Why winning the auction usually means you overpaid

Auctions, by construction, work against the buyer. The seller has the information. The buyer has the bet. The mechanism selects for the highest bid. The highest bid comes from the bidder with the highest estimate of what it's worth. The highest estimate is, on average, the most wrong estimate.

That's not a story about bidders being irrational. It's a mechanism problem.

The UK 3G spectrum auction in 2000 raised £22.5B from five telecoms. The winners lost money; the largest had unwound the deal by 2005.

M&A processes, even when labelled "negotiated," frequently are auctions: sell-side advisor, multiple bidders under NDA, deadline bids. The label doesn't change the structure.

This pattern was named almost six decades ago. In the 1950s, three petroleum engineers — each at a different oil company — looked at the same offshore tract and submitted sealed bids based on private geological data. The tract kept going to the highest bidder. The highest bidder almost always lost money. The pattern repeated across auctions.

In 1971, three economists — Capen, Clapp, and Campbell — gave it a name: the winner's curse.

The mechanism they identified is the same one driving 3G. When every bidder has a different estimate and the auction goes to the highest bid, the winning estimate is, on average, above the true value. The difference isn't noise. It's structural. The bidder with the median estimate loses to the 90th-percentile bidder. The 90th-percentile loses to the 99th. The bid that wins is the bid whose underlying assumption is most aggressively wrong. The asset itself can be fundamentally sound — the mechanism doesn't care.

Most M&A processes function as auctions, even when labelled "negotiations." Multiple bidders, NDA, financial info shared, bids compared. The mechanism selects for the highest bid regardless of label. The winner's curse applies.

The obvious response from a sophisticated bidder: fine, I'll haircut my bid. Bid less than my estimate to account for the fact that the winner is the optimist. Smart bidders do this. Doesn't save them.

Haircutting only changes who wins. The auction still selects the bidder whose haircut bid is highest. A 20% haircut on a $100M estimate bids $80M. A 20% haircut on a $130M estimate bids $104M. The second bidder still wins. The second bidder still overpays. The curse consequently persists.

The test that works: what would you bid if you were the only bidder? Not second-highest. Not third. The only bidder, in a quiet room with the seller, no competitive process at all. If your only-bidder bid equals your auction bid, you're buying the asset. If your auction bid is higher, you're paying for the thrill of winning.

The only-bidder bid is, by construction, what the asset is worth to you. I know this sounds obvious. It isn't — most bidders, in the room, lose track of which number is which. And, frankly, which number is theirs.

The only-bidder bid isn't the lowest price you could force the seller to accept. It's what the asset is worth to you. Only the second question matters for the winner's curse.

The cure isn't smarter modeling. The bidder with the best model doesn't win — the bidder with the most optimistic one does. Smarter modeling is a competitive disadvantage, because it tends to make you more cautious.

Decide the maximum price before the data room opens. If the process pushes your bid above it, walk away. The walk is the most expensive thing you'll do in the process. Walk.

For a $20M-$100M acquisition, the spread between your auction bid and your only-bidder bid is often 15-30%. On a $50M target, that's $7.5M-$15M redirected from the asset to the ego of winning.

(Though the math is hard — synergies, integration costs, the dis-synergies nobody mentions in the CIM. See why your model is probably lying to you.) The hard part is the conversation with yourself when the process asks for a number above your only-bidder cap. Write your rationalizations before the process. Pre-process is the only version that survives the room.