In the intro we dangled a generous coin flip in front of you — heads you win $110, tails you lose $100 — and you, like nearly everyone, refused a bet that pays you money on average. We named the culprit (loss aversion) and moved on. This lesson stops moving. We’re going to pin the asymmetry down hard enough to put a number on it: not “people kind of don’t like losing,” but a measured coefficient near 2.25 that one of psychology’s most-replicated experiments keeps coughing up across money, time, effort, and — genuinely — monkeys. By the end you’ll be able to compute the favorable bet you still feel right refusing, read the asymmetry straight off a table, and tell loss aversion apart from its lookalike twin, risk aversion, which fools almost everyone.
As always: commit to a guess before you peek.
Before you read — take a guess
A coin flip pays +$110 on heads and −$100 on tails. To make a typical person willing to risk losing $100 on a single coin flip, roughly how big does the potential WIN usually have to be?
The favorable bet you refuse
The analogy. Picture a set of kitchen scales. On one pan you drop the pleasure of a possible win; on the other, the pain of a possible loss. A pure money-maximizer uses honest scales — a dollar of gain exactly balances a dollar of loss. Your scales are rigged: the loss pan is heavier. To get the scale to tip toward “yes, I’ll play,” you have to keep piling extra dollars onto the gain pan until it finally overcomes a loss pan that started out roughly twice as heavy.
The precise setup. Back to the coin flip: heads, +$110; tails, −$100. The rational benchmark you met in the thinking in probabilities course is expected value (EV) — the average outcome if you played forever, each outcome weighted by its odds. Here both outcomes are equally likely, so:
- Half the time you get +$110.
- Half the time you get −$100.
- EV = (0.5 × +$110) + (0.5 × −$100) = +$55 − $50 = +$5 per flip.
A money-maximizer plays this bet every chance they get; turning down a +$5 edge is like declining free money. And yet you won’t play once. That refusal is data: it tells us the loss isn’t being weighed at its face value of $100. It’s being weighed as if it were much larger — large enough that even a $110 gain on the other pan can’t tip the scale.
Worked example — find the indifference point. Now let’s reverse the question. Keep the $100 possible loss fixed and ask: how big must the possible win get before a typical person finally says yes? The experimental answer is about $200. Watch what that does to the EV, and to the felt scales:
| Possible win | Possible loss | Expected value (the rational score) | Typical verdict | What the refusal/acceptance reveals |
|---|---|---|---|---|
| +$110 | −$100 | +$5 | Refuse | Loss outweighs gain even with a positive EV |
| +$150 | −$100 | +$25 | Mostly refuse | A $50 edge still isn’t enough |
| +$200 | −$100 | +$50 | Finally tempted | Gain ≈ 2× loss — the scales roughly balance |
| +$250 | −$100 | +$75 | Accept | Now the gain pan clearly wins |
The headline isn’t in the EV column — every row there is positive, so a money-maximizer takes all four bets without blinking. The headline is the gap between that column and the verdict column. People keep refusing perfectly favorable bets until the potential gain reaches roughly twice the potential loss. That 2-to-1 demand is the asymmetry caught in the act: it is the daylight between expected value (how a bet should be scored) and how losses are actually weighted in the human head.
The refusal is the measurement
Notice the trick we just pulled. We didn’t ask anyone how they feel about losses — feelings are slippery and people are terrible at reporting them. We watched what bet they’d accept and read the asymmetry off their behavior. The win they demand (≈$200) versus the loss they’ll risk ($100) gives us a ratio, and that ratio is a number we can compare across people, stakes, and even species. Turning a vague ache into a measurable coefficient is the whole move that made this a science instead of a proverb.
When to use it
Pull this out the instant you (or someone selling to you) frame a choice as a bet with a downside. Compute the EV first — that’s the rational benchmark — then notice your gut demanding a much sweeter upside before you’ll bite. If you’re refusing bets with clearly positive expected value purely because a loss is possible, that’s loss aversion overriding arithmetic, and it’s worth knowing whether the refusal is wisdom or reflex.
The loss-aversion coefficient (≈ 2.25)
The analogy. Think of an exchange rate between two currencies — “gain-dollars” and “loss-dollars” — that refuses to trade at par. When you convert a loss into how-bad-it-feels, the exchange rate is brutal: each loss-dollar costs you about 2.25 gain-dollars of felt pain. A $100 loss doesn’t feel like losing $100 of happiness; it feels like losing roughly $225 worth.
The precise definition. Loss aversion is the finding that a loss of a given size produces a stronger psychological response than a gain of the same size produces in the opposite direction. The loss-aversion coefficient (often written with the Greek letter lambda) is the multiplier that captures how much stronger. Kahneman and Tversky measured its median at about 2.25 — meaning the displeasure of losing weighs roughly 2.25 times the pleasure of an equal gain. It is one of the most robust findings in behavioral science: the same rough doubling shows up not just for money but for time (a wasted hour stings more than a gifted one delights), effort, social standing, and — in later work — even capuchin monkeys trained to trade tokens, who hate losing a grape they were shown far more than they enjoy an unexpected one. When a result survives that many domains and a different species, it’s not an artifact of one quiz.
Worked example — felt value vs. face value. Take the face value of four outcomes and run them through a roughly 2× multiplier on the loss side (we’ll use 2.25 to match the famous number) to get their felt value — the size of the emotional hit:
| Outcome (face value) | Multiplier applied | Felt value (psychological impact) |
|---|---|---|
| +$100 (gain) | ×1 | +100 units of “good” |
| +$500 (gain) | ×1 | +500 units of “good” |
| −$100 (loss) | ×2.25 | −225 units of “bad” |
| −$500 (loss) | ×2.25 | −1,125 units of “bad” |
Read the table top-to-bottom and the asymmetry jumps out. A $100 gain and a $100 loss are equal on a bank statement — but the loss lands 2.25× harder in the only ledger that drives behavior: the felt one. And it scales: a $500 loss (−1,125 units) outweighs even a $500 gain (+500 units) by more than two to one. This is exactly why the +$110/−$100 coin flip fails. In felt units it’s roughly (+110) versus (−225) — a terrible-feeling proposition — even though in dollars it’s a winner. The math says yes; the scales say absolutely not, and the scales win.
Using a loss-aversion coefficient of about 2.25: which single outcome carries the LARGEST psychological impact (felt value), regardless of sign?
Two ways to misuse the number 2.25
The coefficient is a powerful headline and a dangerous one if you carry it around like a physical constant. (1) It’s a median, not a law. Loss aversion varies — by person (some people barely show it, a few show none), by domain (money vs. health vs. time), and by stakes (the ratio can shift for trivial sums or life-changing ones). “2.25” is a robust central estimate, not the exact, universal exchange rate inside every head. Quote it as “roughly 2,” not “precisely 2.25, always.” (2) Don’t confuse the response with the proverb. Loss aversion is a measured asymmetry in felt value, not a moral claim that “losing is bad.” It’s the multiplier, and the multiplier is what makes the model predictive.
When to use it
Reach for the coefficient whenever you want to predict the strength of someone’s reaction to a downside, not just its direction. Selling something? The fear of losing what they have will move them about twice as hard as the hope of gaining something new — so “don’t miss out” out-pulls “look what you’d get.” Designing a policy, a refund, a guarantee? Remember you’re fighting a roughly 2:1 headwind whenever you ask someone to accept even a small possible loss. Just hold the number loosely: it’s a reliable ballpark, not a guarantee about any one person.
A loss hurts more than a forgone gain
The analogy. Two friends each end the week $50 poorer than they could have been. Ana had $50 in her pocket and lost it through a hole in her coat. Ben had a $50 scratch-card that would have won $50, but he forgot to claim it before it expired. Same final wallet — both are $50 short of the best case. But Ana is furious and Ben just shrugs. A loss of money you had and a failure to win money you didn’t have land completely differently, even when the arithmetic is identical.
The precise idea. Loss aversion bites hardest on things measured below your reference point — and “money I already had” sits right at that reference point, so giving it up registers as a genuine loss. “Money I might have won but didn’t” never entered your pocket or your reference point, so it registers only as a forgone gain — a non-event, a shrug. The wallet ends in the same place; the felt value does not, because one path crosses the painful loss threshold and the other doesn’t.
Worked example — same wallet, different sting. Both people are $50 worse off than the ideal. Watch the felt-value column diverge:
| Person | What happened | Final position vs. ideal | Felt as a… | Approx. felt impact |
|---|---|---|---|---|
| Ana | Lost $50 she already had | −$50 | Loss (below reference point) | ≈ −112 units (2.25 × 50) |
| Ben | Missed winning $50 he never had | −$50 | Forgone gain (a gain that didn’t arrive) | ≈ −50 units, often far less |
Identical balance sheets, very different Tuesdays. This is the engine under a hundred everyday phrases. “We lost the deal” gnaws at a sales team far more than “we didn’t get the deal,” even though the company’s books are identical either way — because “lost” frames the outcome as something taken from us (a true loss) while “didn’t get” frames it as a gain that simply never showed up. Same outcome, different reference point, wildly different pain. Hold onto that, because the next lesson is entirely about how movable that reference point is.
Because a clever operator can decide which side of your reference point an outcome lands on, just by wording it. A “limited-time discount” turns paying full price into a loss you’re avoiding (powerful, ~2.25× motivating) rather than a discount being a mere forgone gain. A free trial puts the product in your hands — now keeping it avoids a loss, while never having signed up would have been a forgone gain you’d shrug at. The endowment effect and framing, two later lessons, are basically this paragraph industrialized. Don’t try to wield it yet — just notice the lever exists, and that other people are already pulling it on you.
When to use it
Deploy this whenever you catch yourself unusually upset about not getting something. Ask: did I actually lose this, or did I merely fail to gain it? If it never entered your pocket, your gut is taxing it at the loss rate (2.25×) when it should be a shrug. The same move works in reverse when persuading others: reframing a forgone gain as a loss (or vice-versa) changes how hard the outcome hits, without changing the outcome at all.
Loss aversion ≠ risk aversion
The analogy. Two people both decline a gamble, so they look identical from across the room. One declines because they dislike uncertainty itself — the swing, the variance, the not-knowing. The other declines because the specific outcome on the table is a loss, and losses are taxed at 2.25×. Same visible behavior, two completely different machines underneath — and the moment you change the setup, they walk in opposite directions.
The precise distinction.
- Risk aversion is about disliking variance — the spread of possible outcomes. A risk-averse person prefers a sure thing to a gamble with the same average, regardless of whether the outcomes are gains or losses. It’s a statement about the shape of the distribution you’ll accept.
- Loss aversion is about the reference point and the kink between gains and losses — the fact that the value curve bends sharply right at zero, dropping steeply into loss territory. It’s a statement about which side of your baseline an outcome falls on, not about variance.
They overlap in the gain domain, which is why they’re so easy to confuse: when you might win money, both loss aversion and risk aversion push you toward the cautious, sure-thing choice, and you can’t tell them apart. But they come apart — dramatically — in the loss domain. Here’s the part that should make you sit up: loss aversion can make people risk-seeking. Facing a sure loss versus a gamble that might avoid the loss entirely, people often grab the gamble — taking on more variance, not less — to dodge the certain pain of crossing into loss. A purely risk-averse person would never do that; they’d take the sure thing to minimize variance. So the same person who refuses a favorable bet over a possible gain (looking risk-averse) will chase a worse bet to avoid a sure loss (looking risk-seeking). That flip is impossible to explain with risk aversion alone — and it’s the whole subject of the next lesson, so we won’t resolve it here. Just file the headline: loss aversion is not risk aversion, and it can flip your appetite for risk depending on which side of the reference point you’re standing on.
Which statement best captures the difference between loss aversion and risk aversion?
Fill in the contrast between the two ideas:
Pick the right option for each blank, then check.
A person who always prefers a guaranteed outcome to a gamble with the same average, whether the outcomes are wins or losses, is showing — a dislike of . By contrast, the asymmetry in which a $100 loss hurts about 2.25 times as much as a $100 gain feels good is , and it depends on a . Surprisingly, this second idea can make people when they’re trying to escape a sure loss.
When to use it
Use this distinction to diagnose why someone (including you) is dodging a choice. If they’d dodge it even when only gains are involved, you’re probably looking at plain risk aversion — a dislike of the swing. If their caution evaporates or even reverses the moment the framing crosses into losses — if they suddenly start gambling to avoid a sure hit — that’s loss aversion, and the lever to pull is the reference point, not the variance. Mistaking one for the other leads to exactly the wrong fix.
Recap
You walked in with a vibe (“losses sting”) and you’re leaving with a measured tool. Pin down these:
- The favorable bet you refuse. A +$110/−$100 coin flip has a positive expected value (+$5), yet people refuse it until the possible win reaches about $200 — a roughly 2:1 demand that reveals losses are weighted far above their face value.
- The loss-aversion coefficient ≈ 2.25. A loss produces a stronger psychological response than an equal gain by a multiplier of about 2.25. It’s a median, not a constant — robust across money, time, effort, and even capuchin monkeys, but variable by person, domain, and stakes.
- A loss beats a forgone gain. Losing $50 you had (a true loss, ≈2.25×) stings far more than failing to win $50 you never had (a forgone gain, a shrug) — even with identical wallets. This is why “we lost the deal” hurts more than “we didn’t get the deal.”
- Loss aversion ≠ risk aversion. Risk aversion dislikes variance; loss aversion is about the reference point and the kink at zero — and it can make people risk-seeking when escaping a sure loss. They look alike for gains and split apart for losses.
Check yourself: the asymmetry
A coin flip pays +$120 on heads and −$100 on tails, an expected value of +$10 per flip. Most people still refuse. What does loss aversion say is going on?
Check your answer to continue.
Where this goes next
You can now put a number on the sting, read the asymmetry off a table, and tell loss aversion apart from its risk-aversion lookalike. But we’ve left two giant promissory notes unpaid: what exactly is this “reference point” everything is measured from, and why does loss aversion sometimes make people cautious and sometimes make them reckless? Lesson 2, Reference Points & the Value Function, pays both off at once. It reveals the two-part engine humming under everything you just learned: a reference point you can slide left or right (so the very same outcome can be a gain or a loss depending on where you set the baseline), bolted to an S-shaped value function — concave for gains, convex for losses, and steeper on the loss side — that bends sharply at the origin. That kink is the 2.25 you just met, drawn as a curve; and that convex loss side is the risk-seeking flip we just refused to resolve. You’ll get to grab the curve and drive it yourself.