House money

What Thaler and Johnson actually found

Standard economics says a decision should depend only on the wealth it leaves you with. Richard Thaler and Eric Johnson showed in real-money experiments that it does not, and they found two effects pulling in opposite directions Manage Sci 1990.

• The house money effect. A prior gain made people accept gambles they had turned down from a clean start. The winnings sat in their own mental account, and losing them did not feel like losing your own money.
• The break-even effect. A prior loss made any gamble offering a route back to even unusually attractive, including gambles the same people had refused a moment earlier.

Both halves matter here. A section about a big win that only quotes the first half has taken the convenient one.

How well it has held up: less well than you have been told

This is the part almost every article on the subject leaves out.

In the laboratory it is conditional, not universal. A 2023 replication put 241 participants through four endowment conditions and found no significant effect of the first round's outcome on risk-taking in three of them. The authors concluded the effect is far less ubiquitous than assumed and appears to need skewed lotteries, many rounds, or both J Risk Uncertain 2023. What did survive their design was something slightly different: money that arrived unearned was risked more readily than money that was worked for, whether or not it had been won in the game.

In real casino data it can reverse outright. A study of 5,169 slot players at a Swiss casino over seven months found that after hitting a jackpot, players wagered significantly less, on the order of 60% less, and that the size of the jackpot made no further difference Econ Lett 2019. The authors call it a reverse house money effect. That is the opposite of what the popular version predicts.

So the honest position is this. The mechanism is real and well described. The prediction that a winner will automatically get reckless is not something the evidence supports, and we are not going to tell you it does.

The part that does hold: money is not fungible in your head

A dollar is a dollar to an economist. It is not a dollar to a person. Thaler's later framework describes how people sort money into separate accounts and apply different spending and risk rules to each one, with money arriving outside the normal income stream getting its own account and its own looser rules J Behav Decis Mak 1999.

Researchers who tested why winners take more risk put the mechanism in one sentence: the stake comes from prior gambling profits, and its potential loss has relatively low psychological value Judgm Decis Mak 2013. Gambling profits and ordinary income open different mental accounts, and the two get spent quite differently.

That is a claim about filing, not about recklessness, and it is the one worth carrying.

Cashed out or still on the screen

One study explains a good deal of the mess. Risk-taking after a loss depends on whether the loss has been realised: people who had cashed out a loss became cautious, while people looking at the identical loss still open on the screen took more risk Am Econ Rev 2016. The mental account has to close before the behaviour changes.

There is an obvious read-across to a prediction market, and it is worth stating as a read-across rather than a finding, because nobody has tested it here. Money still sitting in an exchange balance is an open account. Money in a bank is a closed one. That is the mechanism behind the plainest advice in this section, which is to move it.

What this does to a unit

Here is the arithmetic the effect collides with. If your unit is 1% of bankroll and a win adds 40% to the bankroll, the rule says the unit rises 40%. Mental accounting predicts it rises by more than that, because the new money is being sized against a different internal rule than the old money.

Nothing in the literature puts a number on that gap for bettors, and we are not going to invent one. What is measured, in real betting records rather than laboratory choices, is that stakes do rise after wins: about 3.4% per standard deviation of win size on slots and 13.5% to 16.7% on blackjack and roulette across 1,913,251 bets Sci Rep 2024; a mean stake of £18.47 after a win against £10.89 after a loss across 565,915 sports bets Cognition 2014; more bets placed and more numbers covered after a win at a roulette table J Risk Uncertain 2005.

That is the better-supported claim, and it is a different claim. Size goes up after a win. Whether it goes up because the money feels like the house's is a question the evidence has not closed. The practical answer does not depend on which explanation wins: decide the next bankroll as a figure, write it down before you reopen the account, and apply the percentage to that.

Two claims about winners that do not hold up

“Most big winners end up broke.” The figure normally quoted is folklore. The organisation it is universally credited to states plainly that the statistic is not backed by its research and cannot be confirmed by it, and traces it to an unverified remark at a discussion it hosted in 2001 NEFE 2018. The best-identified study of the question followed 34,987 Florida lottery winners and found large winners were about half as likely to file for bankruptcy in years one and two, equally more likely in years three to five, and no different over five years Rev Econ Stat 2011. The money moved the date, not the direction.

“Winning makes gamblers reckless, and the house money effect proves it.” Covered above. It does not prove it. Stakes do rise after ordinary wins in real data, which is the better-supported and more useful claim, and it is not the same claim.

Next: when a win is a warning is about the run that is turning into something else, and tilt and loss limits is the stop rule. Back to after a big win.