Last lesson handed you a number — losses loom about 2.25× larger than equal gains — and a slogan to go with it: we judge outcomes from a reference point, not as final wealth. That number is the headline. This lesson is the engine room. Because “losses hurt more” only becomes a prediction machine once you can answer two questions precisely: more than what? (the reference point) and how much more, at every size of outcome? (the value function). Get those two gears turning and you can forecast a startling amount of behavior — including the bit where the same person plays it safe with gains and gambles with losses in the same breath.
Two gears, then. The reference point decides where zero sits. The value function decides how steeply the world bends on either side of it. Loss aversion is what you get when you bolt them together.
As always: lock in a guess before you peek.
Before you read — take a guess
You were quietly expecting a $2,000 raise this year. Your boss announces a $5,000 raise. Your coworker was expecting $10,000 and also got $5,000. Same $5,000 in both pockets — why does it feel like a gift to you and a slap to them?
The reference point: where you put zero changes everything
The analogy. Stick one hand in ice water and the other in hot water for a minute, then plunge both into the same lukewarm bowl. The cold hand screams warm; the hot hand screams cold — from the same water. Your skin doesn’t report absolute temperature; it reports the change from where it just was. Your sense of money, status, and outcomes works the identical way. There’s no absolute “good paycheck” — only a paycheck that’s higher or lower than the one your mind was braced for.
The precise definition. A reference point is the baseline against which you judge an outcome as a gain (a step up) or a loss (a step down). It’s usually the status quo (what you have right now) or an expectation (what you were braced to get). Crucially, prospect theory’s whole claim is that you evaluate the change from this point — not the final state of wealth. Two people can end the day with the exact same bank balance and feel opposite things, because they started the day pointed at different baselines.
Worked example — one paycheck, two feelings. Take the pretest’s two raises and write them out. Same $5,000 hits both accounts; the only thing that differs is the baseline each person carried into the room.
| You | Your coworker | |
|---|---|---|
| Reference point (what you expected) | $2,000 | $10,000 |
| Actual raise | $5,000 | $5,000 |
| Felt as | a +$3,000 gain | a −$5,000 loss |
| Emotional reading | a happy surprise | a punch in the gut |
Nothing about the money explains the gap — both got $5,000. Everything about the reference points does. You beat your baseline by $3,000 and felt the glow of a gain; your coworker fell $5,000 short of theirs and felt the sting of a loss, even while taking home a raise. The exact same dollars, scored against different zeros, produce opposite emotions.
Worked example — the friendly reference point at the pump. Now watch someone choose your zero for you. A gas station charges $3.00 a gallon for credit and $2.95 for cash. Two ways to post that identical pair of prices:
| Sign | Reference price (the “real” price) | Cash feels like | Credit feels like |
|---|---|---|---|
| ”Credit $3.00 — cash discount $2.95” | $3.00 | a 5¢ gain (a reward) | the normal price |
| ”Cash $2.95 — credit surcharge $3.05” | $2.95 | the normal price | a 10¢ loss (a penalty) |
Same two numbers at the pump. But the word discount plants your reference point at the higher price, so paying cash feels like winning a nickel — and loss aversion means that framing massively outperforms a “surcharge” sign that makes credit users feel penalized. This is not an accident; credit-card lobbyists fought hard to keep stations saying “cash discount” instead of “credit surcharge.” Same prices, friendlier zero.
The reference point is movable — and you don't always hold the pen
The single most useful thing to remember about reference points is that they move, and that other people move them on purpose. A marketer’s “was $120, now $80” sets your zero at $120 so $80 reads as a $40 win. A negotiator’s outrageous opening offer resets your zero so their real target later feels like a relief. An employer’s hinted bonus becomes the baseline against which the actual bonus disappoints. None of these change the underlying outcome by a cent — they change the baseline you measure it from, which is often the whole ballgame.
When to use it
Reach for the reference point the instant an outcome feels much better or worse than the bare numbers justify — a raise that disappoints, a discount that thrills, a “loss” that’s actually still a profit. Ask the operative question: what baseline am I scoring this against, and did I choose it or did someone hand it to me? If a feeling seems out of proportion to the dollars, the reference point is almost always doing the lifting. And when you’re describing an outcome to someone else, remember you’re choosing their zero whether you mean to or not.
The value function: the curve that turns outcomes into feelings
The analogy. Imagine a machine that takes in an objective outcome — ”+$100,” ”−$400” — and spits out a subjective feeling, “how good or bad that actually lands.” That machine has a personality, and its personality is a curve. Plot objective outcome along the bottom and felt value up the side, and you don’t get a straight 45-degree line (which is what a cold calculator would draw). You get a lopsided, kinked S.
The precise definition. The value function is the curve mapping objective outcomes to subjective value — how good a gain feels and how bad a loss hurts, as a function of size. Kahneman and Tversky’s value function has three defining properties, and together they are prospect theory’s psychology of risk:
- Reference-dependent. The curve is centered at the reference point — the origin is your zero, gains stretch right, losses stretch left. (That’s gear one, bolted in.)
- Diminishing sensitivity. The curve is concave for gains (bends and flattens as gains grow) and convex for losses (bends and flattens as losses grow). In plain terms: the jump from $0 to $100 feels bigger than the jump from $1,000 to $1,100, even though both add $100. The first $100 of a windfall thrills; the hundredth barely registers. Same on the loss side — the first $100 lost stings far more than the difference between losing $1,000 and losing $1,100.
- Loss aversion. The curve is steeper on the loss side than the gain side — it takes a sharp downward bend right at the origin, the famous kink. That kink is the 2.25× from last lesson, drawn as geometry: the loss arm plunges faster than the gain arm climbs.
That’s the whole engine: a curve, centered on a movable point, gentle going up, steep going down. Drive it yourself below — drag the loss-aversion slider and watch the loss arm dive while the gain arm only ambles up.
Losses loom larger
Drive the value function: gentle up, steep down, kinked at zero
Drag the loss-aversion slider and watch the curve. Gains bend gently up; losses plunge — and the more loss-averse you are (higher λ), the deeper the drop. The kink at zero is the whole model.
At λ = 2.25, winning $100 feels like +58, but losing $100 feels like -129 — the loss hurts about 2.25× as much as the equal gain feels good. That asymmetry is loss aversion.
Same numbers, opposite choice: the framing flip
A disease threatens 600 people. The two programs below are numerically identical across the frames — only the wording changes. Toggle the frame and watch which option the curve above prefers.
Program A: save 200 people for sure
★ people lean hereProgram B: ⅓ chance to save all 600, ⅔ chance to save no one
In the gain frame the curve is concave, so a sure 200 saved beats the risky gamble: most people play it safe (risk-averse).
Notice what the slider reveals. At the default λ ≈ 2.25, winning $100 feels like a modest climb but losing $100 feels like a steep fall — the loss point sits more than twice as far below zero as the gain point sits above it. Crank λ higher and the loss arm crashes toward the floor; pull it down to 1.00 and the curve becomes symmetric (a person who weighs losses and gains exactly the same). The bend of each arm is diminishing sensitivity; the difference in steepness between the arms is loss aversion. Two distinct features of one curve.
Two gears, two jobs — don't blur them
Diminishing sensitivity and loss aversion are easy to confuse because they live on the same curve. Keep them apart: diminishing sensitivity is about size — each extra dollar (won or lost) matters less than the one before, which is why both arms bend. Loss aversion is about direction — losses are weighted more than equal gains, which is why the loss arm is steeper. You could have one without the other. A curve can bend without being lopsided; ours does both.
Diminishing sensitivity flips your risk attitude
Here’s the payoff that makes the curve worth drawing: those two bends don’t just describe feelings, they predict gambles. And they predict something genuinely strange — the same person is cautious about gains and reckless about losses.
The analogy. Think of the value curve as a hill on the gain side and a cliff on the loss side. On a gently rounding hill, a sure step up is worth more than a coin-flip between a bigger step and no step — the curve “rewards” certainty. On a plunging cliff, a sure drop is so painful that a coin-flip offering any chance of not falling looks tempting — the curve “rewards” the gamble. Same person, opposite instinct, because the ground bends the opposite way on each side.
The precise idea. Because the gain arm is concave, a sure gain beats a risky gain of equal average value — so people are risk-averse for gains. Because the loss arm is convex, a sure loss feels worse than a risky loss of equal average value — so people are risk-seeking for losses: they’ll gamble to avoid a certain loss. Let’s walk both with real numbers.
Worked example — the gain side (concave ⇒ risk-averse). I offer you a choice. Most people grab the sure thing, even though the gamble pays the same on average.
| Option | Outcomes | Average value | What most people pick |
|---|---|---|---|
| A — sure thing | $50 guaranteed | $50 | ✅ the sure $50 |
| B — gamble | 50% of $100, 50% of $0 | $50 | passed over |
Equal averages, yet the concave gain arm makes the certain $50 feel worth more than the bumpy ride to an average $50 — so you pocket the sure thing. Risk-averse for gains.
Worked example — the loss side (convex ⇒ risk-seeking). Now mirror it. Identical structure, flipped into losses. Watch the preference flip too.
| Option | Outcomes | Average value | What most people pick |
|---|---|---|---|
| C — sure thing | lose $50 guaranteed | −$50 | passed over |
| D — gamble | 50% of losing $100, 50% of losing $0 | −$50 | ✅ the gamble |
Same equal averages — and now most people refuse the sure loss and roll the dice, hoping to escape losing anything at all. The convex loss arm makes a certain $50 loss feel worse than the average-$50 gamble, so they gamble to dodge the sure hit. Risk-seeking for losses. Same human, same $50 stakes, opposite choice — purely because one frame is gains and the other is losses.
This flip is the spine of the next lesson: if a sure outcome reverses from “grab it” to “avoid it” just by recoloring it as a gain or a loss, then how you word a choice can flip what people choose — with the actual outcomes untouched. (Kahneman and Tversky pinned the curvature of these bends with a parameter near 0.88 — the exponent the explorer above uses to draw each arm — so this isn’t hand-waving; it’s a measured, fitted shape.)
Sort each situation by which instinct the value function predicts. Are people leaning safe because it's framed as a gain, or gambling because it's framed as a loss?
Place each item in the right group.
- A company doubling down on a failing project, gambling on a turnaround, rather than booking the certain write-off now
- Refusing a sure $200 settlement and rolling the dice in court hoping to owe nothing instead of a certain $200
- Cashing out a small but certain investment profit rather than holding for a possibly larger, possibly zero gain
- A gambler down $300 making one big reckless bet to "get back to even" rather than accept the sure $300 loss
- Locking in a guaranteed prize on a game show instead of risking it for a bigger one
- Taking a guaranteed $500 bonus over a coin flip for $1,000 or nothing
Cause and effect: which property of the value function is directly responsible for people being risk-SEEKING when they face a sure loss?
No — and this is the heart of why prospect theory is a model and not an insult. People aren’t random or stupid; they’re following a consistent, fitted curve. The risk-averse-for-gains, risk-seeking-for-losses pattern shows up reliably, across cultures, with a measurable shape (curvature ≈ 0.88, loss-aversion ≈ 2.25). “Irrational” implies unpredictable; this is the opposite — it’s so predictable you can exploit it (which marketers and casinos do) or defend against it (which lesson 5 will teach you). The departure from cold expected-value math is systematic, not chaotic. That’s exactly what makes it dangerous and useful at the same time.
The pitfall: assuming people optimize final wealth
The two traps that ambush this lens
Trap 1 — assuming people optimize final wealth. The tidy textbook agent of expected value and expected utility cares only about final states: end the day with $1,005,000 and it doesn’t matter whether you got there by gaining $5,000 or losing $5,000 from $1,010,000. Real humans care enormously which path it was, because they live at a reference point and feel the change. If you model people as cold final-wealth maximizers, you’ll mispredict nearly every risky choice they make — the coin flip they refuse, the stock they won’t sell at a loss, the lawsuit they fight past all reason.
Trap 2 — forgetting the reference point is a choice someone can set. It’s tempting to treat your baseline as a neutral fact of nature. It isn’t. The marketer’s “was $120,” the negotiator’s anchor, the employer’s hinted figure — each one plants your zero so the real outcome reads as a win or a loss. Ignoring your reference point doesn’t make it neutral; it just hands the pen to someone else. The only defense is to notice the baseline you’re using and ask who chose it.
When to use it
Pull out the value function whenever a choice involves risk and you can spot which side of zero it’s framed on. Ask: is this dressed as a gain or a loss? If it’s a gain, expect yourself (and others) to play it safe — sometimes too safe, cashing out winners early. If it’s a loss, expect the reckless lean — fighting sunk costs, chasing losses, doubling down on the doomed project. And the master move, the one that defuses half of these traps at once: before you decide, re-anchor. Restate the choice in terms of final outcomes, name your reference point out loud, and check whether someone handed you that zero on purpose. A choice that flips when you move the baseline was never really about the outcomes.
Recap
The two-gear engine behind loss aversion, in five bolts:
- A reference point is the baseline (usually the status quo or an expectation) you score outcomes against. You judge the change from it — a gain or a loss — not final wealth. The same $5,000 is a gift or a slap depending on the zero you carried in.
- The reference point is movable, and marketers, negotiators, and employers routinely set it for you (“cash discount,” “was $120,” outrageous opening offer). Ignoring yours just hands someone else the pen.
- The value function maps objective outcomes to subjective feeling. It’s reference-dependent (centered at your zero), shows diminishing sensitivity (concave for gains, convex for losses — each extra dollar matters less), and exhibits loss aversion (steeper on the loss side — the kink at the origin, the 2.25× drawn as geometry).
- Diminishing sensitivity flips your risk attitude: concavity makes you risk-averse for gains (sure $50 over a 50/50 shot at $100), convexity makes you risk-seeking for losses (gamble over a sure $50 loss). Same person, opposite instinct, set only by which side of zero the choice sits on. K&T fit the curvature near 0.88.
- The pitfalls: don’t model people as cold final-wealth maximizers (they feel the path, not just the destination), and don’t treat your reference point as neutral — it’s a choice, and often someone else’s.
Check yourself: reference points and the value function
Two investors both end the year with a $500,000 portfolio. One started at $400,000; the other started at $600,000. Prospect theory predicts:
Check your answer to continue.
Where this goes next
You now hold the full engine: a movable reference point that sets zero, and an S-shaped value function — concave up, convex down, kinked and steeper on the loss side — that turns outcomes into feelings and quietly flips your appetite for risk depending on which side of zero you’re standing. Lesson 3, Framing Effects, takes that engine and does something almost unsettling with it: it flips a real, high-stakes choice using nothing but wording. The same medical program, the same lives, the same odds — described once as “lives saved” and once as “lives lost” — and people reverse their preference, exactly as the value function predicts. You’ve already met the flip in miniature in the explorer’s lower panel and the sure-thing-versus-gamble tables. Next we let it loose on the famous Asian-disease problem and watch a frame change minds without changing a single fact.