Why do firms exist?

And since they do, why aren’t we all in one big firm?

i. Why Firms?

Why do firms sometimes combine, and other times do not? It is surprisingly difficult to create a model of the world which does not lead to the equally absurd conclusions that no firms exist, or that only one firm, of infinite size, exists. Klein, Crawford, and Alchian (1979) offer an elegant solution in “Vertical Integration, Appropriable Rents, and the Competitive Contracting Process”. The heuristic for whether or not firms vertically integrate is how specific the products are. If they are generic, and can be easily replaced, then firms choose to acquire them through market processes outside of the firm. If they are extremely specific, and require specialized investment, then firms vertically integrate and do it all under the same management.

Companies often choose to contract out parts of a task they need done. Rather than hire something to do the task full time for them, they get a contract for each individual job. Imagine a construction company building a house. Perhaps the client is asking for a particular interior style requiring specialized craftsmanship. They would hardly have enough demand to keep someone employed full time just to do that, so they hire an outside contractor. 

Focusing on length of contract, or the quantity of work done, is the wrong thing to focus on though. That the business be large enough to be done in-house is a necessary, but not sufficient condition, for vertical integration. Firms are perfectly happy to hire contractors for employment-like contracts. A substantial portion of flight attendants are actually not hired directly by the airline, but by a third company which contracts with the airline. Some trucks are driven by direct employees of trucking companies, but many are driven by owner-operators, who lease the equipment and may have a relationship with a specific company for getting freight to haul. In the context of the Industrial Revolution, many early textile factories did not truly employ workers, but essentially rented out equipment to individual contractors, who were paid on a piece-rate system. 

For Klein, Crawford, and Alchian, the critical factor which leads to vertical integration is specificity. If the inputs can be put to other uses, then there is no compelling reason not to use market processes. If you are building a house, and are in no particular rush to find 2x4s for the project, you just buy from whoever has the best prices. If you are a car manufacturer who wants the body of a car built to extremely specific specifications, as in the paper, then you are going to integrate, because if you have no other options, then the supplier or buyer can take advantage of you. What are you to do if they say that the unit cannot be supplied on time, and they need more money to do it? Go elsewhere? Undo the specialized investments you made? Everyone would be better off if nobody could engage in this sort of opportunistic sniping for rents, but it is very hard to stop. 

This gives us a simple heuristic to make predictions about what will happen to vertical integration in response to different conditions. A book publisher will contract out book printing, because it has the time to search for the best prices, and is agnostic as to which city manufacturing will take place in. Meanwhile, a newspaper publisher will almost always own its own printing presses, because they need the newspapers printed that day, every day, in that city only. 

ii. Stronger than it looks

I think the case for vertical integration being due to specificity of inputs can be made even stronger. KC&A are implicitly focusing on games where the optimal strategy is not subgame perfect. More specifically, they focus on scenarios where the game takes place over time and play takes place sequentially, giving opportunities for opportunistic defections. Social welfare would be maximized, in the long run, if people abided by their contracts. But, if specialized investment is undertaken, and *only then* does the other party take action, they could demand a larger slice of the pie. This can be strengthened to include simultaneous games by considering the Myerson-Satterthwaite theorem. When an input is sufficiently specialized, both the buying and selling firms are monopolist and monopsonist. They do not know the exact value of the other, only the range of values in which it may occur. Given this, the optimal private bargaining strategies include being willing to not sell the item if it does not exceed some value, even if that value exceeds your private value. Myerson-Satterthwaite does not hold in repeated games, but if there is some non-zero negotiating cost, then the number of times the game can be repeated is also bounded, and there’s no guarantee we arrive at an optimal outcome. Moreover, simply having a long enough expected relationship need not guarantee cooperation. If people know when the game will end, they know that they can defect then – and knowing this, other people want to beat you to the punch and defect first. 

iii. The Tirole and Maskin Critique, and Insurance

You might be saying, why not simply make the contract unbreakable, and let the payments fall where they may? Due to the revelation principle, there exists some game which will make everyone reveal their true beliefs – if you can ascertain payoffs, and prevent renegotiation, then there being various indescribable contingencies doesn’t matter. This is the point of Tirole and Maskin 1999 – you don’t need to lay out all the possibilities, just what happens if different payoffs are reached.

To put it clearly, suppose there is one firm which wants to induce investment by another party into producing some item, the nature of which cannot be specified in advance. I want you to make me a piece of art, for example. I might have general directions, which might make it useful only for me, but I want to leave you free to use all possible knowledge in creating it. I don’t even know what I’m actually looking for. Ahead of time, I state the amount that I am willing to pay for the work of art, and the artist puts in the optimal amount of effort. It is necessarily possible to construct a game in which everyone reveals their true beliefs about the world. (Such a game is given in this explanation of the paper by Eric Maskin. If you don’t get this paper, don’t worry – there exists not only this, but a 40 page lecture by Tirole, simply explaining the original idea. It’s tough). Then you can simply make payments and everyone is just as well off

I think this challenge, while interesting, assumes too much about ascertainable payoffs. In the real world, it is very hard to accurately assess the quality of inputs or outputs. Much compensation is in the form of better or worse on the job conditions, which could not be reliably assessed by anyone, much less an outside party. Once you break this, you are once again back to ownership mattering. I would also not underrate the possibility that people do not want anything they cannot understand. If firms work well enough, then you’ll have a hard time making anyone change.

Firms exist. Why is not always so clear. Klein-Crawford-Alchian, while they make convincing arguments about the reasons for vertical integration, cannot really explain horizontal integration – why should firms own many departments, all making totally different things, under the same roof? Still, for vertical integration, “is this asset specific” is the clearest heuristic we have for determining how people will arrange their business.