Tuesday, 30 April 2013

Why parties may deliberately write incomplete contracts

Incomplete contracts are big news in economic theory these days. Economists, and lawyers, would generally agree that almost all contracts are incomplete. It is simply too costly for the contracting parties to anticipate the many contingencies that may occur and to write down unambiguously how to deal with them.Incomplete contracts have been used to explain, among other things, the boundaries of firms, the internal organisation of firms, firms’ financial decisions, the costs and benefits from privatisation; and the organisation of international trade between inter- and intrafirm trade. But there is an obvious question you have to ask: Why are contracts incomplete?

In a new NBER working paper, More is Less: Why Parties May Deliberately Write Incomplete Contracts, Maija Halonen-Akatwijuka and Oliver D. Hart look at this question.

First what doesn't (fully) work when thinking about the reasons for incompleteness:
The idea that transaction costs or bounded rationality are a total explanation for this is not convincing. In many situations some states of the world or outcomes are verifiable and easy to describe, appear relevant, and yet are not mentioned in a contract. A leading example is a breach penalty. A contract will usually specify the price the buyer should pay the seller if trade occurs as intended, but may not say what happens if there is a breach or under what conditions breach is justified. Of course, sophisticated parties often do include breach penalties in the form of liquidated damages but this is far from universal.

A second example concerns indexation. Since a worker’s marginal product varies with conditions in the industry she works in as well as the economy as a whole we might expect to see wages being indexed on variables correlated with industry profitability such as share prices or industry or aggregate unemployment, as well as to inflation. Such an arrangement might have large benefits, allowing wages to adjust and avoiding inefficient layoffs and quits of workers (see, e.g., Weitzman (1984) and Oyer (2004)). Indeed Oyer (2004) argues that high tech firms grant stock options to employees to avoid quits. Yet the practice does not seem a common one overall. Similarly, in the recent financial crisis many debt contracts were not indexed to the aggregate state of the economy; if they had been the parties might have been able to avoid default, which might have had large benefits both for them and for the economy as a whole.

How do we explain the omission of contingencies like these from a contract? One possibility is to argue that putting any contingency into a contract is costly – some of these costs may have to do with describing the relevant state of the world in an unambiguous way – and so if a state is unlikely it may not be worth including it (see, e.g., Dye (1985), Shavell (1980)). This is often the position taken in the law and economics literature (see, e.g., Posner (1986,p.82)). However, this view is not entirely convincing. First, states of the world such as breach are often not that unlikely and not that difficult to describe. Second, while the recent financial crisis may have been unlikely ex ante, now that it has happened the possibility of future crises seems only too real. Moreover, finding verifiable ways to describe a crisis does not seem to be beyond the capability of contracting parties. Thus one might expect parties to rush to index contracts on future crises. We are not aware of any evidence that this is happening.

A second possibility is to appeal to asymmetric information (see, e.g., Spier (1992)). The idea is that suggesting a contingency for inclusion in a contract may signal some private information and this may have negative repercussions. Such an explanation does not seem very plausible in the case of financial crises – where is the asymmetry of information about the prospects of a global crisis? – but it may apply in other cases. For example, if I suggest a (low) breach penalty you may deduce that breach is likely and this may make you less willing to trade with me. Or if you suggest that my wage should fall if an industry index of costs rises I may think that you are an expert economist who already knows that the index is likely to rise.

Even in these cases asymmetric information does not seem to be a complete answer. Asymmetric information generally implies some distortion in a contract but not that a provision will be completely missing. For example, in the well-known Rothschild-Stiglitz (1976) model, insurance companies offer low risk types less than full insurance to separate them from high risk types. But the low risk types are not shut out of the market altogether – they still obtain some insurance (and the high risk types receive full insurance). Indeed to explain why a contingency might be omitted from a contract Spier assumes a fixed cost of writing or enforcing contractual clauses in addition to asymmetric information.
What do Halonen-Akatwijuka and Hart offer that is new? They analyse when and why parties will deliberately write incomplete contracts even when contract-writing costs are zero.Their approach is based upon the ideas first formulated in Hart and Moore (2008) which sees contracts as "reference points". Under this approach a contract is viewed as delineating what parties feel they are entitled to. Importantly  the parties to a contract do not feel entitled to outcomes which are outside the set of outcomes specified by the contract but they may feel entitled to different outcomes within those specified in the contract. If a party does not receive what he feels entitled to he is "aggrieved" and "shades" (or cuts back) on performance. This creates deadweight losses. Halonen-Akatwijuka and Hart write,
We have argued that adding a contingency of the form, “The buyer will require an extra good or service in event E”, has a benefit and a cost. The benefit is that there is less to argue about in event E; the cost is that the reference point provided by the extra service in event E may increase argument costs in states outside E. Similarly indexing a price or wage to an exogenous variable has the benefit that if this variable tracks the buyer’s value and seller’s cost closely then breakdown in trade can be avoided; but the cost that if the index does not track value and cost closely the reference point provided by the indexation may make renegotiation harder when trade does break down.

Our principal result is that the relative benefit and cost of adding a contingency or indexing will be sensitive to how closely the parties agree about what is a reasonable division of surplus when an incomplete contract is renegotiated. The benefit is likely to exceed the cost when parties have very different views about what is a reasonable division of surplus, but the opposite will be the case if they have shared views. Under the latter conditions an incomplete contract will be strictly optimal. Our results can shed light on why wage indexation, although observed in some situations (see Card (1986) and Oyer (2004)), is not more common.

It is worth considering how our theory’s implications differ from those of a theory based on asymmetric information. Consider the Nanny example [see below] in the introduction where the question is why a late fee is not introduced. The asymmetric information explanation would be that introducing the late fee might signal to the Nanny that the employer knows that he is unpunctual, which makes the job less attractive. But this problem could be presumably solved through the choice of a high late fee. Or take the case of wage indexation. If an employee is offered a contract whereby the wage is indexed on some signal, the employee might think that the employer already knows that the signal will be such that the employee’s wage is low, making the contract less attractive. But this would suggest that in an optimal contract the wage should not vary much with the index, not that it should not vary at all. Only by introducing costs of contractual clauses (as in Spier (1992)) can one explain a complete lack of indexation.

In contrast in our theory, introducing a late fee or any amount of indexation has a discontinuous effect: it introduces a brand new reference point. We have seen that in some circumstances the cost of doing this outweighs the benefit.

Our theory also has different implications from the asymmetric information one regarding the timing of incompleteness. Signaling favorable private information is particularly important at the beginning of a relationship. In our theory one possible explanation for similar views about the division of surplus is the history of the relationship between the buyer and the seller. If the parties have interacted before they may have grown to know and like each other, with the implication that each will become more generous about sharing surplus (see the social influence theory of Kelman (1958)). Therefore we would expect contracts to become less complete in long-term relationships, but be more complete when such relationships are formed -- in contrast to the asymmetric information theory.

Finally, our approach may also be able to explain why parties often use general rather than specific language in contracts. For example, parties negotiating acquisitions frequently include a clause that excuses the buyer if the target seller suffers a “material adverse change” (see Schwartz and Scott (2010)). According to our theory the advantage of a general clause is that it creates a neutral reference point: In terms of the model of Section 2 it is like describing states s2-s4, rather than event E, as a situation where the add-on should be provided. In contrast spelling out particular contingencies that qualify as a material adverse change may complicate renegotiation in other contingencies that are not easily described but where the parties also intended to excuse the buyer. Asymmetric information theories do not seem to have much to say about this issue.
The Nanny example,
Suppose that you hire a Nanny to work Monday-Friday from 9am-5pm for $600 per week ($15 per hour). There is a chance that you will get stuck in traffic and will be late. Should you include a late fee of, say, $30 per hour in the Nanny’s contract? (Being late is a verifiable contingency.) Including the late fee could prevent bad feelings later on about how much the Nanny should be paid when you are late. But if you include the late fee, it may create some expectation by the Nanny concerning what she should receive if, say, you need her to work on the weekend. (There may be several reasons for you to want her to work on the weekend—some business, some pleasure— and it may be difficult to distinguish between these in advance.) She might feel that $30 per hour is the appropriate reference point for such an arrangement, whereas you might feel that $15 per hour is. If you and the Nanny have similar views about what is reasonable absent a reference point, it may be better to leave the late fee out and renegotiate as needed.

No comments: