Two people take adjoining seats in the same theatre and discover that they have paid different prices. The seats offer the same view, and admitting one customer has cost no more than admitting the other. One bought a student ticket; the other paid the ordinary charge. The seller has found a way to divide a demand that the preceding plates treated as one. Where the monopolist posted a single price and lost the customers who could not meet it, a discriminating seller tries to charge different buyers different amounts without allowing the cheaper transaction to undo the dearer one. The object is the surplus the uniform price leaves with those who would have paid more. Consumer surplus is not a possession the buyer holds securely; it is a prize the seller is perpetually trying to take. The instruments of that pursuit range from a visible group discount to a menu designed to make buyers disclose themselves.
The limiting case makes the consequences unusually clear. Imagine three buyers, each seeking one unit, whose maximum willingness to pay is twelve, nine, and five. Each unit costs four to supply, with no other costs or effects. A seller confined to one price earns the largest profit by charging nine and serving the first two buyers: eighteen in receipts, eight in costs, and ten in profit. The first buyer retains three of surplus; the second retains none; the third goes without. Now give the seller exact knowledge of all three valuations and the ability to prevent resale. He can charge twelve, nine, and five separately and serve everyone, receiving twenty-six, spending twelve, and keeping fourteen. Assuming the buyers accept a trade at their maximum valuation, none retains surplus. The extra unit creates one additional unit of value, because its buyer values it at five and it costs four. Every beneficial trade is made, and the whole surplus belongs to the seller. The gain in total surplus and the change in its division are both real.1
Pigou distinguished three degrees of discrimination, a division later developed into the modern account of individual prices, menus, and groups. In the first, just described, each unit can be charged at the buyer's own maximum. In the second, the seller offers a common schedule and buyers reveal differences through their choices: the number of units purchased, the package selected, or the conditions they accept. A discount for a larger block can be one part of such a schedule. In the third, an observable classification separates groups to which different prices can be quoted, as a student credential separates one audience from another. The distinctions concern information and the available pricing instrument. A choice from a menu is evidence the seller induces; membership of a group is evidence he observes. Neither gives him perfect knowledge of a person's private value. Nor does every difference in price establish discrimination: different production, service, timing, or delivery costs can also account for different charges. The question is how prices differ once the relevant differences in supply have been allowed for.2
The seller must be able to sustain the separation on which the price difference depends. A customer offered the low price could otherwise buy for a customer facing the high one, retaining a small commission and erasing the expensive segment. Resale restrictions, personal tickets, eligibility checks, or the character of a service may keep that route closed. The seller also needs some discretion over price and sufficient information to form useful groups or schedules. An uninformative classification may lose the very customers it was intended to charge more. The relevant demand response belongs to the segment at the proposed price: under the usual interior conditions, a group less responsive to a price increase bears a larger proportional markup. That group need not be the richer one. Urgency, location, switching costs, and the absence of another supplier can make a buyer relatively insensitive to price even when the payment is burdensome. The information that helps a seller find a high willingness to pay may reveal a valuable service, or a costly inability to leave.3
The welfare of imperfect discrimination turns on which trades it changes as well as on who pays. A lower price in a previously unserved segment can bring in customers whose valuations exceed the cost of supplying them. A higher price elsewhere can drive out other worthwhile purchases. The totals have to be counted together. In the standard comparison with a uniform monopoly price, constant marginal cost, and the relevant markets already served, an increase in total output is necessary for third-degree discrimination to improve total surplus; the increase alone does not guarantee the improvement. Discrimination can also change the allocation of the units among buyers, assigning a unit to someone who values it less while excluding someone in a dearer segment who values it more. The outcome therefore depends on the demands and on the prices chosen. A discount may enlarge access, while a discriminatory system as a whole still reduces surplus. A more profitable system may increase or decrease it. The division of whatever surplus remains is a further question.4
A seller can also separate the price of entering from the price of using what is inside. Walter Oi's analysis of the amusement park makes the distinction concrete. A low charge for each ride can encourage use, while the admission charge collects part of the value that cheap rides create. With identical customers whose demands are known and unaffected by income, pricing each use at marginal cost and charging an admission fee equal to the resulting consumer surplus can reproduce the efficient quantity while transferring that surplus to the proprietor. Different customers complicate the device. Set a large admission fee to capture the enthusiastic visitor's value and the occasional visitor may stay away; reduce it to admit the occasional visitor and the enthusiast keeps some benefit. The proprietor has to weigh participation against extraction, perhaps offering a menu that separates them. A schedule can thus obtain information from what people choose, but every condition that persuades one kind to identify itself also affects the choices available to the others. The customer's response remains part of the constraint the scheme must satisfy.5
The segment brings the volume back to the distinction between making value and deciding who keeps it. The perfectly informed seller can eliminate the trades forgone under a uniform monopoly price while leaving the buyer no gain from any trade that remains. Less perfect sellers face a more tangled account, in which additional access, withdrawn access, information, and the division of surplus all matter. The price difference itself cannot settle the judgment. It can be necessary to ask which customers would otherwise have been served, what each extra unit costs, what separates the groups, and what freedom each customer has to refuse the offered terms. These questions acquire another force when a single network can serve the whole market more cheaply than several rivals could. Then the undertaking has common costs to recover, and the choice of charges helps decide both whether it survives and who can use it. The next plate turns to that network, and to the rules that restrain its price.