Behavioral Finance: Six Biases, and What Each One Actually Costs

Almost every article on behavioral finance lists the same biases and attaches no number to any of them. That is the problem with the genre: a bias you cannot measure is indistinguishable from a personality trait, and there is nothing to do about it on Monday morning. The research does contain numbers — measured on real brokerage accounts, over defined periods, with named samples — and they are more interesting than the lists, partly because the largest of them is not caused by what the lists say causes it, and partly because they cannot be added together, which is exactly what the genre does next.

This article takes six biases that have been measured on real money, states what each one cost, and is explicit about what the number does and does not cover. Five of the six come from the same body of work on US discount-brokerage records; the sixth comes from fund flows. None of them is a law of nature, and all of them are averages across populations in which the median investor did something quite ordinary.

The six biases with a number attached

Each row is a separate study on a separate sample. The right-hand column is the comparison the authors actually ran, which is not always the comparison the number gets used for.

Notice that the last row is a cost in risk, not in return. Holding four stocks is not a mistake that shows up as a lower average return — it shows up as a much wider range of outcomes, most visibly when the employer is the issuer. Enron’s employees had 62% of their retirement plan assets in company stock at the end of 2000.

The biggest number is not caused by bad stock picking

The 7.1-point gap between the most and least active households is the largest measured cost in the table, and the usual telling of it is that frequent traders are bad at picking stocks. The paper says something more specific and more useful.

Before costs, the households in that sample did roughly as well as the market: the average household earned a gross return of 18.7% a year against 17.9% for a value-weighted index of NYSE, AMEX and Nasdaq stocks. After the bid-ask spread and commissions, the average household earned 16.4%. Among the most active households — turning over more than 250% of their portfolio a year — the net figure falls to 11.4%. In the authors’ words, gross performance differs very little between frequent and infrequent traders.

So the mechanism is not that overconfident investors choose worse stocks. It is that they choose roughly the same stocks, more often, and pay for the privilege each time. That matters because it changes the remedy: the fix is not better analysis, it is fewer transactions. It is also why this particular number should not be quoted at investors paying zero commission on a modern platform without adjusting for the fact that half of the friction the study measured has since been competed away — the spread and the tax have not.

The other side of that coin is Odean’s 1999 result on a different sample, where selection was poor: the stocks bought underperformed the stocks sold by 3.2 points over the next year. Two datasets, two different dominant mechanisms, and anyone quoting one of them as “the” cost of overconfidence is picking a favourite.

Loss aversion has a measurable footprint: the disposition effect

Loss aversion is the bias everyone can name and nobody can price, because feeling a loss twice as keenly as a gain is not itself a cost. What it produces is: investors realise gains and postpone losses, which is the disposition effect.

Odean measured what happened next. Over the following 252 trading days, the winners investors sold beat the losers they kept by 3.4 percentage points. Over 84 days the gap was 1.0 point; over 504 days, 3.6. The investors were not merely unlucky in their timing — they were systematically selling the holding that went on to do better, in the belief that the loser was due a rebound.

There is a tax wrinkle worth stating plainly, because it is the honest objection: in a taxable account, realising gains and deferring losses is backwards on tax grounds too, so the behaviour costs twice. In a tax-sheltered account the 3.4 points stand alone.

The gap between what the fund earned and what its investors earned

The fifth row is the only one not measured on individual brokerage records. Morningstar’s Mind the Gap compares a fund’s total return with the return of the average dollar invested in it — the two differ when money arrives after good runs and leaves after bad ones. For the decade to 31 December 2025, across roughly 23,000 US funds and ETFs, the average dollar earned 8.7% a year while the funds themselves returned 9.9%: a gap of 1.2 percentage points.

Two caveats belong with that figure, and most articles carry neither. First, the method is contested: a 2026 paper in the Financial Analysts Journal argues the calculation is sensitive to how cash flows are weighted and that it overstates the cost of bad timing. Second, the gap is not evenly spread — it is widest in the most volatile, narrowest categories and smallest in plain allocation funds, which is a statement about which products invite mistiming rather than about investors in general.

It is also the smallest number in the table, which is awkward for the industry that quotes it most — usually inflated to three to five points by borrowing from a different study, Dalbar’s Quantitative Analysis of Investor Behavior, which compares investor returns against an index rather than against the funds those investors actually held. That comparison folds in fees, bonds, cash and international exposure. None of those are timing mistakes.

Why you cannot add these numbers up

Here is the arithmetic that ends the genre. Sum the measured costs in the table — 7.1 + 3.2 + 3.4 + 2.65 + 1.2 — and you get 17.55 percentage points a year. Subtract that from a long-run 8% and the investor earns −9.55% a year: $10,000 becomes $1,343 over twenty years, while the untouched benchmark grows to $46,610. Nobody believes that, and yet “behavioral mistakes cost investors X% a year” is built by exactly this addition.

Three reasons the sum is meaningless:

• The samples are different people. Sixty-six thousand US brokerage households in the 1990s, ten thousand accounts in the late 1980s, and everyone who owned a US fund in the 2010s and 2020s are not the same population, and the studies run over different decades with different costs and different markets.
• Several rows measure the same underlying behaviour. Overtrading, the gender partition, and buying what caught your attention are three views of one thing. The gender comparison is not a finding about men and women; it is a way of testing the overconfidence hypothesis by splitting the sample on a variable that predicts overconfidence.
• They are not all costs of the same kind. Two of them are gaps between one portfolio and another, one is a gap between a fund and its own investors, and the last is not a return at all but a concentration of risk.

What you can legitimately do is take one gap and let it compound. The 1.2-point timing gap, applied to an 8% base over twenty years, costs about 20% of the final pot — $9,334 out of $46,610 on a $10,000 start. The 3.4-point disposition gap costs 47% of it. Those are illustrations of a single measured effect compounding, not a forecast of your account.

What the numbers imply, in order of size

Ranking by measured cost rather than by how often a bias gets mentioned changes the advice:

1. Cut the number of transactions before anything else. The largest measured gap comes from frequency, not from judgement. A rule that limits when you may trade — a rebalancing schedule or a threshold — removes most of the opportunities to express the bias.
2. Judge a holding on what you would pay for it today. The disposition effect lives entirely in the purchase price, which is information about your past, not about the asset. If you would not buy it now at this price, the fact that you are down on it is not a reason to keep it.
3. Automate the timing decision. The fund-to-investor gap is created by discretionary arrival and exit. Removing the discretion — fixed contributions, fixed dates — removes the gap by construction; our piece on lump sum vs dollar-cost averaging covers what that trade-off costs on average.
4. Count your positions. Four holdings is not a portfolio, and if one of them is your employer you have concentrated your salary and your savings in the same balance sheet.
5. Size positions so that being wrong is survivable. Every bias above becomes terminal at the wrong position size and merely annoying at the right one — the arithmetic is in our guide to position sizing and risk of ruin.
6. Write the rules down before the market moves. That is the entire practical content of behavioral finance, and it is also what our guides to investor mindset and common investing mistakes are for. If you do not know which kind of investor you are, the risk profile questionnaire is a starting point.

Frequently asked questions

Which behavioral bias costs investors the most?
On the measured evidence, overconfidence expressed as trading frequency. In the largest study of US brokerage households, those who traded most earned 11.4% a year net against 18.5% for those who traded least — a gap of 7.1 percentage points over 1991–1997. The caveat is that the gap was created almost entirely by transaction costs, not by worse stock selection.

Do these studies still apply now that commissions are zero?
Partly, and this is the honest limitation of the table. Commissions on US equities have largely been competed away since 2019, and they were a large share of what the 1990s studies measured, so the 7.1-point gap should not be quoted as a current figure. What has not gone anywhere: the bid-ask spread, the tax bill triggered by realising a gain, and the selection and timing effects, which were measured as return differences between portfolios rather than as fees.

What is the disposition effect in plain terms?
The tendency to sell what is up and hold what is down. Measured on US brokerage records, the winners that investors sold went on to beat the losers they kept by 3.4 percentage points over the following year, so the behaviour is not merely tax-inefficient — it is the wrong way round on the merits as well.

Is the “investors underperform their own funds by 3 to 5 points” claim true?
Not as stated. Morningstar’s like-for-like comparison, which measures the average dollar in a fund against that same fund, puts the gap at 1.2 percentage points for the decade to December 2025. The larger figures come from comparing investor returns to an index, which folds in fees, bonds, cash and international exposure — none of which are timing mistakes.

Do men really invest worse than women?
That is not what the study tested. Barber and Odean partitioned a sample on gender because psychological research finds men are more overconfident in financial matters, and used it to test whether overconfidence predicts excessive trading. It did: men traded 45% more, and trading reduced their net returns by 2.65 points a year against 1.72 for women. The finding is about trading frequency, and it says nothing about any individual.

Can you remove these biases with enough self-awareness?
The evidence is not encouraging. The clearest test of learning comes from Brazilian day traders: average daily net profit was −US$47.34 over an investor’s first 250 day trades and −US$51.65 over the ones after that, so practice made results slightly worse rather than better. What does work is removing the decision. Rules, schedules and automation reduce the number of moments at which a bias can be expressed, which is why the largest measured cost in the table — trading frequency — is also the most tractable one.

Sources

• Brad M. Barber and Terrance Odean, “Trading Is Hazardous to Your Wealth: The Common Stock Investment Performance of Individual Investors”, The Journal of Finance 55(2), April 2000. 66,465 households at a large US discount broker, February 1991 to January 1997. Gross return of the average household 18.7% vs 17.9% for a value-weighted NYSE/AMEX/Nasdaq index; net 16.4%; most-active households 11.4% net vs 18.5% for the least active; average household held 4.3 stocks worth $47,334 and turned over 75% of the portfolio a year.
• Terrance Odean, “Do Investors Trade Too Much?”, American Economic Review 89(5), December 1999, pp. 1279–1298. 10,000 accounts at a nationwide discount broker, January 1987 to December 1993, a file of 162,948 trade records. Over the following 12 months, market-adjusted returns to purchases were 3.2 points below those to sales (3.3 points on raw returns).
• Terrance Odean, “Are Investors Reluctant to Realize Their Losses?”, The Journal of Finance 53(5), October 1998. Excess return of winners sold minus paper losses kept: 1.0 point over 84 trading days, 3.4 over 252, 3.6 over 504 (p-values 0.002, 0.001, 0.014).
• Brad M. Barber and Terrance Odean, “Boys Will Be Boys: Gender, Overconfidence, and Common Stock Investment”, The Quarterly Journal of Economics 116(1), February 2001. More than 35,000 households, February 1991 to January 1997. Men traded 45% more than women; trading reduced men’s net returns by 2.65 percentage points a year and women’s by 1.72.
• Morningstar, Mind the Gap 2026. Roughly 23,000 US open-end funds and ETFs in existence since 1 January 2016; for the decade to 31 December 2025 the average dollar earned 8.7% a year against 9.9% for the funds themselves. The methodology is disputed in “Bad Timing Does Not Cost Investors 15% of Their Funds’ Returns”, Financial Analysts Journal, 2026.
• Barber and Odean, “The Behavior of Individual Investors”, in Handbook of the Economics of Finance (2011), section 6, for the diversification figures: Poterba (2003) on 44% of assets in company stock across the 20 largest US corporate defined-contribution plans, and the Enron holding of 62% of plan assets at the end of 2000.
• Fernando Chague, Rodrigo De-Losso and Bruno Giovannetti, “Day trading for a living?”, Department of Economics FEA-USP, Working Paper 2019-47, for the learning test: average daily net profit of −US$47.34 over an investor’s first 250 day trades against −US$51.65 thereafter, among 19,646 Brazilians who began day trading mini-Ibovespa futures between 2013 and 2015.
• Own arithmetic: compounding each measured gap against an 8% annual base on a $10,000 starting sum over 10, 20 and 30 years, and the sum of the five return gaps (17.55 points) applied to the same base.

This article is educational and is not investment advice. We are not licensed advisers, and nothing here is a recommendation to buy, sell or hold any security. Figures quoted are from the studies named above, measured on the samples and periods stated; they are averages across populations and are not forecasts of any individual result.