value anything they own — and when you scale that distortion from individual psychology to an entire asset market, the consequences stop being personal finance stories and become macroeconomic ones.

The endowment effect, a term Richard Thaler coined in 1980, is the observation that people demand significantly more to give up something they possess than they would pay to acquire the same thing. Kahneman, Jack Knetsch, and Thaler tested this experimentally in 1990 by giving half the participants in a study a coffee mug and then opening a market. Sellers demanded roughly $5.25 on average. Buyers offered roughly $2.25 to $2.75. The willingness-to-accept was approximately double the willingness-to-pay — for the same mug. Not a family heirloom. Not a house. A mug they’d been handed minutes earlier.

The mechanism is loss aversion applied to ownership. Once you own something, giving it up is a loss. Acquiring something you don’t own is a gain. And losses weigh roughly twice as much as gains. The result: the psychological cost of selling is systematically higher than the psychological benefit of buying, even when the object is identical and the person is the same.

Coffee mugs don’t crash economies. Houses do.

In 2001, Dean Genesove and Christopher Mayer published a study in the Quarterly Journal of Economics using data from Boston’s downtown condominium market during the 1990s downturn. They found that homeowners facing nominal losses — whose property’s market value had fallen below their original purchase price — set asking prices 25 to 35 percent above the gap between the expected selling price and their purchase price. They achieved 3 to 18 percent higher realised prices than sellers not facing nominal losses. But they paid for it with dramatically lower sale rates. They sat on the market for months. Many didn’t sell at all.

The aggregate effect is the one that matters for this argument. When a market declines, loss-averse sellers refuse to adjust their asking prices to match what the market will actually pay. Not because they’ve analysed the fundamentals and concluded their property is undervalued. Because selling below the purchase price means realising a loss, and the architecture treats that loss as roughly twice as painful as the equivalent gain. Volume collapses. The market doesn’t clear. It seizes.

These sellers aren’t stupid. They aren’t greedy. They are operating from a value function in which the original purchase price is more psychologically real than the current market price — because the architecture makes it so.

You can explain the endowment effect to a homeowner sitting on a property worth $50,000 less than they paid for it. They will nod. They will understand the mechanism. They may even recognise it in their own behaviour. And they will still set their asking price too high, because the mechanism that produces the behaviour is not located at the level of understanding. It never was.

Mental accounting and the weight of reference points

There's a related phenomenon that helps explain why the purchase price feels so heavy. Richard Thaler called it mental accounting — the tendency to treat money differently depending on where it came from or which mental category it occupies. A $5,000 tax refund gets spent differently from $5,000 of salary, even though the dollars are identical. Casino winnings — "house money" — get gambled more freely than earned savings. And the price you paid for an asset gets locked in as the reference point against which all future prices are judged, not because it reflects any economic reality but because the brain files it in a category that carries psychological weight. This is why the Genesove and Mayer findings feel inevitable once you understand the mechanism: the purchase price isn't just a number. It is the anchor of an entire mental account, and selling below it means closing that account at a loss — an act the architecture is built to resist.

The optimism trap

If the story so far were just about loss aversion — an architecture too sensitive to loss, too quick to protect what it has, too reluctant to let go — it would be troubling enough. But the architecture fails from the other direction too.

The planning fallacy, first named by Kahneman and Tversky in 1979 and empirically demonstrated by Roger Buehler, Dale Griffin, and Michael Ross in 1994, is the systematic tendency to underestimate the cost, duration, and difficulty of projects we plan to undertake. Buehler and colleagues showed that participants consistently predicted task completion times more optimistically than their own prior experience with identical tasks would justify. Not other people’s experience. Their own. They had done similar things before, they knew how long those things had taken, and they predicted faster anyway.

The mechanism is what Kahneman later called the inside view versus the outside view. When you plan a project — renovating a kitchen, launching a business, saving for retirement — you instinctively construct the plan from the inside: you imagine the specific sequence of steps, assume they’ll proceed more or less as intended, and arrive at a time and cost estimate that reflects your plan rather than reality. The outside view — what actually happened to other people who renovated kitchens, launched businesses, or projected their retirement savings — is available, but it feels less relevant than your specific plan. So you ignore it.

The financial consequences are concrete. According to the Bureau of Labor Statistics, roughly 20 percent of new businesses fail within their first year. About half are gone within five years. Business plans, almost universally, do not project this. The Employee Benefit Research Institute’s Retirement Confidence Survey has documented the gap between expectation and reality for decades: only 12 percent of current workers plan to retire before age 60, but 27 percent of actual retirees report that they did. Workers plan for the career path they imagine. The actuarial record delivers something else.

Daniel Lovallo and Kahneman extended the mechanism in a 2003 Harvard Business Review article, showing how it produces cost overruns and benefit shortfalls in business contexts — not because executives are dishonest in their projections, but because the inside view is the default cognitive mode and the outside view requires deliberate, effortful override that rarely happens.

But here is what makes the planning fallacy genuinely unsettling in the context of everything that precedes it. Loss aversion, the endowment effect, hyperbolic discounting — these are all failures of excessive caution, fear of loss, reluctance to let go of what you have. The planning fallacy runs in the opposite direction. It is a failure of excessive optimism. The architecture isn’t just tilted one way. It fails from both sides — too afraid of loss when facing the past, too confident of success when facing the future. And the combination is worse than either alone, because it means there is no stable vantage point from which the architecture gives reliable financial guidance. The whole system is unreliable. Not in one direction. Comprehensively.

Naming the planning fallacy does not interrupt it. You can read this paragraph, understand the mechanism, and still underestimate the cost of your next home renovation. Because the fallacy operates at the moment you believe you are making an accurate assessment. That’s the trap.

The wrong cure

The institutional answer, for the better part of fifty years, has been financial education. Teach people the maths. Explain compound interest. Run workshops on budgeting. Integrate personal finance modules into school curricula. The logic seems sound: people make bad financial decisions because they don’t know enough, so give them more knowledge.

In 2014, Daniel Fernandes, John Lynch, and Richard Netemeyer published a meta-analysis in Management Science that tested this logic at scale. They analysed 168 papers covering 201 studies of financial literacy interventions. The finding: financial literacy programmes explain 0.1 percent of the variance in downstream financial behaviour.

Not 10 percent. Not 5 percent. Not 1 percent. Zero point one percent. And even that fraction decayed. Beyond twenty months, the effects of financial education on actual behaviour approached zero. Large interventions with many hours of instruction had negligible impact once enough time had passed. The relationship between financial education and financial behaviour is, at population scale, statistically indistinguishable from noise.

The counter-evidence deserves honest engagement. Kaiser and Menkhoff, in a 2017 meta-analysis of 126 impact evaluation studies published through the World Bank, found effect sizes three to five times larger than the Fernandes estimate. That is a genuine methodological disagreement, and waving it away would be dishonest. But even taking Kaiser and Menkhoff’s most generous figures, multiplying Fernandes’ 0.1 percent by five gives you 0.5 percent of the variance in behaviour. At population scale, this is operationally negligible. If the strongest case for financial education explains half a percent of the difference in how people actually handle money, the intervention is not working. It is generating a measurable signal that is too small to matter.

And the counter-evidence, on closer inspection, supports the architectural argument rather than undermining it. Kaiser and Menkhoff’s most favourable results cluster around what they call “just-in-time” interventions — financial counselling delivered at the moment of a specific decision. But just-in-time interventions work precisely the way the SMarT programme worked: by changing the decision environment at the critical moment, not by improving the knowledge base months or years in advance. That the most effective “educational” interventions turn out to be structural ones is not a refutation of the architectural argument. It is a confirmation.

So the evidence says financial education doesn’t work at the relevant level. And yet. The United States mandates financial literacy education in a majority of states. The OECD runs an International Network on Financial Education with close to 300 public institutions from over 130 countries. Employers fund financial wellness programmes. Banks sponsor literacy curricula. The infrastructure is vast, expanding, and, on the evidence, essentially useless for its stated purpose.

Why does it persist?

The inference is not complicated. Financial education locates the problem in individuals — in their knowledge deficits, their mathematical gaps, their failure to understand compound interest. And locating the problem in individuals is extraordinarily convenient for everyone who is not the individual. It requires no changes to how financial products are designed, marketed, or regulated. It generates demand for curricula and programmes, which generates institutional funding and visible activity. It produces certifications, assessments, and completion rates that can be reported as progress. And it absolves the designers of the financial environment — the institutions that structure the choices people actually face — of any obligation to redesign the environment itself.

When the evidence base for your intervention shows it explains between 0.1 and 0.5 percent of the variance in the outcome you claim to address, and you continue to expand that intervention, the explanation is not that you need more evidence. The explanation is that the intervention serves a purpose other than the one on the label.

The researchers who understood the architecture best knew this. Thaler and Benartzi didn’t design a curriculum. They designed a commitment structure. Madrian and Shea, in a 2001 study published in the Quarterly Journal of Economics, found that switching a large company’s 401(k) plan from opt-in to automatic enrolment — no seminar, no module, no worksheet — raised new-hire participation by approximately fifty percentage points. Fifty. Not by teaching anyone anything. By changing the default.

Teaching people a vocabulary for their cognitive architecture is a categorically different operation from reducing the failures that architecture produces. The evidence is unambiguous about which of the two financial education has achieved.

Return to those 10,000 investors in Odean’s data. Seven years of selling winners and holding losers. A pattern that cost them real money, that was predictable in advance from a theory published in 1979, and that has persisted through three decades in which that theory became one of the most widely known ideas in social science. These were not unintelligent people. They were not uneducated. The research existed. In the years since, it won Nobel Prizes. It sold millions of books. It entered the vocabulary of educated conversation.

The pattern didn’t change.

What changed was the story we told about why. The institutional narrative held — still holds — that the problem is information. That if people understood their biases, they would correct them. That the gap between knowing and doing is a gap education can close. The evidence says otherwise. The gap between knowing and doing, for these mechanisms, is the architecture itself. It is not a gap. It is the system working as designed — designed, that is, by evolution for an environment that did not include stock markets, 401(k) plans, thirty-year mortgages, or the financial instruments whose complexity now exceeds the cognitive machinery that must navigate them.

We have known this for decades. And the institutional response, for decades, has been to keep teaching the maths. Not because the evidence supports it. Because the alternative — redesigning the decision environments that produce the failures — would require changing the interests that benefit from leaving them as they are.

The problem was never what people know. It was always what the architecture does with what they know. And the honest answer to what we’ve done about that, across fifty years of research and institutional effort, is: almost nothing.

Gen AI Disclaimer

Some contents of this page were generated and/or edited with the help of a Generative AI.

Media

Malik Cil – Pexels

Key Sources and References

Daniel Kahneman and Amos Tversky, “Prospect Theory: An Analysis of Decision under Risk,” Econometrica, vol. 47, no. 2, 1979, pp. 263–291.

Hersh Shefrin and Meir Statman, “The Disposition to Sell Winners Too Early and Ride Losers Too Long: Theory and Evidence,” Journal of Finance, vol. 40, no. 3, 1985, pp. 777–790.

Terrance Odean, “Are Investors Reluctant to Realize Their Losses?” Journal of Finance, vol. 53, no. 5, 1998, pp. 1775–1798.

Richard Thaler, “Toward a Positive Theory of Consumer Choice,” Journal of Economic Behavior and Organization, vol. 1, no. 1, 1980, pp. 39–60.

Richard Thaler, “Some Empirical Evidence on Dynamic Inconsistency,” Economics Letters, vol. 8, no. 3, 1981, pp. 201–207.

Daniel Kahneman, Jack Knetsch, and Richard Thaler, “Experimental Tests of the Endowment Effect and the Coase Theorem,” Journal of Political Economy, vol. 98, no. 6, 1990, pp. 1325–1348.

David Laibson, “Golden Eggs and Hyperbolic Discounting,” Quarterly Journal of Economics, vol. 112, no. 2, 1997, pp. 443–478.

Dean Genesove and Christopher Mayer, “Loss Aversion and Seller Behavior: Evidence from the Housing Market,” Quarterly Journal of Economics, vol. 116, no. 4, 2001, pp. 1233–1260.

Daniel Kahneman and Amos Tversky, “Intuitive Prediction: Biases and Corrective Procedures,” TIMS Studies in Management Science, vol. 12, 1979, pp. 313–327.

Roger Buehler, Dale Griffin, and Michael Ross, “Exploring the Planning Fallacy: Why People Underestimate Their Task Completion Times,” Journal of Personality and Social Psychology, vol. 67, no. 3, 1994, pp. 366–381.

Daniel Lovallo and Daniel Kahneman, “Delusions of Success: How Optimism Undermines Executives’ Decisions,” Harvard Business Review, July–August 2003.

Richard Thaler and Shlomo Benartzi, “Save More Tomorrow: Using Behavioral Economics to Increase Employee Saving,” Journal of Political Economy, vol. 112, no. S1, 2004, pp. S164–S187.

Daniel Fernandes, John Lynch, and Richard Netemeyer, “Financial Literacy, Financial Education, and Downstream Financial Behaviors,” Management Science, vol. 60, no. 8, 2014, pp. 1861–1883.

Tim Kaiser and Lukas Menkhoff, “Does Financial Education Impact Financial Literacy and Financial Behavior, and If So, When?” World Bank Economic Review, vol. 31, no. 3, 2017, pp. 611–630.

Brigitte Madrian and Dennis Shea, “The Power of Suggestion: Inertia in 401(k) Participation and Savings Behavior,” Quarterly Journal of Economics, vol. 116, no. 4, 2001, pp. 1149–1187.

Joshua Coval and Tyler Shumway, “Do Behavioral Biases Affect Prices?” Journal of Finance, vol. 60, no. 1, 2005, pp. 1–34.

U.S. Bureau of Labor Statistics, Business Employment Dynamics, Establishment Age and Survival Data.

Employee Benefit Research Institute and Greenwald Research, Retirement Confidence Survey (annual series, 2019–2025).

Daniel Kahneman, Thinking, Fast and Slow, Farrar, Straus and Giroux, 2011.

OECD International Network on Financial Education (INFE), oecd.org/en/networks/infe.html.

Daniel Kahneman, open letter to colleagues in social priming research, September 2012.