Losses Hurt Twice as Much: Prospect Theory

For most of the twentieth century, economics ran on a single elegant assumption about how people choose under risk: you look at each possible outcome, weigh it by its probability, and pick whatever gives you the highest expected satisfaction. That's expected utility theory, and it is beautiful, tidy, and mathematically tractable.

It also does not describe actual human beings.

In 1979, two psychologists, Daniel Kahneman and Amos Tversky, published a paper in an economics journal (a slightly hostile act in itself) that documented exactly how real people deviate from the textbook, and then proposed a replacement model that fit what people actually do. They called it prospect theory. It went on to win a Nobel Prize, spawn behavioral economics as a field, and quietly explain half the dumb things investors do.

The problem: the textbook person does not exist

Kahneman and Tversky did something economists mostly weren't doing: they asked people. They ran simple choice experiments, often with students, offering pairs of gambles and recording which one people preferred. Then they checked whether those preferences were consistent with expected utility theory.

They weren't. And the failures were not random noise, they were systematic, repeatable, and predictable. A few of the patterns they hammered on:

The certainty effect. People overvalue outcomes that are certain relative to outcomes that are merely very likely. Going from a 95 percent chance to a 100 percent chance feels like a much bigger upgrade than going from 60 percent to 65 percent, even though it's the same five points of probability.
The reflection effect. Take a gamble people reject when it's framed as a gain, then flip the signs so it's about losses. Very often their preference flips too. People are typically cautious about gains and reckless about losses.
The isolation effect. People simplify choices by throwing away components the options share, and which components they throw away depends on how the problem is described. So the same underlying choice, described two ways, produces two different answers.

That last one is fatal for the classical theory. In expected utility theory, a choice is a choice. If describing it differently changes your answer, you are not maximizing anything stable.

The key idea via analogy: a thermostat, not a bank balance

Here is the heart of prospect theory, and it's simpler than the math suggests.

Classical theory says your happiness depends on your total wealth, like a bank balance. A person with 1,010,000 dollars is a fixed, knowable amount happier than a person with 1,000,000 dollars, full stop.

Kahneman and Tversky said no: people are more like a thermostat. A thermostat doesn't care what the absolute temperature is, it cares whether you are above or below the setting. Humans evaluate outcomes as gains and losses relative to a reference point, usually wherever they currently are. Winning 10,000 dollars feels like a win whether you started with nothing or with a million. What registers is the change, not the level.

Once you accept that, three consequences fall out, and together they are prospect theory:

Losses loom larger than gains. The pain of losing 100 dollars is roughly twice the pleasure of gaining 100 dollars. This is loss aversion, and it is the single most consequential idea in the paper. It's why people cling to a losing stock, why they refuse a coin flip that pays 110 dollars on heads and costs 100 on tails (a bet with clearly positive expected value), and why "don't lose what you have" beats "go get more" as a motivator.
Diminishing sensitivity, in both directions. The difference between gaining 10 dollars and gaining 20 dollars feels big. The difference between gaining 1,010 dollars and 1,020 dollars feels like nothing, even though it's the same 10 dollars. The same is true on the loss side. This produces the S-shaped value function: risk-averse when you're up, risk-seeking when you're down. Ahead for the day, you lock in the win. Down for the day, you swing for the fences to get back to even. Every casino, and every trading desk, has watched this happen live.
Probabilities get bent. People don't use probabilities as stated, they use a decision weight that overweights small probabilities and underweights moderate-to-large ones. That single warp explains why the same person buys a lottery ticket (overweighting a tiny chance of a huge win) and buys insurance (overweighting a tiny chance of a huge loss). Classical theory struggles to make one person do both. Prospect theory makes it obvious.

Why it mattered

Prospect theory did not just tweak a model, it changed who was allowed to make claims about economic behavior.

It made "irrationality" scientific. Before this, deviations from rationality could be waved away as noise that averages out. Kahneman and Tversky showed the deviations are directional and reproducible, which means they don't average out. They aggregate. And things that aggregate can move prices.
It gave behavioral finance its engine. The disposition effect (selling winners too fast, riding losers too long), myopic loss aversion (why stocks seem to pay too much given their risk), mental accounting, the equity premium puzzle, and much of the sentiment literature all trace their DNA back to loss aversion and the reference point.
It gave practitioners a lens for their own behavior. Every trader who has moved a stop loss "just this once", cut a winner early to "book the profit", or doubled down on a losing position to get back to breakeven is running the S-shaped value function on live capital. Naming the bias is the first step to building a system that overrides it.
It won the Nobel. Kahneman received the 2002 Nobel Memorial Prize in Economic Sciences for this line of work. Tversky had died in 1996 and the prize is not awarded posthumously, which is one of the field's genuine tragedies.

The honest limitations

It describes, it does not explain. Prospect theory is a superb description of what people choose. It does not tell you why the brain works this way, or what the reference point will be in any given situation, which is a large loophole. If I can pick the reference point after the fact, I can explain anything, and a theory that explains anything predicts nothing.
The reference point is the soft underbelly. Is your reference point your purchase price? Your peak portfolio value? What your neighbor made? Your expectation from last week? The theory needs one and doesn't fully supply one. Later work has tried to pin this down, with partial success.
The original version had technical problems. The 1979 formulation could, in some setups, produce nonsensical answers when you had many possible outcomes. Kahneman and Tversky fixed this themselves in 1992 with cumulative prospect theory, which handles the probability weighting more carefully. When people say "prospect theory" today they usually mean the 1992 version.
Lab gambles are not markets. The evidence came from students choosing between hypothetical gambles for modest stakes. Whether the same distortions survive high stakes, expertise, competition, and repeated feedback is a genuinely open question, and it is exactly the question that matters if you want to trade on it. The lab shows the bias exists. It does not prove you can profit from it.

The one-line takeaway

Kahneman and Tversky showed that people do not evaluate outcomes against their total wealth but against a reference point, that losses hurt roughly twice as much as equivalent gains feel good, and that these distortions are systematic enough to be modeled, which turned "people are irrational" from an excuse into a research program.