The Commission

The agent stands between the client and a bargain the client cannot easily make alone. He may know where to find the other party, what an offer is likely to become, or which terms matter after the signature. The commission pays for using that knowledge on someone else's behalf. It also gives the agent an interest in the result, and the interest need not coincide with the client's. A dealer who buys and resells commits his own balance sheet to the difference between two prices. An agent can influence the transaction while leaving most of its consequences with the person represented. The distinction makes the middleman's position more intimate: the client seeks help in judging the very decision on which the helper's payment depends. This is the problem of incentives within a relationship of agency. A service can be valuable, its provider can be well informed, and the arrangement can still reward a choice the client would not make with the same information.

A share of the price aligns the parties in one direction while leaving them differently exposed at the margin. Consider an illustrative seller whose agent receives two per cent of the sale price. A sale at one hundred thousand pays the agent two thousand and leaves the seller ninety-eight thousand before other costs. If additional work raises the price by ten thousand, the agent receives only two hundred more while the seller receives nine thousand eight hundred more. Suppose that securing the increase costs the agent three hundred in time and expense and the seller five hundred in delay. On those stipulated terms, the seller gains nine thousand three hundred from continuing while the agent loses one hundred. Both prefer a higher price when it arrives without further cost; they differ over the effort and waiting needed to obtain it. The arithmetic does not establish what the next offer will be. It shows why a recommendation to accept can depend on whose additional cost is being compared with whose additional receipt. The conflict becomes consequential when the client relies on the agent to assess the uncertain prospect of waiting.¹

The difference can be studied where the same kind of expert sometimes acts for a client and sometimes sells an asset of his own. Levitt and Syverson examined single-family home sales listed in thirty-four Cook County suburbs from 1992 to 2002. Their principal sample contained about ninety-eight thousand sales, including roughly thirty-three hundred in which the agent had an ownership interest. With detailed controls for property attributes, place and time, their preferred comparison found agent-owned homes selling for about 3.7 per cent more and remaining on the market about 9.5 days longer. The joint pattern was consistent with an information advantage that let agents advise clients to accept earlier than the agents would accept for themselves. It was an observational comparison, and the authors examined alternative explanations, including unmeasured property quality, effort and different costs of waiting. Its force comes from the specified comparison and those checks. It does not establish that every agent undersells a client's property, that the historical magnitude applies elsewhere, or that selling without an agent would produce the same benefit. The study makes an incentive problem visible without measuring away all the services that accompanied it.²

The measure used to pay an agent also determines which part of the client's interest is easiest to neglect. A payment per completed transaction rewards completion, even where another search or a decision to abstain would serve the client. A percentage of expenditure rewards a different margin from a reward for savings. If the provider of one product pays the adviser more than another provider does, the adviser's ranking of the products can acquire a term absent from the client's ranking. These are possibilities created by the payment rule, not a verdict that every recipient follows the rule without regard to reputation or duty. They become especially troublesome when the agent has several tasks and only one is readily measured. Quantity sold can be counted at once; the quality of the advice may emerge much later. Holmström's account of multitasking explains why a stronger reward for the measurable task can draw attention from the rest. A perfectly accurate count of transactions can still be a poor measure of how well the person represented has been served.³

Changing the payment changes the conflict, and can move it to another part of the work. A fixed fee removes the direct reward for increasing the transaction price or volume, while leaving the client with the problem of obtaining adequate effort after the fee is agreed. A payment conditional on a successful conclusion gives the agent a reason to pursue completion but also makes unsuccessful work unpaid. Increasing the agent's share may strengthen the incentive to improve an outcome, while exposing the agent to more of the chance that enters it. In the standard model with a risk-averse agent and imperfectly observed effort, inducing effort and insulating income from luck have to be reconciled. The contract cannot be judged by the strength of its incentive alone. The Tenant made the same issue visible in the division of a harvest. In a brokerage relationship, the price and timing of a sale similarly depend on more than the agent's diligence. Making the agent the outright owner would alter that exposure again, bringing financing and inventory costs into an arrangement that previously delegated advice and execution.⁴

The quality of delegation also depends on what the client can learn about the agent's choices. A record of offers received, an explanation of alternatives rejected, or a comparison that separates common market movements from the agent's contribution can make a payment more informative. Holmström's informativeness principle asks whether an additional signal reveals something about the action beyond what the outcome already reveals; mere abundance of data is not the criterion. In practice, obtaining and interpreting such evidence also costs resources. Disclosure of a payment can expose a conflict while leaving the client unable to evaluate the proposed bargain unaided. Reputation can make disappointing a client costly, but its force depends on whether the disappointment is detected and affects future business. These considerations turn the commission into one part of an arrangement: the assignment, the information available, the limits on discretion, the way performance is assessed, and the client's means of seeking an alternative. An announced intention to represent someone does not perform this work of alignment by itself.⁵

The answer to whose side the agent is on lies partly in what the arrangement makes it worthwhile to do, and partly in the judgments and obligations it can make effective. The commission can reward a service the client could not cheaply obtain alone. It can also reward bringing the client's search to an end before the best attainable terms have been found, or directing attention toward the transaction that pays the agent most. Both possibilities belong inside the account of the service. A low fee accompanied by poor advice can cost more than an expensive engagement that materially improves the bargain; a high payment, by itself, proves neither improvement nor abuse. What matters is the completed exchange compared with a feasible alternative, including the cost of finding out how well the representative acted. The volume's closing plate, The Friction, gathers these costs with search, waiting, inventory and information. The price of standing between has led to the cost of making any exchange happen—and to the institutions that a frictionless price quietly leaves outside its frame.