Oliver Hart and Bengt Holmström have just won the Nobel Prize in Economics. They’re both important economists in the area known as “contract theory”, which is important for the theory of the firm — so this work is significant for legal scholars who work in business associations, antitrust (e.g. vertical integration), etc. You can look at Marginal Revolution for a general discussion of Hart and Holmström. (See also here.)
Hart was my professor and one of my advisors in graduate school. This blog post tries to give a taste — in language that might be comprehensible to the non-economist — of what his contribution is to the theory of the firm. Here’s an excerpt from his article An Economist’s Perspective on the Theory of the Firm, 89 Colum. L. Rev. 1757 (1989). You should read the whole thing, where he spends the first part summarizing the various other theories of the firm, since Coase’s seminal 1937 article, before describing his own (“property rights”) theory, which I excerpt below. After this excerpt, he discusses in greater detail how his theory differs from the previous theories.
One way to resolve the question of how integration changes incentives is spelled out in recent literature that views the firm as a set of property rights. This approach is very much in the spirit of the transaction cost literature of Coase and Williamson, but differs by focusing attention on the role of physical, that is, nonhuman, assets in a contractual relationship.
Consider an economic relationship of the type analyzed by Williamson, where relationship-specific investments are important and transaction costs make it impossible to write a comprehensive long-term contract to govern the terms of the relationship. Consider also the nonhuman assets that, in the postinvestment stage, make up this relationship. Given that the initial contract has gaps, missing provisions, or ambiguities, situations will typically occur in which some aspects of the use of these assets are not specified. For example, a contract between GM and Fisher might leave open certain aspects of maintenance policy for Fisher machines, or might not specify the speed of the production line or the number of shifts per day.
Take the position that the right to choose these missing aspects of usage resides with the owner of the asset. That is, ownership of an asset goes together with the possession of residual rights of control over that asset; the owner has the right to use the asset in any way not inconsistent with a prior contract, custom, or any law. Thus, the owner of Fisher assets would have the right to choose maintenance policy and production line speed to the extent that the initial contract was silent about these.
Finally, identify a firm with all the nonhuman assets that belong to it, assets that the firm’s owners possess by virtue of being owners of the firm. Included in this category are machines, inventories, buildings or locations, cash, client lists, patents, copyrights, and the rights and obligations embodied in outstanding contracts to the extent that these are also transferred with ownership. Human assets, however, are not included. Since human assets cannot be bought or sold, management and workers presumably own their human capital both before and after any merger.
We now have the basic ingredients of a theory of the firm. In a world of transaction costs and incomplete contracts, ex post residual rights of control will be important because, through their influence on asset usage, they will affect ex post bargaining power and the division of ex post surplus in a relationship. This division in turn will affect the incentives of actors to invest in that relationship. Hence, when contracts are incomplete, the boundaries of firms matter in that these boundaries determine who owns and controls which assets. In particular, a merger of two firms does not yield unambiguous benefits: to the extent that the (owner-)manager of the acquired firm loses control rights, his incentive to invest in the relationship will decrease. In addition, the shift in control may lower the investment incentives of workers in the acquired firm. In some cases these reductions in investment will be sufficiently great that nonintegration is preferable to integration.
Note that, according to this theory, when assessing the effects of integration, one must know not only the characteristics of the merging firms, but also who will own the merged company. If firms A and B integrate and A becomes the owner of the merged company, then A will presumably control the residual rights in the new firm. A can then use those rights to hold up the managers and workers of firm B. Should the situation be reversed, a different set of control relations would result in B exercising control over A, and A’s workers and managers would be liable to holdups by B.
It will be helpful to illustrate these ideas in the context of the Fisher Body-General Motors relationship. Suppose these companies have an initial contract that requires Fisher to supply GM with a certain number of car bodies each week. Imagine that demand for GM cars now rises and GM wants Fisher to increase the quantity it supplies. Suppose also that the initial contract is silent about this possibility, perhaps because of a difficulty in predicting Fisher’s costs of increasing supply. If Fisher is a separate company, GM presumably must secure Fisher’s permission to increase supply. That is, the status quo point in any contract renegotiation is where Fisher does not provide the extra bodies. In particular, GM does not have the right to go into Fisher’s factory and set the production line to supply the extra bodies; Fisher, as owner, has this residual right of control. The situation is very different if Fisher is a subdivision or subsidiary of GM, so that GM owns Fisher’s factory. In this case, if Fisher management refuses to supply the extra bodies, GM always has the option to fire management and hire someone else to supervise the factory and supply extra bodies (they could even run Fisher themselves on a temporary basis). The status quo point in the contract renegotiation is therefore quite different.
To put it very simply, if Fisher is a separate firm, Fisher management can threaten to make both Fisher assets and their own labor unavailable for the uncontracted-for supply increase. In contrast, if Fisher belongs to GM, Fisher management can only threaten to make their own labor unavailable. The latter threat will generally be much weaker than the former.
Although the status quo point in the contract renegotiation may depend on whether GM and Fisher are one firm rather than two, it does not follow that the outcomes after renegotiation will differ. In fact, if the benefits to GM of the extra car bodies exceed the costs to Fisher of supplying them, we might expect the parties to agree that the bodies should be supplied, regardless of the status quo point. However, the divisions of surplus in the two cases will be very different. If GM and Fisher are separate, GM may have to pay Fisher a large sum to persuade it to supply the extra bodies. In contrast, if GM owns Fisher’s plant, it may be able to enforce the extra supply at much lower cost since, as we have seen in this case, Fisher management has much reduced bargaining and threat power.
Anticipating the way surplus is divided, GM will typically be much more prepared to invest in machinery that is specifically geared to Fisher bodies if it owns Fisher than if Fisher is independent, since the threat of expropriation is reduced. The incentives for Fisher, however, may be quite the opposite. Fisher management will generally be much more willing to come up with cost-saving or quality-enhancing innovations if Fisher is an independent firm than if it is part of GM, because Fisher management is more likely to see a return on its activities. If Fisher is independent, it can extract some of GM’s surplus by threatening to deny GM access to the assets embodying these innovations. In contrast, if GM owns the assets, Fisher management faces total expropriation of the value of the innovation to the extent that the innovation is asset-specific rather than management-specific, and GM can threaten to hire a new management team to incorporate the innovation.
So far, we have discussed the effects of control changes on the incentives of top management. But workers’ incentives will also be affected. Consider, for example, the incentive of someone who works with Fisher assets to improve the quality of Fisher’s output by better learning some aspect of the production process. Suppose further that GM has a specific interest in this improvement in car body quality, and that none of Fisher’s other customers cares about it. There are many ways in which the worker might be rewarded for this, but one important reward is likely to come from the fact that the worker’s value to the Fisher-GM venture will rise in the future and, due to his additional skills, the worker will be able to extract some of these benefits through a higher wage or promotion. Note, however, that the worker’s ability to do this is greater if GM controls the assets than if Fisher does. In the former case, the worker will bargain directly with GM, the party that benefits from the worker’s increased skill. In the latter case, the worker will bargain with Fisher, who only receives a fraction of these benefits, since it must in turn bargain with GM to parlay these benefits into dollars. In consequence, the worker will typically capture a lower share of the surplus, and his incentive to make the improvement in the first place will fall.
In other words, given that the worker may be held up no matter who owns the Fisher assets—assuming that he, himself, does not—his incentives are greater if the number of possible hold-ups is smaller rather than larger. With Fisher management in control of the assets, there are two potential hold-ups: Fisher can deny the worker access to the assets, and GM can decline to pay more for the improved product. As a result, we might expect the worker to get, say, a third of his increased marginal product (supposing equal division with Fisher and GM). With GM management in control of the Fisher assets, there is only one potential hold-up, since the power to deny the worker his increased marginal product is concentrated in one agent’s hands. As a result, the worker in this case might be able to capture half of his increased marginal product (supposing equal division with GM).
The above reasoning applies to the case in which the improvement is specific to GM. Exactly the opposite conclusion would be reached, however, if the improvement were specific to Fisher, such as the worker learning how to reduce Fisher management’s costs of making car bodies, regardless of Fisher’s final customer (a cost reduction, furthermore, which could not be enjoyed by any substitute for Fisher management). In that event, the number of hold-ups is reduced by giving control of Fisher assets to Fisher management rather than GM. The reason is that with Fisher management in control, the worker bargains with the party who benefits directly from his increased productivity, whereas with GM management in control, he must bargain with an indirect recipient; GM must in turn bargain with Fisher management to benefit from the reduction in costs.
Up to this point we have assumed that GM management will control GM assets. This, however, need not be the case; in some situations it might make more sense for Fisher management to control these assets—for Fisher to buy up GM. One thing we can be sure of is that if GM and Fisher assets are sufficiently complementary, and initial contracts sufficiently incomplete, then the two sets of assets should be under common control. With extreme complementarity, no agent—whether manager or worker—can benefit from any increase in his marginal productivity unless he has access to both sets of assets (by the definition of extreme complementarity, each asset, by itself, is useless). Giving control of these assets to two different management teams is therefore bound to be detrimental to actors’ incentives, since it increases the number of parties with hold-up power. This result confirms the notion that when lock-in effects are extreme, integration will dominate nonintegration.
These ideas can be used to construct a theory of the firm’s boundaries. First, as we have seen, highly complementary assets should be owned in common, which may provide a minimum size for the firm. Second, as the firm grows beyond a certain point, the manager at the center will become less and less important with regard to operations at the periphery in the sense that increases in marginal product at the periphery are unlikely to be specific either to this manager or to the assets at the center. At this stage, a new firm should be created since giving the central manager control of the periphery will increase hold-up problems without any compensating gains. It should also be clear from this line of argument that, in the absence of significant lock-in effects, nonintegration is always better than integration—it is optimal to do things through the market, for integration only increases the number of potential hold-ups without any compensating gains.
Finally, it is worth noting that the property rights approach can explain how the purchase of physical assets leads to control over human assets. To see this, consider again the GM-Fisher hypothetical. We showed that someone working with Fisher assets is more likely to improve Fisher’s output in a way that is specifically of value to GM if GM owns these assets than if Fisher does. This result can be expressed more informally as follows: a worker will put more weight on an actor’s objectives if that actor is the worker’s boss, that is, if that actor controls the assets the worker works with, than otherwise. The conclusion is quite Coasian in spirit, but the logic underlying it is very different. Coase reaches this conclusion by assuming that a boss can tell a worker what to do; in contrast, the property rights approach reaches it by showing that it is in a worker’s self-interest to behave in this way, since it puts him in a stronger bargaining position with his boss later on.
To put it slightly differently, the reason an employee is likely to be more responsive to what his employer wants than a grocer is to what his customer wants is that the employer has much more leverage over his employee than the customer has over his grocer. In particular, the employer can deprive the employee of the assets he works with and hire another employee to work with these assets, while the customer can only deprive the grocer of his custom and as long as the customer is small, it is presumably not very difficult for the grocer to find another customer.