%Paper: ewp-le/9506002 %From: Eric Rasmusen %Date: Wed, 14 Jun 95 13:37:38 -0500 \documentstyle[12pt,epsf] {article} \parskip 10pt \textheight 8in \begin{document} % \begin{raggedright} \parindent 24pt \parskip 10pt \titlepage \vspace*{12pt} \begin{center} \begin{large} {\bf The Economics of Agency Law and Contract Formation }\\ \end{large} \vskip 15pt May 8, 1995 \\ \bigskip Eric Rasmusen \\ \vskip 1in {\it Abstract} \end{center} This article addresses issues that arise in agency law when agents make contracts on behalf of principals. The main issue is whether the principal should be bound when the agent makes a contract with some third party on his behalf which the principal would immediately wish to disavow. The resulting tradeoffs resemble those in tort law, so the least-cost-avoider principle is useful for deciding when contracts are valid and may be the underlying logic behind a number of different legal doctrines applied to agency cases. In particular, an efficiency explanation can be found for the undisclosed principal rule, which says that the principal is generally bound even when the third party is unaware that the agent is acting as an agent for him. \vskip .3in \begin{small} \noindent \hspace*{20pt} Indiana University School of Business, Rm. 456, 10th Street and Fee Lane, Bloomington, Indiana, 47405-1701. Office: (812) 855-9219. Fax: 812-855-3354. Internet: Erasmuse@indiana.edu.\\ Formerly titled: ``The Undisclosed Principal Problem in Contract Law''. Indiana University Working Paper in Economics \# 95-006. %Draft: 0.1 (Draft 1.1, February %1991). I would like to thank Edwin Greenebaum and participants in seminars at the Indiana University School of Business and Georgetown Law School for helpful comments. \end{small} %%-----------------------------%-------------------------------------- ---- \newpage \noindent {\bf 1. Introduction} Economists have long used the principal-agent model to explain the intricacies of labor compensation, the organization of hierarchies, the design of securities, and a host of other problems. In the paradigmatic model, a principal hires an agent to exert effort, which he cannot directly observe. The agent is tempted to be slack in his effort, and the principal tries to overcome this moral hazard by designing a contract under which agent compensation is based on output. The terminology of ``principal'' and ``agent'' comes from the law of agency, which has ancient roots and was formerly considered one of the fundamental ideas of the common law. Although agency composes a considerable body of law, it is now rarely taught as a separate course in law schools and as an area of law it has not attracted the attention of economists. The issues are quite different from those in the principal-agent model. Instead of the economist's concern that the contract will not induce appropriate effort by the agent, the lawyer's concern is over what happens when the agent takes inappropriate action, action which may involve no less effort. The agent places an order with a supplier when he has been forbidden to do so; he drives the principal's delivery truck into a schoolbus; he hires the wrong employee for the principal's business. For the economist, the agency problem is how to give the agent incentives for the right action; for the lawyer, it is how to mop up the damage when the agent has taken the wrong action. The economist's problem involves only the principal and the agent. The agent's slack effort affects both of them: the principal, because the task is done poorly, and the agent, because the principal, foreseeing the low effort, will not pay as high a salary. The solution is also internal to the transaction: the parties voluntarily enter into a contract that ameliorates the moral hazard by basing the agent's pay on his output. As with economic transactions in general, the law is needed only to enforce the contract. The lawyer's problem usually does involve a third party.\footnote{A prominent area of agency law which does not involve third parties is the law of fiduciaries: whether the agent has acted properly on behalf of the principal when the contract between them is vague, and how much the agent must compensate the principal for mistakes. See Cooter \& Freedman (1991). } The agent's misbehavior may not have any direct adverse effect on the principal. The principal is not in the schoolbus that is wrecked by the agent, and unless the law enforces the contract, he is unhurt by the agent's foolish or unauthorized purchases. Third parties are harmed, however, so government intervention can aid efficiency. When the agent takes a mistaken action, the damage must be allocated to someone---principal, agent, or third party. The best-known issues involve torts: what happens when an agent injures a third party? The law deals with these involuntary creditors according to the doctrines of {\it vicarious liability} or {\it respondeat superior}, which make the principal liable for torts committed by his agent in the course of the agent's duties.\footnote{Economic analyses of vicarious liability include Chu \& Quian (1992), Kornhauser (1982), and Sykes (1984, 1988).} The present article will focus on a different problem: an agent who makes contracts on the principal's behalf which the principal himself would not have made. Section 2 will lay out the relevant concepts in agency law and illustrate them with the classic case of {\it Watteau v. Fenwick}. Section 3 will construct a model of mistaken contracts, in which the similarities between tort and contract make the least-cost-avoider principle useful. Section 4 will apply the least-cost-avoider principle to a variety of illustrations that show when the principal is bound by contracts and when he is not. Section 5 will analyze a special set of problems involving contracting when the third party is not aware that the agent is acting for a principal--- the ``undisclosed principal'' problem. \bigskip \noindent {\bf 2. The Law of Agency} The concept of agency has long been an important part of the common law. When the American Law Institute began to compile the influential summaries of the common law known as the ``restatements,'' the Restatement of Agency was important enough to be second in the series (after contracts). The {\it Restatement } defines agency as ``the fiduciary relation which results from the manifestation of consent by one person to another that the other shall act on his behalf and subject to his control, and consent by the other so to act.''\footnote{ American Law Institute, {\it Restatement of the Law, Second: Agency 2d}, St. Paul, Minnesota: American Law Institute, 1958, \S 1. All references are to this second restatement, hereafter called the {\it Restatement}. Although I will rely on the {\it Restatement} as my guide to the law in this article, it should be kept in mind that it is not binding on judges, who are supposed to follow the decisions in their particular state. } \newpage According to Roscoe Steffen, a noted authority, \begin{quotation} \begin{small} The essentials of agency are few... First, the relation is a {\it consensual} one; an agent {\it agrees}, or at least {\it consents} to act under the {\it direction} or {\it control} of the principal. Second, the relation is a {\it fiduciary} one; an agent agrees to act {\it for } and {\it on behalf} of the principal. He is in no sense a proprietor entitled to the {\it gains} of enterprise-- nor is he expected to carry the {\it risks}.\footnote{Steffen (1977), p. 26.} \end{small} \end{quotation} In the present article, we shall take for granted that the agent will not assume the risks of the enterprise, either because the efficient contract puts the risk there, or because the agent lacks the resources to assume liability.\footnote{This limitation is for reasons of length only. Questions involving the agent's liability for contracts are both important and interesting, but require a different approach.} We will focus instead on contracts made by the agent with a third party on behalf of the principal. The {\it Restatement}'s ``general principle of interpretation'' says that \begin{quotation} \begin{small} ``An agent is authorized to do, and to do only, what it is reasonable for him to infer that the principal desires him to do in the light of the principal's manifestations and the facts as he knows or should know them at the time he acts.''\footnote{{\it Restatement}, \S 33.} \end{small} \end{quotation} More specifically, the law provides at least six reasons why the principal may be bound by contracts made by the agent:\footnote{This discussion is taken from Chapter 1 of Ramseyer \& Klein (1994), from Steffen (1977), and the {\it Restatement}. I have adopted the categories of Ramseyer and Klein, and added estoppel as a separate category. } {\it 1. Actual Express Authority.} The principal has entered into an explicit agreement with the agent authorizing him to take particular action. When the board of directors of a company votes to authorize the president to purchase a new office building, this is actual express authority. {\it 2. Actual Implied Authority.} The principal has entered into an explicit agreement to employ the agent, and although he has not specifically authorized the particular action at issue, the agent can reasonably infer that authority for that action has been delegated to him. If the general manager of a department store hires clerks, the store is bound by his contract even if he was not expressly granted that authority.\footnote{{\it Restatement}, \S 52, 73.} {\it 3. Apparent Authority}. The principal has no agreement with the agent authorizing the action, but a third party could reasonably infer that the agent was authorized.\footnote{``Apparent authority is the power to affect the legal relations of another person by transactions with third persons, professedly as agent for the other, arising from and in accordance with the other's manifestations to such third persons.'' {\it Restatement}, \S 8. } If a sales manager claims he has authority to sell flour without confirmation from the home office, and such sales are customary in the flour business, then he has apparent authority even if other authority is lacking.\footnote{ North Alabama Grocery Co. v. J.C. Lysle Milling Co. (1921) 205 Ala. 484, 88 So. 590, as described in Steffen, p. 128). See also {\it Restatement}, \S 49. } Note that this differs from actual implied authority in that apparent authority may exist even if the principal has expressly forbidden the agent's action. Apparent authority depends on the beliefs of the third party, not on the actual relation between principal and agent. {\it 4. Estoppel}. The principal is ``estopped'' from objecting to the agreement made by the agent if the principal could have intervened to prevent the confusion over authority; e.g., if the principal overheard the agreement being made and failed to assert that the agent was unauthorized.\footnote{ But see Black (1990) , p. 63, which seems to conflate estoppel with apparent authority: ``Agency by estoppel: When the principal, by his negligence in supervising the agent, leads a third party to reasonably believe that the agent has authority to act for the principal.''} {\it 5. Ratification}. If no other authority exists, but the principal agrees to the contract once he learns about it, this ratification binds the principal. If the flour salesman has no authority to sell wheat, but he makes a contract anyway, that contract is binding if the flour company agrees to it upon learning of the salesman's actions. {\it 6. Inherent Agency Power}. This is the least well-defined reason for liability. \S 8a of the {\it Restatement} says that ``Inherent agency power is a term used in the restatement of this subject to indicate the power of an agent which is derived not from authority, apparent authority or estoppel, but solely from the agency relation and exists for the protection of persons harmed by or dealing with a servant or other agent.'' The agency relationship may give the agent the power to harm third parties even if there is no manifestation by the principal that the agent is acting on his behalf. Vicarious liability is based on inherent agency power, and these powers also arise in contract. As an illustration, let us consider perhaps the best-known case involving agency and contracts, {\it Watteau v. Fenwick}. \begin{quotation} \begin{small} ``From the evidence it appeared that one Humble had carried on business at a beer-house called the Victoria Hotel, at Stockton-on-Tees, which business he had transferred to the defendants, a firm of brewers, some years before the present action. After the transfer of the business, Humble remained as defendants' manager; but the license was always taken out in Humble's name, and his name was painted over the door. Under the terms of the agreement made between Humble and the defendants, the former had no authority to buy any goods for the business except bottled ales and mineral waters; all other goods required were to be supplied by the defendants themselves. The action was brought to recover the price of goods delivered at the Victoria Hotel over some years, for which it was admitted that the plaintiff gave credit to Humble only: they consisted of cigars, bovril, and other articles. The learned judge allowed the claim for the cigars and bovril only, and gave judgement for the plaintiff for 22$l$. 12$s$. 6$d$. The defendants appealed.''\footnote{{\it Watteau v. Fenwick}, 1 Q.B. 346 (1892). The appeal lost.} \end{small} \end{quotation} This case raises problems for both standard agency concepts and everyday notions of fairness. There is no actual authority, either express or implicit, for the agent to order the cigars, because he was expressly instructed not to order them. There is no apparent authority, because the principal did nothing to convey the idea that the manager was acting as an agent. Estoppel and ratification do not apply. All that remains is ``inherent agency powers'': the ability of the manager, based on his employment by the principal, to harm third parties by making contracts. Everyday notions of fairness do not help either. Who should bear the cost of the mistaken order for cigars? Not the principal, it seems, since he never ordered the cigars and expressly forbade the agent from doing so. Not the agent, it seems, since he was not buying the cigars for himself and received no benefit from them. Not the third party, it seems, since he had no way of knowing that the manager was an agent who lacked authority. As in other areas of the law where morality and legal rules come into conflict, efficiency may come to our aid. Section 3 will look at the implications of different legal rules on the incentives to avoid entering into mistaken contracts. \newpage \bigskip \noindent {\bf 3. A Model of Contracts Made by Agents} The principal wishes to buy a good he values at $V_p$, which costs the third party $V_t$ to produce. The principal can either buy the good directly, at transaction cost $c_{na}$, or hire an agent, at the lower cost $c_a$. With probability $f$, the agent mistakenly orders the wrong good, which the principal values at only $(V_p-R)$ and which cannot be resold for more than that amount. The agent error has a probability given by the convex function $f(c_p, c_t)$ which is decreasing in $c_p$ and $c_t$, the care the principal and the third party take to prevent the error.\footnote{ Convexity of $f$ implies that the second derivatives $f_{pp}$ and $f_{tt}$ are positive (so there are diminishing returns to care), and $f_{pp}f_{tt} - f_{pt}^2 >0$ (so the effect of each party's care on the marginal benefit of the other party's care is small relative to the diminishing returns). } Let us assume that if the contract is mistaken and the principal is liable, he accepts delivery of the wrong good, but if he is not liable, the contract is renegotiated at a price that is lower by amount $R$, so the third party bears the cost of the mistake. Our focus will be on the choice of the care levels $c_p$ and $c_t$. The principal's care has a large number of interpretations. It is he who has made an agreement with the agent, and at some cost he can incorporate incentives to deter agent error. This cost includes the cost of formulating and negotiating the agreement, the cost of compensating the agent for his increased effort to avoid errors, and the cost of incentives under imperfect information (e.g., increased risk-bearing by the agent and the real costs of punishments). The principal can also exert care by choosing agents carefully and by monitoring them to increase the agent's incentives to avoid error and to catch erroneous contracts before any reliance costs are incurred. The care level $c_p$ incorporates all these avenues of error avoidance. The third party has less scope for avoiding agent error because he does not select or compensate the agent. His chief advantage is that he is on the spot when the contract is being formed, so he can detect some errors more easily. Even if he does not know the principal's desire perfectly, the third party can detect gross agent errors, take care to avoid inducing agent error, ask about the agent's instructions, and contact the principal directly for confirmation. These comprise the care level $c_t$. The transaction price, $P$, is the result of bargaining between the buyer and the seller. We will assume the price is negotiated to split the expected surplus evenly between the principal and the third party given their common knowledge of the probability of agent error. If no agent is used, and the principal pays the transaction cost, the payoffs of the principal and third party are $\pi_p= V_p -P - c_{na}$ and $\pi_t= P - V_t$. The total surplus from the transaction is therefore \begin{equation} \label{e1} Surplus (no \; agent) = V_p -V_t - c_{na}. \end{equation} If the price splits the surplus equally between the two parties, it will yield \begin{equation} \label{e2} P = V_t + \frac{V_p -V_t - c_{na}}{2}. \end{equation} The price is falling in $c_{na}$, so the principal and the third party share any reduction in the transaction cost. The principal receives a direct benefit, paying out a smaller $c_{na}$, and the third party receives an indirect benefit, a higher value for the price $P$. If the agent is used, an error may occur, in which case it is efficient to breach the erroneous contract and write a new contract correcting the error. The individual payoffs depend on who pays the agent, and on the legal rule. If the legal rule is that the principal is bound by an erroneous contract, the expected payoffs are \begin{equation} \label{e3} \pi_p= (1-f(c_p, c_t)) (V_p-P) + f(c_p, c_t)(V_{p} -P- R) - c_{a}- c_p \end{equation} and \begin{equation} \label{e4} \pi_t= P- V_t - c_t, \end{equation} for an expected social surplus of \begin{equation} \label{e5} Surplus (with\; agent)= V_p- V_t - f(c_p, c_t) R - c_{a} - c_p - c_t. \end{equation} This equation describes the social surplus regardless of the legal rule. It consists of the gains from trade, $(V_p- V_t)$, the expected loss due to error, $f(c_p, c_t) R$, the cost of the agent's effort, $c_{a}$, and the care taken to avoid error, $(c_p+c_t)$. All these exist regardless of the legal rule, which will only affect the levels of $c_p$ and $c_t$, unless it completely chokes off use of the agent. The socially optimal care levels, $c_p^*$ and $c_t^*$, are found from the first order conditions for maximizing the social surplus in (\ref{e5}), \begin{equation} \label{e6} -f_p(c_p^*, c_t^*) R - 1 =0 \end{equation} and \begin{equation} \label{e7} -f_t(c_p^*, c_t^*) R - 1 =0.\footnote{The notation ``$f_p$'' and ``$f_t$'' refers to the derivatives of $f$ with respect to $c_p$ and $c_t$.} \end{equation} In the first-best, the agent is used if he reduces expected transaction costs, i.e., if \begin{equation} \label{e8} f(c_p^*, c_t^*) R + c_{a} + c_p^* + c_t^* < c_{na} . \end{equation} The equilibrium price under these care levels will be \begin{equation} \label{e9} P = V_t + \frac{V_p -V_t - f(c_p^*, c_t^*) R - c_{a} - c_p^* + c_t^* }{2}. \end{equation} We will ordinarily assume that the agent will indeed be used. Because the parties split the surplus, they split the gains from using an agent. If the probability of agent error falls exogenously, the principal receives a direct gain, bearing the mistake cost of $R$ less often, but the third party receives an indirect gain, a higher price $P$. The agent reduces transaction costs, to the benefit of both parties, and both parties should be interested in a legal rule that encourages efficient monitoring. It is not true that when the principal hires the agent it helps the principal alone: a reduction in transaction costs helps both sides of the transaction. We will consider four legal rules for dealing with mistaken agreements: 1. Enforce, 2. Void, 3. Split the cost, and 4. Enforce unless the third party was careless. \noindent {\underline {\it Rule 1: An Erroneous Agreement is Enforced.}} {\it (Strict Liability of the Principal).} Under this legal rule, the principal is always bound by his agent's contract. The third party does not bear any of the cost of error, and so will choose zero care. Predicting this, the principal will choose $c_p$ to maximize his expected payoff, \begin{equation} \label{e10} \pi_p= (1-f(c_p, 0)) (V_p-P) + f(c_p, 0)(V_{p} - R - P) - c_{a}- c_p, \end{equation} yielding the first order condition which characterizes his equilibrium care, $\tilde{c_p}$: \begin{equation} \label{e11} -f_p(\tilde{c_p}, 0) R -1= 0. \end{equation} The third party's care to avoid agent error is inefficiently low: $0 < c_t^*$. The principal's care may be either higher, lower, or unchanged relative to the first-best level, depending on how the care levels of the principal and the third party interact. When they complement each other, the cross-partial $ f_{tp}$ is positive, and the principal's effort will be low: $\tilde{c_p} < c_p^*$. If the third party never reads letters from the principal, the principal will devote little effort to explaining what the agent is to buy. When $ f_{tp}$ is negative, principal and third party effort substitute for each other, and principal effort will be high. If the third party never checks dubious orders with the principal, the principal will hire more able agents than in the first-best. When $ f_{tp}$ is zero, the principal's care will be at the first best: $\tilde{c_p}=c_p^*$. If the third party's effort choice is to not check for scrivener's errors made by the agent, the principal's incentive to clearly explain his preferences to the agent is close to being unaffected. \noindent \underline{ {\it Rule 2: An Erroneous Agreement is Void.} } {\it (No Liability of the Principal).} This rule puts the entire liability on the third party. If the agent makes an erroneous contract, the contract is rescinded and the third party bears the cost $R$. The payoffs are \begin{equation} \label{e12} \pi_p= V_p-P - c_{a}- c_p \end{equation} and \begin{equation} \label{e13} \pi_t= P- V_t -f(c_p, c_t)R - c_t. \end{equation} The principal would choose zero care, and the third party would choose care to solve \begin{equation} \label{e14} -f_t(0, \tilde{c_t})R - 1=0. \end{equation} The comparison with the first-best is analogous to that under Rule 1, but it is now the principal who chooses zero care and the third party whose care depends on the cross-partial. \noindent \underline{{\it Rule 3. The Principal and the Third Party Split the Cost of the Error.}}\footnote{The closest tort analogy to this is comparative negligence. Comparative negligence, however, becomes an issue only if both parties are negligent, whereas Rule 3 splits the cost of the error even if both parties have taken the efficient level of care.} This rule splits the cost of error between the principal and the third party. Let the share of the mistake cost $R$ paid by the principal be some value $\theta$ between, but not including, zero and one. \begin{equation} \label{e15} \pi_p= (1-f(c_p, c_t)) (V_p-P) + f(c_p, c_t)(V_{p} - \theta R - P) - c_{a}- c_p \end{equation} and \begin{equation} \label{e16} \pi_t= P- V_t - f(c_p, c_t)(1-\theta) R - c_t. \end{equation} The first order conditions for the principal and third party are \begin{equation} \label{e17} -f_p(\hat{c_p}, \hat{c_t}) \theta R - 1=0 \end{equation} and \begin{equation} \label{e18} -f_t(\hat{c_p}, \tilde{c_t}) (1-\theta)R - 1=0. \end{equation} One might think that since the two parties are each bearing only a fraction of the loss from error, each exerts less than the socially optimal level of care. This is not true, because Rules 1 and 2 are just extreme forms of Rule 3, with $\theta=1$ and $\theta=0$, yet they do not always result in low care. Consider the case where the care levels of the two parties are substitutes. If the third party would exert very little care in the first-best, but must bear most of the cost when an error occurs, he will increase his level of care, while the principal reduces his care because he bears so little liability. Though we cannot predict its direction, choice of care is distorted because the presence of $\theta$ makes the first-order-conditions differ from those in the social optimization problem, equations (\ref{e6}) and (\ref{e7}). This is a variant of the ``moral hazard in teams'' problem of Holmstrom (1982). His Theorem 1 implies that any liability rule will be inefficient unless it does one of two things: $\bullet$ The rule destroys value by imposing punishments on the principal and the third party following agent error that lead to a total cost to them of more than $R$. This is impractical in the context of contract law because the two parties to the contract would be unwilling to go to court if they anticipated being punished there. In Holmstrom's original context of labor contracts, this approach takes the form of paying the workers zero wages if their joint output is so low as to show that at least one worker shirked. Workers will agree to this in advance knowing that in equilibrium nobody will shirk under this threat, and if someone did, the employer would gladly carry out the threat. In the context of an erroneous contract, the court would impose the entire loss on each partiy. Each would then provide the efficient care level, but they would have an actual disincentive to enforce the agreement in court. $\bullet$ The rule makes use of the care levels $c_p$ and $c_t$ in allocating liability. This is what often happens in tort law: liability is allocated depending on which parties were negligent, and Rule 4 is of this type. \underline{{\it Rule 4. An Erroneous Agreement is Enforced Unless the Third Party was Careless.}} {\it (Strict Liability of the Principal, with a Defense of Third Party Negligence).} Under this rule, the principal is liable under the contract unless the third party took too little care. In equilibrium, the principal maximizes \begin{equation} \label{e19} \pi_p= (1-f(c_p, c_t^*)) (V_p-P) + f(c_p, c_t^*)(V_{p}-R -P) - c_{a}- c_p, \end{equation} and the third party maximizes \begin{equation} \label{e20} \begin{array}{ll} \pi_t &= P- V_t - c_t \;\; {\rm if}\;\; c_t \geq c_t^*\\ & = (1-f(c_p^*, c_t )) (P- V_t) + f(c_p^*, c_t)(P -R-V_t) - c_t \;\; {\rm if}\;\; c_t < c_t^*.\\ \end{array} \end{equation} The first order condition for the principal is \begin{equation} \label{e21} -f_p(c_p, c_t^*) R -1 = 0, \end{equation} which yields $c_p= c_p^*$. The third party will choose either $c_t^*$, for a payoff of $P- V_t - c_t^*$, or a lower care level that solves the first order condition \begin{equation} \label{e22} -f_t(c_p^*, c_t)R -1 =0. \end{equation} This first order condition, however, also has $c_t^*$ as a solution. Thus, it is an equilibrium for both the principal and the third party to choose the first-best care level. Any rule which imposes liability on a negligent party will result in an equilibrium with efficient care by both parties in this model, regardless of how liability is apportioned when no party is negligent.\footnote{ A variety of other tort-like legal rules condition on negligence, like Rule 4, but differ in how they allocate liability due to non-negligent error. The analog of ``simple negligence with a defense of contributory negligence'' would be for the principal to be bound by the contract only if he failed to exert the efficient level of care to avoid agent error, but even then to be released from the contract if the third party also failed to exert care. The analog of ``comparative negligence'' would be for the principal to be bound by the contract only if he failed to exert the efficient level of care, but for the parties to share the costs if both failed to exert care. } Contract law, however, unlike tort law, rarely looks to the level of care of each party, or splits damages between the parties.\footnote{ Posner (1986, p. 165) suggests that unlike in tort, in contract it is usually very clear that only one party could have prevented breach. The performer failed to perform, which the payer could not prevent, or the payer failed to pay, which the performer could not prevent. An apparent exception to the indivisibility of liability is the rule of {\it Hadley v. Baxendale} under which the breacher is not liable for unforeseeable damages. ({\it Hadley v. Baxendale}, 9 Ex. 341, 156 Eng. Rep. 145 (1854)) Since he does pay compensation only for the immediate damages from breach, this splits total damages. It allocates each type of damage completely to one party or the other, however, and the unforeseeability of the uncompensated damages from the point of view of the breacher means that it would be inefficient to make him liable for them. } If courts are unwilling to consider care levels or split damages, the legal rule must be strict liability, and we are left with a choice between Rules 1 and 2. The social surpluses under these two rules are: \begin{equation} \label{e25} Surplus (Rule\; 1: Principal \; liable)= V_p- V_t - f(\tilde{c_p}, 0) R - c_{a} - \tilde{c_p}, \end{equation} and \begin{equation} \label{e26} Surplus (Rule \;2: Third\; Party \; liable)= V_p- V_t - f(0, \tilde{c_t}) R - c_{a} - \tilde{c_t}. \end{equation} Rule 1 is preferred if \begin{equation} \label{e27} f(\tilde{c_p}, 0 ) R + \tilde{c_p} < f(0, \tilde{c_t}) R + \tilde{c_t}. \end{equation} This is what I will call the least-cost-avoider principle. The easiest interpretation is when the equilibrium probability of agent error is the same under Rules 1 and 2, so $f(0, \tilde{c_p}) =f(0, \tilde{c_t})$. In that case, we need only ask whether $\tilde{c_p} < \tilde{c_t}$; does the principal have the least cost of preventing agent error? More generally, if the equilibrium probabilities of error are different, it may be that even if $\tilde{c_p} > \tilde{c_t}$, the principal should be liable because he will prevent more errors. Suppose, for example, that the principal's care consists of double-checking every contract the agent makes immediately, while the third party's care consists of trying to guess whether the agent has forged his authorization letter. This form of third party care may be so costly that the efficient level of it is very low, resulting in a low $\tilde{c_t}$ but also a low $f(0, \tilde{c_t})$. Even if courts are constrained to use rules of strict liability, they can allow some flexibility by thinking of the error probability $f(c_p, c_t)$ as being the sum of the probabilities of $N$ different kinds of errors, each with their own kind of preventive care and their own choice between legal rules. The transaction cost is then \begin{equation} \label{e28} c_{a} + \sum_{i=1}^N \left( f_i(c_{pi}, c_{ti})R + c_{pi}+ c_{ti} \right). \end{equation} For each of the $N$ mistakes, the court must determine where to put liability. This is the approach that will be taken in the cases in the following section: for each situation, try to determine whether the contract should be enforced or not depending on the parties' relative cost of and effectiveness in preventing the particular type of agent error. In accordance with inequality (\ref{e27}), the agreement should be binding after error $i$ if and only if \begin{equation} \label{e29} f_i(\tilde{c_{pi}}, 0 ) R + \tilde{c_{pi}} < f(0, \tilde{c_{ti}}) R + \tilde{c_{ti}}. \end{equation} An entirely different way to avoid harm is for the principal to drop the agent and deal directly with the third party. The social surplus would then be $V_p - V_t - c_{na}$, as in equation (\ref{e1}). If Rule 1 is used, the principal may decide to dispense with the agent altogether rather than risk the agent making a mistake. Abandoning use of the agent could even be the efficient outcome, if optimal care levels by both principal and third party still result in high agency costs.\footnote{The point that optimal deterrence should consider the activity level as well as the degree of the care in the activity is due to Shavell (1980), who made it in the context of torts. } The same outcome could occur under Rule 2. If the third party must bear the cost of agent mistakes, he may decide to refuse to deal with agents and insist on direct negotiation with the principal. In looking at the illustrations in the next two sections, it will be seen that refusal to deal with an agent in dubious circumstances is often the most efficient means for the third party to prevent mistaken contracts. For both principal and third party, the least-cost way of avoiding agent error may be to dispense with the agent. In sum, the legal rule dealing with the consequences of agreements mistakenly entered into by agents cannot attain first-best levels of care by the principal and the third party unless the legal rules assigns liability based on those care levels. Since contract law principles generally do not allow this, some inefficiency will result, with too little effort by one party and perhaps too little or too much by the other. If different types of agent error are prevented by different kinds of care, however, the legal rule can be improved by placing the liability on whichever party is the least-cost-avoider for the particular error. \bigskip \noindent {\bf 4. Application of the Least-Cost-Avoider Principle } \noindent {\it 4.1 The Sources of Authority} The six sources of agent authority described in Section 2 can be derived from the least-cost-avoider principle. When the agent has actual express authority, as in Illustration 1, the principal will usually be the least-cost-avoider. \begin{small}\begin{quotation} {\it Illustration 1: Actual Express Authority.} $P$ hires $A$ to buy goods from $T$. $A$ orders the goods from $T$. Soon afterwards, $P$ realizes that he cannot make use of the goods. Must $P$ accept delivery and pay $T$? \end{quotation} \end{small} The role of the agent is trivial: he carries out the wishes of the principal to the letter, and when the principal is mistaken {\it ex post}, it is not because he employed an agent. It is a useful starting point, however, to explain why $P$ should be bound by the agreement, because both $P$ and $T$ could conceivably have prevented the harm. $P$ is the natural least-cost-avoider because he should know better than anyone what goods he wants and how much he is willing to pay for them. Determining even one's own tastes is costly, but trying to determine someone else's is usually more difficult. This is particularly true when an agent is used, because $T$ does not even meet $P$ face to face. Even so simple a point, however, has its exceptions. Although $P$ knows his tastes best, if $T$ knows the goods he is selling best, it may be $T$ who can most cheaply avoid mistaken contracts. This is the basis for the formation defenses of fraud and misrepresentation which a defendant may use when accused of breach of contract. It is also the basis for many common business practices such as warranties and explicit or implicit money-back guarantees. If the legal default is that $P$ is responsible, but $T$ is the least-cost-avoider in transactions of a particular type, the contract can be modified to shift the risk of mistake onto $T$. In Illustration 1, the considerations involved are very little different from when a contract is made without an agent, except that $T$'s cost of discovering the tastes of the party with whom he is ultimately contracting are even higher. When actual implied authority is involved, the situation is more special to agency, because it creates the possibility that the agent acts in a way that the principal did not intend and would repudiate immediately. \begin{small}\begin{quotation} {\it Illustration 2: Actual Implied Authority.} $P$ hires $A$ as manager for his grocery store. $A$ orders fresh fruit from $T$, but orders a much greater quantity than $P$ desired. Is $P$ liable on the contract? \end{quotation} \end{small} $P$ is liable on the contract because general managers have the implied authority to place orders for fruit. This is a more difficult case than Illustration 1, however. We cannot just say that the buyer is responsible for his contracts, because in Illustration 2, $A$ is acting contrary to $P$'s wishes. Unlike in Illustration 1, if $A$ were not employed, the mistaken order would not have been made, so $A$ is not simply acting in place of $P$. $P$ has, however, intended for $A$ to make contracts of this kind, and knows there is some possibility that $A$ will place the wrong order, and in this sense $P$ intended that $A$ place a certain number of wrong orders. This was the tradeoff between reducing transaction costs and increasing error modelled in Section 3. Viewed this way, actual implied authority is close to inherent agency power: the principal has put the agent in a position where the agent can inflict harm on third parties. The least-cost-avoider principle helps clear away some of the confusion over authority. We must ask what general rule of liability would result in the least cost of avoiding this kind of mistake. The principal has a variety of means to reduce the risk of agent mistakes. He hires the agent, and so can select an agent with the appropriate talents and negotiate a contract to give him incentive to use those talents properly. He instructs the agent to a greater or lesser extent, choosing the level of detail in light of the cost. He can expressly instruct the agent not to take certain actions, and tell third parties about the restrictions.\footnote{ If not communicated to the third party, such withdrawal may not affect the manager's {\it power} to take such actions, because he may still have apparent authority, but the principal can punish the agent for taking forbidden actions, and this is the chief use of hidden instructions of which the third party is unaware, such as those in {\it Watteau v. Fenwick}.} He can monitor the agent, asking for progress reports on or randomly checking transactions that are in progress. The principal's control over the agent, a basic feature of the agency relationship, gives him many levers with which to reduce the probability of mistakes. Analogs of some of these levers are available to the third party. He can refuse to deal with with an insufficiently talented agent, but this requires incurring costs to gauge the agent's talent. Although this may be no more difficult for the third party than for the principal, the principal often will be able to spread the fixed cost of testing the agent over many transactions the agent will make for him. Similarly, the third party could sign a contract with the agent, specifying that the agent will be punished if the principal disavows the contract, but economies of scale are more available to the principal than to the third party. Or the third party could monitor the agent by checking with the principal to confirm the transaction. Sometimes this may be efficient, but the trouble to which it puts the principal reduces the value of hiring an agent. In light of this, it makes sense to have the principal liable as a general rule when the agent has actual implied authority. The next category, apparent authority, rests on manifestations made by the principal which reasonably lead the third party to believe that the agent is acting on the principal's behalf. \begin{small}\begin{quotation} {\it Illustration 3: Apparent Authority.} ``P discharges A, his general selling agent, but gives no notice to others. Shortly thereafter, A contracts to sell goods by separate contracts to X, Y and Z. X had known of A's former employment. A showed Y a letter from P to A signed by P, stating that A is employed as P's selling agent. Z had not known of A's employment and relied wholly on A's oral statement.''\footnote{{\it Restatement}, \S 159. } Do $X, Y,$ and $Z$ have valid contract claims against $P$? \end{quotation} \end{small} The {\it Restatement} says that $X$ and $Y$ have valid claims, but $Z$ does not. This corresponds to the least-cost-avoider principle, because for $X$ and $Y$ to check the authority of every well-known or documented agent for every transaction would be more costly than to require $P$ to notify his customers that $A$ has been fired and to demand return of letters of authority.\footnote{This is close to the {\it former dealer} rule in partnership law, that when a partnership is dissolved, an erstwhile partner still has the power to obligate the firm until former dealers on credit are notified of the dissolution. See Steffen (1977), p. 51 or U.P.A. \S 35.} Furthermore, even if the cost to $P$ of retrieving his letter of authority is high, imposing the liability on $P$ discourages him from issuing such a dangerous letter in the first place. Some other sign of authority may be less easily abused. $Z$, on the other hand, has exerted no care to determine that $A$ is $P$'s agent, and $P$'s cost of advertising to every potential customer that every potential salesman is not his agent is prohibitively high. The principle of estoppel is even more clearly related to the least-cost-avoider principle. Estoppel is based on the ease with which someone could have prevented harm to himself; having failed to prevent the harm, he is ``estopped'' from asserting what would otherwise be a valid claim. \begin{small}\begin{quotation} {\it Illustration 4: Estoppel.} ``P learns that A, who has no authority or apparent authority to sell P's goods, is negotiating with T as P's agent for their sale. He does nothing although he could easily notify T. T pays A for the goods, as is customary in such a transaction.''\footnote{{\it Restatement}, \S 8B. } Must $P$ deliver the goods? \end{quotation} \end{small} On pure agency grounds, $P$ would be free to refuse to deliver the goods, because $A$ is not his agent and $P$'s manifestations did not deceive $T$. Instead, the law requires him to deliver the goods on grounds of estoppel: he could have prevented the misunderstanding but did not. Even though $P$ did not hire $A$, he is the least-cost avoider because he can prevent the mistake more cheaply than $T$ can. He is therefore liable for the resulting harm. Estoppel has two special features in contrast to other sources of liability.\footnote{See Steffen (1977), p. 128.} First, the third party's recovery is limited to the losses caused by the principal's failure to prevent the mistake. Strictly speaking, that failure does not make an invalid agreement into a contract; it just makes the principal liable for damages. Second, estoppel is a one-way street. The third party can obtain damages from the principal, but the principal cannot enforce the agreement against the third party (unless it is made valid by ratification, as discussed below). These features make it even clearer that the driving idea is that of the least-cost-avoider, rather than some idea from contract idea. Ratification, the fifth source of liability, occurs when the principal assents to an agreement after it is made by someone who lacks authority. Ratification is similar to actual express authority, because when the principal ratifies the agreement, he is saying that it is satisfactory to him and he sees no mistake worth the cost of renegotiation. The principal will generally be the least-cost-avoider for the same reason as when the agent has actual express authority. The last source of liability, inherent agency power, is illustrated by {\it Watteau v. Fenwick}, here abstracted as Illustration 5. \begin{small}\begin{quotation} {\it Illustration 5: Inherent Agency Power.} P buys A's tavern and hires him as manager, instructing him not to buy cigars for the tavern. A orders cigars for the tavern anyway, from T, who believes A to be the owner. Must P honor the agreement? \end{quotation} \end{small} $A$ lacks actual authority to buy cigars, because of $P$'s instructions, and lacks apparent authority because $T$ does not believe $A$ to be $P$'s agent. $P$'s liability is based on inherent agency power: $A$ has been put into a situation, where he can impose losses on third parties unless $P$ is made liable. The least-cost-avoider principle reaches the same result, because $P$ can control $A$ more cheaply than can $T$. Not only do the arguments made earlier for $P$'s lower cost of mistake avoidance apply, but the fact that $P$ is undisclosed strengthens the argument because $T$ does not even realize there is an agency problem. Thus, every one of the six sources of liability can be justified by the least-cost-avoider principle. \bigskip \noindent {\it 4.2 When Should the Third Party Bear the Cost of Mistaken Contracts?} As Section 3 pointed out, the transaction benefits both principal and third party. Since both can take care to avoid agent error, which party is liable should depend on the type of error. In almost all of the illustrations in Section 4.1, the principal was the least-cost-avoider of error. Since the principal has more control over the identity and incentives of the agent, it is usually efficient that hebe liable. ``Usually'' is not ``invariably,'' however, as the illustrations in this section will show. The third party does have the advantage of being present at the time of contracting and he does have control over actions of his own which might lead the agent to make an inefficient contract. In Illustrations 6 through 9, the least-cost-avoider principle leaves the principal free from liability for the agent's contracts. Let us begin with the care to check authority, in Illustration 6. \begin{small}\begin{quotation} {\it Illustration 6: Not Checking Authority Carefully.} ``P tells T that A is authorized to buy sheep for him when the market price of wool in another country has reached a certain point. T sells sheep to A, relying upon A's untruthful statement that the price of wool has reached the specified point.''\footnote{{\it Restatement}, \S 168. } Is $P$ bound by the contract? \end{quotation} \end{small} In Illustration 6, $T$ is the least-cost-avoider, because his cost of verifying $A$'s claim is lower than the cost to $P$ of ensuring that $A$ does not make false claims. The {\it Restatement} agrees and $P$ is not bound by the contract. $P$ does have means to prevent the false claims--- through threat of discharge, if nothing else--- but the cost for $T$ to check the claims is lower.\footnote{A case like Illustration 6 is Mussey v. Beecher (1849) 57 Mass. (3 Cush.) 511. The defendant authorized his agent to buy up to \$2,000 in books for his store. The agent bought more than that amount, and ordered even more from the plaintiff, falsely telling the plaintiff that the limit was not yet exceeded. Judge Shaw ruled for the defendant. } The outcome would be different if it were more costly for $T$ to discover the information on which $A$'s authority relies. We have already seen in {\it Watteau v. Fenwick} that secret instructions do not protect $P$. This is equally true of information known to $A$ but not to $P$ or $T$. If $P$ authorizes $A$ to buy one sheep, and tells $T$ that $A$ may be coming by to do so, then $P$ is bound by $A$'s purchase from $T$ even if $A$ had previously terminated his authority by buying a different sheep from someone else.\footnote{{\it Restatement}, \S 171. } $T$ cannot easily discover the information in these cases. A second reason why $T$ may be the least-cost-avoider is that he is on the spot at the time of contracting and can observe the agent's behavior. \begin{small}\begin{quotation} {\it Illustration 7: Incapacity of the Agent.} $P$ hires $A$ as an agent to buy goods from $T$. While intoxicated, $A$ orders the wrong goods from $T$, who knows that $A$ is intoxicated. Must $P$ pay for the goods? \end{quotation} \end{small} $P$ should not have to pay for the goods, because it is $T$ who has the lowest cost of monitoring $A$'s sobriety at the time of contracting.\footnote{Lack of capacity is a standard formation defense for breach of contract. See {\it Restatement (Second) of Contracts}, \S 18. } If $T$ has no way of knowing that $A$ is intoxicated, the answer is more difficult.\footnote{{\it Restatement}, \S 122 seems to say that the contract remains invalid: the agent's authority to contract ends after an event which ``deprives the agent of capacity to make the principal a party to it''.} Contract law does not premise lack of capacity on the knowledge of the other party of that lack. Here, however, $P$ stands behind $A$, and can take measures to prevent him from contracting while intoxicated. If $P$ has hired a habitual drunkard as his agent, $P$ surely should bear the loss from $A$'s frivolous agreements. A third reason to release the principal from liability is if the third party colludes with the agent against him. \begin{small}\begin{quotation} {\it Illustration 8: Collusion With the Agent.} ``T sells a horse to A, P's authorized agent. T represents the horse to be sound. A knows the horse to be unsound. P does not have this knowledge.''\footnote{{\it Restatement}, \S 144. } Is $P$ bound by the contract? \end{quotation} \end{small} The {\it Restatement} bases the result in Illustration 8 on whether $T$ and $A$ have colluded. Note first that $T$'s misrepresentation is not the cause of the erroneous purchase, which it would have been had he been selling directly to $P$, because $A$ was not fooled by it. Therefore, if there was no collusion between $T$ and $A$, $P$ would be bound by the contract; he could have gone to the trouble to hire a more responsible agent than $A$. If $T$ had been colluding with $A$, on the other hand, the contract would be invalid, because $T$ could have prevented the harm by refraining from colluding with $A$, a very low cost to $T$. A fourth reason to release the principal from liability is if the agent's malfeasance should be obvious to the third party. \begin{small}\begin{quotation} {\it Illustration 9: Obvious Agent Malfeasance.} $P$ authorizes $A$ to buy a refrigerator for him. $T$ has listed a refrigerator for sale at \$400, but $A$ offers \$700 if it can be delivered slightly more quickly. $T$ knows that $A$ is an agent for $P$ and that $P$ would not value speed of delivery at \$300. Is $P$ bound by the agreement? \end{quotation} \end{small} According to the {\it Restatement}, ``Unless otherwise agreed, authority to buy or sell with no price specified in terms includes authority to buy or sell at the market price if any; otherwise at a reasonable price.''\footnote{{\it Restatement}, \S 61.} This implies that $P$ is not bound by the agreement in Illustration 9. $T$ has the least cost of controlling the agent's misbehavior because he is on the spot, unlike $P$. The price of \$700 is so excessive that it should be easy for $T$ to see that $A$ is misrepresenting his authority from $P$. If the price paid had only been \$450, the mistake would not be so clear. An extra \$50 is not an unreasonable premium for speed, and it is not easy for $T$ to tell that $A$ is misbehaving. The cost to $P$ of preventing overpayment, on the other hand, is little different whether the amount is \$50 or \$300. The least-cost-avoider principle suggests that $P$ should be liable for small mistakes but not for large mistakes. Illustration 9 also raises the issue of the level of damages that $P$ must pay to $T$ for breaching the contract. The usual remedy the law provides is expectation damages: if the principal breaches, he must pay enough to make the third party as well off as if the contract had not been breached. An alternative is reliance damages: if the principal breaches, he must pay enough to make the third party as well off as if the contract had never been written. Expectation damages induce efficient breach by the promisor, but reliance damages induce efficient reliance by the non-breaching party.\footnote{See chapter 5 of Polinsky (1989) for a discussion of these contract remedies.} In the model of Section 3, those issues did not arise, because the model was constructed so that breach of contract would take place when efficient and the reliance expenditure was made exogenous. In Illustration 9, if the contract price is \$700 and the market price is \$400, the expectation damages would be \$300. Reliance damages, on the other hand, would be zero if $T$ incurred no cost in making the contract, and very little more if $T$'s loss were limited to the transaction cost. The social cost of making and breaching the inefficient contract is limited to the transaction cost, since the \$300 is just a transfer, a windfall profit for $T$. Reliance damages are more efficient than expectation damages, because expectation damages would lead to excessive care by $P$.\footnote{This point is made in Rasmusen \& Ayres (1993) in the context of scrivener's errors. If contracts with such errors are enforced to the letter, parties making contracts will take overly great precautions to avoid the errors; it would be more efficient simply to reform the contract once the error was discovered. } This suggests that even if the court finds $P$ liable for a contract made by an agent, reliance damages may be more appropriate, especially if $P$ announces his intention to breach immediately upon discovering what $A$ has done. \bigskip What this variety of illustrations has shown is that the least-cost-avoider principle can be useful in guiding agency law. The next section will continue to use the principle, but in a special context--- the undisclosed principal problem. \pagebreak \bigskip \noindent {\bf 5. The Undisclosed Principal Problem and Other Issues} This section will address the basic problem of the undisclosed principal and use that context to address a number of other issues in agency law. \bigskip \noindent {\it 5.1 The Undisclosed Principal Problem} The undisclosed principal problem arises when an agent makes an agreement with a third party who does not realize that the agent is acting as an agent rather than on his own behalf. The question then arises of whether the third party has a legal claim against the principal as well as against the agent. {\it Watteau v. Fenwick} is a particularly dramatic example because the agent acted against the express wishes of the principal, but the problem exists even if the agent is obedient, as in Illustration 10. \begin{small}\begin{quotation} {\it Illustration 10: The Undisclosed Principal Problem.} A agrees to buy goods from T. A represents P, unknown to T. Is P bound by the contract? \end{quotation} \end{small} Conventional contract law says that $A$ and $T$ are bound by the contract. Agency law says that $P$ is also bound, but this is s difficult to justify under the usual jurisprudential theories of contract law.\footnote{This discussion is drawn from Barnett (1987), who provides references to the case law and evidence of the discomfort of common law scholars with the undisclosed principal rule. Barnett briefly discusses efficiency theories of contract, but objects to them as not providing a normative theory of contractual obligation.} The rule seems to violate the standard will theory of contract. As the {\it Restatement} says, \begin{small}\begin{quotation} The undisclosed principal rule appears to violate basic principles of contract law. The relation between principal and a person with whom the agent has made an authorized contract is spoken of as contractual, although by definition there has been no manifestation of consent by the third person to the principal or by the principal to him.\footnote{ {\it Restatement}, \S 186.} \end{quotation} \end{small} The rule is no better explained by the bargain theory or reliance theories of contract. The bargain theory looks to whether the two parties bargained with each other. $T$ did not knowingly bargain with $P$, and making $P$ liable provides $T$ with a benefit for which he did not bargain. The reliance theory looks to whether the two parties reasonably rely on each other's promises. $T$ does not rely on $P$'s promises, because he does not know that $P$ exists.\footnote{Barnett (1987) shows that his own ``consent theory'' of contract does explain the common law rules of the undisclosed principal problem. This theory looks to whether (a) the subject of the contract is a morally cognizable and alienable right owned by the transferor and (b) the transferor manifests his consent to transfer the right.} An efficiency theory of contract law has an easier time explaining the undisclosed principal rule. The least-cost-avoider principle says that $P$ should be bound because $T$'s cost of preventing $A$ from making an inefficient contract is greater than $P$'s. $T$, who does not even know that $A$ is an agent, should be allowed to take even less care than if he knew $A$ were an agent. Consider again Illustration 9, in which $A$ pays \$700 for a refrigerator that normally costs \$400. The {\it Restatement} seems to indicate that $P$ is released from obligation regardless of whether he is disclosed or undisclosed. The least-cost-avoider principle suggests that the court should think harder about releasing $P$ if he is undisclosed, because $T$ has less reason to be suspicious and to investigate. Without knowing that $A$ is an agent, $T$ has no reason to doubt that $A$ has some personal reason for paying a large premium for speedy delivery. The rule that undisclosed principals are bound by their agents' contracts is also useful because it encompasses situations in which the agent has used the resources of the principal to appear wealthier or more dependable. {\it Watteau v. Fenwick} is one such case: the manager seemed more creditworthy because he appeared to own the tavern. Indeed, the very act of placing an order may make an agent seem more dependable. If an individual orders ten thousand dollars worth of replacement parts for a nuclear reactor, it is natural to suppose that he is acting on behalf of a larger business. The main objection to making $P$ liable is that it provides a windfall to $T$, because $T$ did not rely on $P$'s credit when entering into the contract. If $T$ has the option of enforcing the contract against either $A$ or $P$, he has gained an advantage for which he had not bargained with $A$: if $A$ is insolvent, $T$ can sue $P$ instead. This windfall matters not only for fairness but efficiency, as Illustration 11 shows. \begin{small}\begin{quotation} {\it Illustration 11$'$: Windfalls$'$.} $A$, who owns no assets, orders goods on credit from $T$ which cost \$90 to produce. If there is no recession, $A$ can resell the goods for \$105, while if there is a recession, which occurs with a 10 percent probability, the goods are worth \$0. The contract takes the form of a contract price of \$101, and $A$ breaches if a recession occurs. \end{quotation} \end{small} \begin{small}\begin{quotation} {\it Illustration 11$''$: Windfalls$''$I.} The same as 11$'$, except: $A$ is backed by undisclosed principal $P$, who must honor the contract and will not become insolvent in doing so. $T$ believes the probability that $A$ is backed by an undisclosed principal is negligible and so still insists on a contract price of \$101. A contract price of \$101, would yield negative profits for $P$, so no agreement is reached. \end{quotation} \end{small} \begin{small}\begin{quotation} {\it Illustration 11$'''$: Windfalls$'''$.} The same as 11$''$, except: The legal rule is that $P$ is not a party to the contract, and so need not pay damages if $A$ breaches. A contract price of \$101 would yield positive profits for $P$, and so agreement is reached. \end{quotation} \end{small} In each of the variants of Illustration 11, it is efficient for $T$ to sell the goods to $A$, but in 11$''$, in which the undisclosed principal is bound, agreement is not reached. Thus, there is indeed a potential efficiency loss associated with making undisclosed principals liable on contracts. Some mutually beneficial contracts will not be made because $T$ does not know the value of the contract being offered to him. Whether because of the windfall problem or for some other reason, there is one class of undisclosed principal cases where American courts have usually exempted the principal from liability. This class is described in Illustration 12. \begin{small}\begin{quotation} {\it Illustration 12: Principal Pays but Agent Breaches.} $A$ agrees to buy goods from $T$, and $T$ delivers the goods. Unknown to $T$, $A$ is an agent for $P$. $P$ pays $A$ the money for the goods, but $A$ becomes insolvent before paying $T$. Can $T$ sue $P$ for payment? \end{quotation} \end{small} Most American courts would deny $T$ the right to sue $P$, on the grounds that $T$ relied only on $A$'s credit, but English courts, and the {\it Restatement}, would give $T$ the right to sue.\footnote{See Barnett (1987), pp. 1973, 1984 and the {\it Restatement} \S 208.} The least-cost-avoider principle leads to the English rule. $P$ has the least cost of inducing $A$ to make an efficient contract, and also has the least cost of preventing $A$ from losing the payment money before transferring it to $T$. \bigskip \noindent {\it 5.2 Special Contract Terms} One of the maxims of the law and economics of contract is that if transaction costs are low, the parties will customize their contract regardless of the legal default rule, so an inefficient default rule can cause only a limited amount of harm. This might seem to be a way out of the undisclosed principal problem, as in Illustration 13. \begin{small}\begin{quotation} {\it Illustration 13: Special Contract Terms.} $A$ agrees to buy goods from $T$, but the contract specifically excludes any liability on the part of anyone but $A$ and $T$. $A$ represents $P$, unknown to $T$. Is $P$ bound by the contract? \end{quotation} \end{small} $P$ has no obligation to $T$ here,\footnote{ {\it Restatement}, \S 189. } but it may be difficult to write a contract of this kind because asymmetric information creates serious trouble for this contract maxim.\footnote{A number of articles have appeared in recent years showing how parties may be reluctant to propose special contractual terms, or to object to adding terms, because such actions would reveal their private information. See Ayres \& Gertner (1989) and Spier (1992).} If $A$ proposes a term exempting other parties from liability, $T$ can deduce that $A$ is backed by a principal, which is information that $A$ may not wish to disclose. Hence, if most agents wished to exempt their principals, it would be important to make this exemption the default rule, or perhaps even to make it mandatory. \bigskip \noindent {\it 5.3 An Agent Acting Outside of His Authority} Section 4.1 discussed why the principal should be liable when the agent has various categories of authority. Why should the principal not be liable when the agent acts without authority? The question may seem vacuous, but the principal might have the least cost of preventing mistakes by the agent even when the mistakes are unconnected with the principal's purpose in hiring him. Illustration 14 is an example. \begin{small}\begin{quotation} {\it Illustration 14: An Agent Acting on his Own Behalf. } ``$P$ authorizes $A$ to purchase a particular horse in $A$'s name and gives $A$ the money to do so. $A$ purchases the horse on his own credit, without disclosing $P$'s existence, intending to abscond with horse and money, which he subsequently does.''\footnote{{\it Restatement}, \S 199.} Is $P$ liable to $T$, the seller of the horse? \end{quotation} \end{small} According to the {\it Restatement}, $P$ is not liable, because $A$ is not acting as his agent. If $P$ were a disclosed principal and $A$ bought the horse on $P$'s credit, $A$'s motivation would not matter and $P$ would be liable. This makes sense because $P$ can control $A$ at lower cost than $T$. Since $P$ is undisclosed, Illustration 14 seems hard to distinguish from the ordinary undisclosed principal problem in Illustration 10. In both, the issue is whether $P$ should be liable for the mistaken contract when $T$ does not know of $P$'s existence. Why should $A$'s motivation matter? The answer may lie in providing efficient incentives for $P$ to decide whether to hire $A$. In Illustration 10, where $A$ is acting on behalf of $P$, the contract would not have been made had $P$ not set the process in motion. If contracting by means of an agent sometimes leads to harm, $P$ ought to bear the cost, or he will be too willing to use an agent. Illustration 14 is different because $P$'s hiring of $A$ in no way caused the harm. If $P$ had never hired $A$, $A$ could still have done the same thing, agreeing to buy the horse from $T$ and absconding. That $A$'s action is to abscond with a horse is irrelevant; if $A$ had chosen to abscond with $T$'s bicycle instead, the problem would be essentially the same. Since the harm does not result from $P$'s desire to buy a horse using an agent, it would be inefficient to impose an extra cost on that activity by making $P$ liable for $A$'s misdeeds. The point that the principal should not be liable for harm caused by an agent acting without authority can be applied more generally. An argument could be made using the least-cost-avoider principle. Perhaps the principal should be liable for harm caused by the agent even if all of Section 3's six sources of liability are lacking, because the principal can still control the agent by means of his contract. This may be easier to see in the context of tort. The legal rule of vicarious liability is that the principal is liable only for torts that the agent commits in the course of employment.\footnote{{\it Restatement}, \S 219.} Imagine going a step further by making the principal liable even for the agent's holiday torts. The justification would be that the principal can control even those torts, by threatening to fire the agent if they occur. Making the principal liable would be distortionary, however, because it would tend to deter the hiring of agents. The extra liability on the principal is like a tax on hiring agents. Thus, it is not enough to discover that the principal is the least-cost-avoider in controlling agents; one must also decide whether imposing the cost of avoiding the harm might overly deter principals from hiring agents in the first place. \bigskip \noindent {\it 5.4 Negotiable Instruments} A negotiable instrument is a transferrable written promise to pay money. Whether an undisclosed principal is bound by a note signed by his agent depends on whether it is negotiable. Consider Illustrations 15 and 16. \begin{small}\begin{quotation} {\it Illustration 15: A Negotiable Instrument.} ``Chicago, June 1, 1928. On demand I promise to pay to bearer \$100.'' (Signed) ``A.''\footnote{{\it Restatement}, \S 152.} \end{quotation} \end{small} \begin{small}\begin{quotation} {\it Illustration 16: A Nonnegotiable Note.} ``Chicago, June 1, 1928. On demand I promise to pay John Smith \$100.''(Signed) ``A.''\footnote{{\it Restatement}, \S 152.} \end{quotation} \end{small} According to the {\it Restatement}, $A$'s undisclosed principal may be found liable in Illustration 16, but not in Illustration 15. It is easiest to explain why the principal could be liable in Illustration 16. This may just be the payment part of the earlier Illustration 10: the third party, John Smith, has done something for the principal, and the agent pays Smith with the note. The only difference in Illustration 15 is that the promise is to ``the bearer,'' yet the undisclosed principal is not liable. This might be due to the need for transferrability. By asking for a note made out to ``bearer,'' $T$ is showing that he may wish to transfer the note to a fourth party, $F$. The value of the note depends on the creditworthiness of $A$, which may make it difficult for $T$ to receive the full value of the note from $F$. This is a problem of adverse selection; if $T$ tries to resell the note to $F$, $F$ will suspect from that very act that $T$ has heard something unfavorable about $A$'s credit. Anything that increases the possibility that $T$ is better informed than $F$ about the value of the note will make it more difficult for $T$ to transfer the note. If the rule is that $P$ is liable on $A$'s note, the asymmetry of information between $T$ and $F$ will most likely increase. Not only will $F$ fear that $T$ has superior information on $A$'s credit, but also on $P$'s existence and $P$'s credit. Thus, $T$ should prefer to receive a note from $A$ relying only on $A$'s credit (but for a slightly larger face value) instead of a note from $A$ relying on the credit of both $A$ and a possible undisclosed principal. \bigskip \noindent {\it 5.5 Liability of the Third Party to the Principal } A different part of the undisclosed principal problem is the obligation of the third party to the principal. Should $T$ be bound contractually to a party he did not know existed? Illustrations 17 and 18 contrast two situations in which this might arise. \begin{small}\begin{quotation} {\it Illustration 17: Inadvertent Sale to an Enemy.} $T$ agrees to sell goods to $A$, who, unknown to $T$, is an agent for $P$. $T$ and $P$ are bitter enemies, and $T$ would never willingly sell goods to $P$, as $A$ and $P$ know. $T$ later discovers that $A$ is acting for $P$. Is $T$ bound by the agreement? \end{quotation} \end{small} \begin{small}\begin{quotation} {\it Illustration 18: Inadvertent Sale to a Rich Buyer.} $T$ agrees to sell goods to $A$, who, unknown to $T$, is an agent for $P$. $P$ is very rich, and $T$ would have held out for a higher price if he had known $P$ was interested in the goods. $T$ later discovers that $A$ is acting for $P$. Is $T$ bound by the agreement? \end{quotation} \end{small} In both of these illustrations, $T$ would have been willing to sell to $A$ at the specified price, but not to $P$. Where they differ is in $T$'s motivation. In Illustration 17, $T$'s motive is real, rather than redistributive. $T$'s utility from holding the goods himself is greater than the sum of his utility from the sale price to $P$ plus the disutility of knowing that $P$ has bought the goods. The result is analogous to when a purchaser buys a defective product: the utility he expects from the transaction is less than the utility he receives. Thus, there is a strong case for invalidating the sale.\footnote{Steffen (1977), p. 188 says that if the third party has once refused to deal with the principal, the principal may not circumvent the refusal by using a secret agent, according to most courts. When the third party's desires are less objectively manifested, courts are less likely to intervene. An example is Kelly Asphalt Block Co. v. Barber Asphalt Paving Co. (1914) 211 N.Y. 68, 105 N.E. 88, in which Judge Cardozo held for an buyer who had remained an undisclosed principal because he suspected a competitor would refuse to deal with him. } In Illustration 18, on the other hand, $T$'s motive is redistributive. $T$'s utility from holding the goods himself is less than his utility from the sale price to $P$, and he would prefer the agreed sale price to not selling the goods at all. $T$ is dissatisfied, but that is only because he could have received an even higher price if he had known that $P$ was the ultimate buyer. Thus, the transfer of the goods from $T$ to $P$ is allocatively efficient, and there is no reason to invalidate the agreement. \bigskip \noindent {\bf 6. Concluding Remarks} This paper has tried to sort out various aspects of agency law involving the power of agents to bind their principals by contracts with third parties. A unifying theme has been the view of the agent as a means to reduce the transaction cost, a role in which he is useful to both parties in the transaction, the principal and the third party. Issues of liability should be related to providing each of the two transactors with incentive to monitor the agent. After constructing a model of the interaction between principal, agent, and third party, the least-cost-avoider principle was applied to a variety of topics in agency law, including the categories of authority for the agent to act on behalf of the principal, the undisclosed principal problem, bribery, criminal agents, negotiable instruments. Many topics in agency still await explanation. This article has only attempted to analyze the main problems that arise, and scholars will find the {\it Restatement of Agency} a stimulating source of further difficulties that arise when one person acts for another. This analysis may also be useful for analyzing two other areas of the law: partnerships, and criminal conspiracies. In both areas, one person bears responsibility for the actions of another, and the methods used here may be helpful in understanding and improving the law. %-----------------------------%--------------------------------------- --- \newpage \bigskip \noindent {\bf References } American Law Institute (1958) {\it Restatement of the Law, Second: Agency 2d}, St. Paul, Minnesota: American Law Institute, 1958. American Law Institute (1981) {\it Restatement of the Law, Second: Contracts 2d}, St. Paul, Minnesota: American Law Institute, 1981. Ayres, Ian and Robert Gertner (1989) ``Filling Gaps in Incomplete Contracts: An Economic Theory of Default Rules,'' {\it Yale Law Journal}, October 1989, 99: 87-130. Barnett, Randy (1987) ``Squaring Undisclosed Agency Law with Contract Theory,'' {\it California Law Review}, December 1987, 75: 1969-2003. Black, Henry (1990) {\it Black's Law Dictionary}, 6th edition, St. Paul, Minn.: West Pub. Co., 1990. Chu, Cyrus \& Yingyi Qian (1992) ``Vicarious Liability Under a Negligence Rule: An Economic Analysis,'' Stanford Law School, June 1992. Cooter, Robert (1985) ``Unity in Tort, Contract, and Property: The Model of Precaution,'' {\it California Law Review}, January 1985, 73: 1-51. Cooter, Robert \& Bradley Freedman (1991) ``The Fiduciary Relationship: Its Economic Character and Legal Consequences,'' {\it New York University Law Review}, October 1991, 66: 1045-1075. Holmstrom, Bengt (1982) ``Moral Hazard in Teams'' {\it Bell Journal of Economics.} Autumn 1982. 13: 324-40. Klein, William \& J. Mark Ramseyer (1994) {\it Teacher's Manual for Cases and Materials on Business Associations}, Second Edition, Westbury, NY: Foundation Press, 1994. Kornhauser, Lewis (1982) ``An Economic Analysis of the Choice Between Enterprise and Personal Liability for Accidents,'' {\it California Law Review}, December 1982, 70: 1345-1392. Landes, Richard \& Richard Posner (1987) {\it The Economic Structure of Tort Law}, Cambridge, Mass.: Harvard University Press, 1987. Polinsky, A. Mitchell (1989) {\it An Introduction to Law and Economics}, second edition, Boston: Little, Brown and Company, 1989. Posner, Richard (1986) {\it The Economic Analysis of Law}, 3rd edition. Boston: Little, Brown, and Company, 1986. Rasmusen, Eric \& Ian Ayres (1993) ``Mutual Versus Unilateral Mistake in Contracts,'' {\it Journal of Legal Studies, } (June 1993), 22: 309-343. Schiff, Martin (1983) ``The Undisclosed Principal: An Anomaly in the Laws of Agency and Contract,'' {\it Commercial Law Journal}, May 1983, 88: 229-237. Shavell, Steven (1980) ``Strict Liability versus Negligence,'' {\it Journal of Legal Studies}, January 1980, 9: 1-25. Spier, Kathryn (1992) ``Incomplete Contracts and Signalling,'' {\it Rand Journal of Economics}, (Autumn 1992) 23: 432-443. Steffen, Roscoe (1977) {\it Agency-Partnership in a Nutshell}, West Publishing Co., 1977. Sykes, Alan (1984) ``The Economics of Vicarious Liability,'' {\it Yale Law Journal }, June 1984, 93: 1231-1280. Sykes, Alan (1988) ``The Boundaries of Vicarious Liability: An Economic Analysis of the Scope of Employment Rule and Related Legal Doctrines,'' {\it Harvard Law Review}, January 1988, 101: 563-609. % \end{raggedright} \end{document}