Avoiding the winner's curse in construction

A tender opens. Six contractors, one number each, lowest wins. The envelopes come out and the spread is wide, the lowest well under the next. Everyone in the room knows what they are looking at, because everyone has been on both sides of it. The winner has probably left money on the table. Not through carelessness. Winning a competitive bid is itself evidence that you may have been the most optimistic estimator in the room, and optimism about a job’s value is the same thing as underpricing its risk.

Economists call this the winner’s curse. In a competitive bid for something of uncertain value, the winner is disproportionately the one who overestimated it. More bidders make it worse, because the winning number is pulled from the extreme tail of everyone’s estimates. It was first seen in oil companies bidding for drilling rights, and Richard Thaler made it famous. His book on it, first published in 1992, was reissued in October 2025, and the updates carry a finding worth sitting with: the anomalies from the old lab experiments show up in real-world data today with almost no decay. This is not a historical curiosity that markets have since learned around.

But the winner’s curse is only the hook here. The argument is about something the research on it revealed almost in passing, and that has quietly become urgent in 2026: experienced contractors were never protected from the curse by skill at estimating. They were protected by three specific things around the estimate, and all three are eroding at the same time. That simultaneity is the problem. Any one of them weakening would be manageable. Losing all three at once changes what it takes to bid safely.

The experiment that should bother a bid desk

In 1989, Dyer, Kagel and Levin ran a study that is uncomfortable to read if you bid for a living. They took experienced construction executives, people who competed for contracts professionally, and put them in a laboratory auction next to students who had never bid on anything.

The executives fell into the winner’s curse as hard as the students did.

That is the whole finding, and it is strange, because these executives did not go bankrupt in real life. They ran contracting businesses that worked. So experience clearly protected them somewhere, just not in the lab. Dyer and Kagel went back to the industry to ask why, and the answer they published in 1996 is the useful part. Experienced contractors avoid the curse in the field for three reasons, and none of them is a better estimate of value. The lab removed all three, and the moment it did, decades of experience counted for nothing.

The three protections

The first is escape mechanisms. In the field, a contractor who bids too low is not sealed into the loss. There are variations, change orders, renegotiations, claims. The bid opens a commercial relationship rather than closing the question. The lab took that away. One number, no recourse.

The second is learned evaluative process. Experienced contractors read a job in industry-specific ways that stop them bidding too low before they ever write the number. Not a formula. A trained sense of where risk hides, assembled over years of watching which jobs went wrong and why. The lab handed them an abstract task their instincts had no grip on, and the instincts had nothing to attach to.

The third is private value. Real contractors do not value the same job identically. One has idle capacity, one has a crew already down the road, one has a strategic reason to want in. So the lowest bidder is not always the most optimistic. Sometimes they are simply the one for whom the job is worth more. The lab made the prize identical for everyone, which left estimation error as the only thing separating one bid from the next.

Strip all three and the veteran has nothing but the estimate, and the estimate was never where the experience lived. It lived in the context around it. That is why the lab humbled them. It kept the estimate and deleted the context.

Why 2026 is not 1989

Now the part that should worry contractors. Each of the three protections is thinning, and they are thinning together.

Start with the one that is easiest to miss until it is gone. The learned evaluative process was never institutional. It sat in the heads of individual estimators and bid managers, built over careers, and firms treated it as a fixed feature of the business when it was nothing of the kind. It was personal, and personal protection does not survive personnel change. That generation is retiring across European construction now, and the bench behind them is thin. When one of these people leaves, the firm does not lose a headcount. It loses the specific thing the 1989 study named as the reason experienced contractors avoid the curse, and there is usually nothing in the systems that holds it once the person is gone.

The escape mechanisms are narrowing too, and hardest in the public work that makes up so much of the market. In private contracting there is room to renegotiate a thin bid, to adjust by agreement. Public procurement closes that room by law. Article 72 of the EU procurement directive bars substantial modification after award. A change big enough to matter counts as a new contract and triggers a re-tender. So the classic recovery route, quietly reshaping a bad bid in delivery, is not available the way it is in private work. The first protection is weakest exactly where the volume has moved.

And the private-value cushion is being competed away. As public procurement leans harder on price and standardises scope, the space for a contractor’s own circumstances to justify a lower number shrinks. Force everyone to value the job the same way and the lowest bid is, once again, mostly just the most optimistic one.

That last point answers the obvious objection. Yes, real tendering is not a pure common-value auction, contractors have different costs and different reasons to want a job. That difference was one of the three protections. It is not a reason the curse does not apply. It is one of the things being taken away. The field is drifting toward the lab from more than one direction at once.

What you can actually do about it

Only one of the three is really within a contractor’s control. You cannot legislate the escape mechanisms back, and you cannot slow the retirements. What you can decide is whether the evaluative judgment stays a personal possession or becomes something the organisation holds: a portfolio memory rather than a personal one.

Be honest about the limit here. You cannot bottle a veteran’s intuition, and anyone claiming a model reproduces it is selling something. The judgment itself does not transfer. But its outputs can. The risk that got flagged, the assumption behind a price, the reason a clause was treated as dangerous, the pattern that a certain kind of provision tends to bite in execution. Capture those at the moment they are made, tie them to the tender they came from, carry them into the next one. Then the next estimator, who may be years short of the one who left, at least starts from the accumulated record instead of a blank sheet and whatever they happen to recall. Some of it can be written down in advance, too: the handful of provisions that move risk in tender pack after tender pack, which is what the Tender patterns notes are for.

This is the real reason a tender should not start from scratch every time, and it is a harder problem than it sounds, because the knowledge worth keeping is exactly the knowledge that was never written down. The winner’s curse is not a story about weak estimators. It is a story about what protects strong ones, and about how much of that protection was sitting in people rather than in the firm, quietly, right up until those people started to leave.

The award and the overpayment are usually the same event. Keeping them apart used to depend on having the right person in the room. Increasingly it depends on whether the firm kept what that person knew.