Hostage capital

Why does a supplier sometimes make an investment that would be worthless if the deal fell apart?

A supplier has a choice between two machines that do the same job at the same price. One can be sold on to anyone. The other is tooled so specifically to one customer's product that if the contract ends it is scrap. Prudence says buy the first. Suppliers routinely buy the second, and sometimes the customer's willingness to sign depends on it. That is a strange preference — the supplier appears to be paying for the privilege of being trapped. Unless being trapped is what is being sold.

Start with the buyer's actual problem. He needs a supplier who will still be delivering, at quality, in five years. He cannot write that into a contract in any complete way, because neither of them can enumerate in advance what "at quality" will mean when circumstances change. So he is relying on a promise — and the trouble with promises is not dishonesty but that a promise which is free to break carries no information. Every supplier, sincere or not, will make it.

So ask what a sincere supplier could do that an insincere one would find too expensive. Not talk. Not a warranty, if the firm can vanish. It has to be an act whose cost falls only on the party who walks away.

Now the specific machine makes sense. By buying an asset with no resale value, the supplier deliberately destroys his own exit option: leaving now costs him the whole outlay. He has not promised harder, he has made the promise expensive to break, and that is the only kind of promise a stranger should believe. Williamson set this out in 1983 as the deliberate posting of a hostage, and the same logic underlies Klein and Leffler's account of brand investment — money sunk into something unrecoverable on exit is a bond a firm intending to cheat would not post.

But test that, because "sunk cost equals commitment" is too loose and gets the design wrong. Suppose the hostage were something the buyer could seize and use — a valuable general-purpose asset lodged with him. He now gains by engineering a breach, since breaking the deal hands him something worth having. The bond has created a reason to break the thing it was meant to protect.

So the property that matters is not sunkness but a peculiar asymmetry: the hostage must be worth a great deal to the giver and close to nothing to the taker. Williamson's illustration is a king taking a daughter as security who takes the plain one rather than the beautiful one, because he must not want to keep her. That is the load-bearing claim here — non-salvageable to the giver and non-appropriable by the taker. Drop either half and it stops being a commitment device: an asset the giver can resell bonds nothing, and one the taker covets invites the breach.

Now let me try to falsify the account rather than admire it. If hostages solved credibility we would see them everywhere, and we do not. The hostage creates exactly the exposure it cures, pointed the other way: once the supplier's plant is worthless elsewhere, the buyer can demand a lower price knowing the alternative is scrap. So a one-sided hostage is often unacceptable and these deals tend to be mutual — the buyer signs a take-or-pay commitment, or funds part of the tooling. Hostage exchange, not hostage posting. And where reputation works, no hostage is needed at all: in a market where performance is observed and buyers talk, the value of future business is itself the bond, and it is free. Where neither reputation nor a workable hostage exists, the parties stop bonding the exchange and merge instead.

What survives is narrower than the textbook version: hostage capital is one instrument among several, chosen when the specificity can be made asymmetric, when it can be made mutual, and when reputation is unavailable — which is why it appears in new entrants, opaque quality, and thin markets.

The refinement that ties the reasoning together: this is not an argument about trust or goodwill. It is an argument about which promises are informative. Costless promises tell you nothing because everyone makes them; a promise becomes evidence only when keeping it is the cheaper option for the promiser, and the specific asset manufactures that.

The misconception to correct is that the supplier is naive or exploited. He is usually paid for the position — in a longer term, a higher price, a volume guarantee, a share of tooling cost. If he is not, he has posted a hostage and received nothing, which is not commitment but a mistake.

Two honest qualifications. This piece is the mirror image of the collection's explanation of the boundaries of the firm, and they should be read together: there, asset specificity is the problem, exposing a party to hold-up and pushing the transaction inside a firm; here the same specificity is chosen on purpose as the cure for a different problem, credibility. Both are true, which is why the sensible question is never whether specificity is good but who becomes vulnerable and whether they are compensated. And the canonical illustration — Fisher Body's dedicated stamping plants and its absorption by General Motors, in a 1978 paper by Klein, Crawford and Alchian — has been seriously contested since, with Coase and others arguing the archival record does not support the hold-up reading. The theory does not depend on that case, but it is often taught as settled, and it is not.

• Why franchisees are so often required to fund their own fit-out, and what that shares with a dedicated stamping plant.
• How escrow and performance bonds differ from hostages, given that the holder can keep them.

Every term the collection defines is gathered in the glossary.