This prize is long overdue. The theory of the firm is one of the big ideas in economics and as far as I can tell the Nobel committee was right to trace it back to Williamson.
A firm is a container for a bunch of highly idiosyncratic, repeated, informal transactions. A great thought experiment is to wonder why these transactions are not conducted through a market, making the firm dissolve away. Instead of giant firms building and selling cars, why aren’t there a bunch of tiny firms each doing a tiny part with all of their interaction governed by the market or by contract? There are three main reasons why.
First, its costly to use the market. If the chassis firm is going to be buying axles from the axle firm all the time, it would save transaction costs by just integrating. Then it can “procure” axles with a memo.
Second, the contracts would be impossibly complicated and unwieldy. Imagine writing a complete blueprint for the car, breaking it down into individual instructions for every actor who is supposed to contribute, laying out the timing when each party is supposed to arrive and do his part, describing payments as a function of the performance of each interdependent action, etc. That is probably already impossible, but imagine you could do that. Now suppose that a supply shock requires you to use different materials for the chassis. This would require a coordinated change in many parts of the car, to keep structural integrity, balance, etc. The entire volume of contracts would have to be re-written.
The third reason is the one that adds richness to the theory of the firm. Most of the transactions that occur within firms require parties to make investments that only make sense within the context of that specific firm. The party making the investment has little or no option to recover the value of the investment outside the firm. When the chassis firm contracts with the firm building auto bodies, it writes down minute specifications that must be met. The bodies produced could not be sold to any other chassis maker. The firm building auto bodies must invest in the machinery that can make these specific bodies. Once sunk that investment has no value outside of the specific relationship with the chassis firm.
The reason this adds richness to the theory is that it explains which transactions must be encompassed within firms. Transactions that require relation-specific investments would be crippled if conducted across firm boundaries. Once the die is cast for building these specific auto bodies, the chassis firm has no reason to pay a price that compensates for the sunk cost because there is no outside option. This hold-up problem implies that the auto body firm has poor incentives to make the investment in the first place, unless it integrates with the chassis firm and becomes a claimaint to the profits of the integrated firm.