Winning the Auction Is the Bad News
In 1971, three petroleum engineers at Atlantic Richfield published a paper about something odd they'd noticed in offshore oil lease auctions. Companies were bidding against each other for drilling rights, winning, drilling — and earning returns well below what they'd projected. Not occasionally. Systematically.
Nobody was incompetent. The bids were based on real geology by real geologists. The problem was structural, and it has a name.
The mechanism
Suppose a tract of seabed contains a quantity of oil worth, in truth, $100 million. Nobody knows that. Each company surveys it and comes up with an estimate. Those estimates cluster around $100 million — some come in at $85 million, some at $115 million, most somewhere between.
Everyone bids their honest estimate. Who wins?
The company whose estimate was highest. Which is to say: the company that was most wrong, in the most expensive direction. Winning isn't a reward for being right. It's evidence that you were the most optimistic person in the room.
That's the winner's curse, and the worst part is what happens when more people show up.
Win the auction, lose the money
Everyone is bidding on the same thing, genuinely worth $100 to whoever gets it. Nobody knows that. Each bidder has an honest estimate that happens to be a bit high or a bit low.
standard deviation of each estimate
Average winning bid
$121.5
the item is worth $100
Average winner's profit
−$21.5
the curse, in one number
Auctions the winner overpaid
100%
Bid shading applied
$0
none — bid as estimated
- Bidding the honest estimate
- Shading the bid down
The honest-bidding line falls as the room fills up. That is the whole phenomenon: with more bidders, the winner is more reliably the person whose estimate was furthest above the truth. Winning is evidence that you were too optimistic — which is why more competition makes the curse worse, not better.
Show the numbers
| Bidders in the room | Winner's profit, bidding the estimate | Winner's profit, shading the bid |
|---|---|---|
| 2 | −$8.9 | −$0.2 |
| 4 | −$15.4 | −$0.9 |
| 6 | −$19.2 | −$1.4 |
| 8 | −$21.5 | −$1.3 |
| 10 | −$23.1 | −$1.1 |
| 12 | −$23.5 | $0.0 |
| 14 | −$25.7 | −$1.0 |
| 16 | −$27.0 | −$1.1 |
| 18 | −$27.6 | −$0.8 |
| 20 | −$27.7 | −$0.0 |
Drag the room from two bidders to twenty with everyone bidding their honest estimate, and the winner's average profit falls off a cliff. With two people, the highest of two estimates is only modestly above the truth. With twenty, you're taking the maximum of twenty draws — and the maximum of a larger sample sits further out in the tail.
More competition doesn't discipline the price toward the truth. It pushes the winning bid further past it.
The condition that makes it real
This only happens when the item has a common value — worth roughly the same to whoever ends up with it, with the uncertainty being about what that value is. Oil in the ground is worth what the oil is worth, regardless of who pumps it.
If values are genuinely private — a painting is worth what it's worth to you, and your neighbour's opinion is irrelevant — there's no curse. You know your own value with certainty. There's nothing to overestimate.
Most real business situations sit somewhere between, and the ones that lean common are the expensive ones:
- Acquisitions. The target's future cash flows are roughly the same whoever buys it. Every bidder is estimating the same unknown, and the winner is the one who modelled the synergies most generously. A large literature finds acquirers frequently overpay; this is a substantial part of why.
- Construction and procurement tenders. The job costs what it costs. The winning contractor is often the one who most underestimated the difficulty — which is how you end up with a bid that wins and a project that loses money.
- Free agency in sport. The player's future performance is what it is. The team that wins the bidding is the one that projected it most optimistically.
- Hiring in a competitive market. Several firms interview the same candidate, form estimates, and bid salaries. The one that wins is often the one that read the strongest signal into the same evidence everyone else saw.
What to actually do
Shade your bid — and shade harder as the room fills. This is the core correction, and the interactive above shows it working: the green line sits near break-even at every room size. The logic is uncomfortable but sound. Ask yourself not "what do I think this is worth?" but "what would this be worth given that I win — given that my estimate turned out to be the highest of everyone's?" That conditional is strictly lower than your raw estimate, and the more bidders, the lower it is.
Set your maximum before the room gets warm, and write it down. The curse is a statistical problem, not an emotional one, but auction fever is a separate and entirely real phenomenon that stacks on top of it. A number decided in advance, by someone who isn't in the room, is worth a lot.
Spend money on reducing the uncertainty. The curse scales with how wrong estimates are. Drag the noise slider down in the figure above and the whole problem shrinks. Better geology, better due diligence, a longer look at the data room — these are not overheads, they're the direct lever on how much you overpay.
Notice when you won easily. If you expected a fight and got the thing cheaply, that is not always good news. It may mean everyone else knows something you don't.
The version worth remembering
In a private-value world, winning means you wanted it most. In a common-value world, winning means you estimated highest. Those feel identical from the inside and they are completely different facts — and the second one comes with a bill attached.