Price Discrimination

What price discrimination is, precisely

Price discrimination is the practice of charging different buyers different prices for the same good or service, where the price difference is not explained by a difference in marginal cost. The cost qualifier matters. A delivery firm that charges more to ship to a remote island is not price-discriminating — fuel and labour really do cost more out there. A pharmaceutical company that charges Americans five times what it charges Canadians for an identical pill manufactured in the same plant is discriminating: the cost-of-service is the same in both countries, and the price split exists only because the firm has detected that the two groups have different willingness to pay and that resale across the border is hard.

The economics is straightforward. A firm with any downward-sloping demand curve — a monopolist, a brand with loyal customers, a regional carrier with limited substitutes — faces a tension: a single price either turns away buyers who would have paid less than the chosen price but more than marginal cost (foregone surplus) or sells too cheaply to high-value buyers (foregone profit). Price discrimination resolves the tension by tailoring the price toward each buyer's willingness to pay. The strategies for doing so fall into three families, classified by the British economist Arthur Pigou in his 1920 Economics of Welfare.

Pigou's three degrees

Pigou's taxonomy is the standard scaffold for thinking about every real-world pricing scheme — and it has held up for more than a century because the three categories really do correspond to three distinct information environments for the seller.

First-degree (perfect) price discrimination

The seller knows every buyer's reservation price — the maximum amount each individual would pay — and charges exactly that. The entire area under the demand curve and above marginal cost becomes producer surplus; consumer surplus is zero. Output expands to the competitive level because the firm is willing to sell to every buyer whose willingness to pay exceeds marginal cost, and so the deadweight loss of single-price monopoly vanishes. Pure first-degree pricing is rare because it requires perfect information about individual buyers. Approximations show up in:

• Bespoke negotiation. Bazaar haggling, car dealerships, art galleries, B2B enterprise sales — the seller invests in learning each buyer's reservation price before quoting.
• College tuition with financial aid. The sticker price is high; the financial-aid award reduces it to roughly what the student's family will actually pay, inferred from a detailed disclosure of income, assets, and competing offers. Net tuition is therefore close to each family's reservation price.
• Personalised algorithmic pricing. An e-commerce platform that posts different prices for the same SKU to different visitors based on browsing history, device, location, and inferred income — the modern technology that brings the textbook ideal within reach.

Second-degree price discrimination

The seller cannot tell buyers apart in advance but designs a menu of price-quantity or price-version options such that high-value and low-value buyers each pick the option intended for them — self-selection. The seller never asks "are you a heavy user?"; instead, it offers a small package at a high per-unit price and a large package at a lower per-unit price, and the heavy users reveal themselves by choosing the larger package. Examples:

• Quantity discounts. Buy one for $5, three for $12, twelve for $36 — the per-unit price falls with volume.
• Two-part tariffs. A fixed entry fee plus a per-unit charge: amusement park admission plus per-ride fee, mobile-phone monthly fee plus per-minute charge, warehouse-club membership plus item prices, Costco.
• Versioning. Software sold as basic, pro, and enterprise tiers at progressively higher prices with progressively more features. The marginal cost to ship a feature flag is near zero; the price discrimination is the entire point.
• Airline seat classes. First, business, premium economy, economy — the same physical journey at very different prices, with cabin differentiation acting as the fence.
• Hardback then paperback. The hardback at $30 captures eager readers immediately; the paperback at $15 a year later captures the price-sensitive who were willing to wait.

Third-degree price discrimination

The seller can directly sort buyers into observable groups with different demand elasticities and posts a different price for each group. This is by far the most common variety in practice because it requires only that the seller can verify group membership at the point of sale.

• Student and senior discounts. A student ID or driver's licence is enough to qualify; the underlying logic is that the two groups have flatter demand (more elastic) than the median buyer.
• Matinee versus evening movie tickets. The afternoon audience is more price-elastic (retirees, parents, the unemployed) and gets a lower price; the evening audience pays full freight.
• Peak versus off-peak pricing. Electricity, public transit, hotel rooms, gym memberships — time-of-day, day-of-week, or season-of-year is the sorting variable.
• Geographic price differences. Pharmaceuticals priced higher in the United States than in Canada or the EU. Software licences priced lower in low-income countries.
• Coupons. The coupon itself is the sorting device: the high-willingness-to-pay shopper does not bother to clip; the price-sensitive shopper does.

The three degrees at a glance

Degree	Information	Mechanism	Examples	Consumer surplus extracted
First (perfect)	Each buyer's reservation price known	Individual price quotation	Haggling, bespoke B2B, financial aid, personalised pricing	All
Second (self-selection)	Distribution of types known, individuals unknown	Menu — buyers self-select	Quantity discounts, two-part tariffs, software tiers, airline cabins	Much, not all
Third (group)	Group identity observable	Different posted price per group	Student/senior discounts, matinee, geographic, peak/off-peak, coupons	Some, varies by group

The three conditions for price discrimination to work

None of the three degrees is workable without all three of the following preconditions. Pull any one and discrimination collapses.

1. Market power

A firm in perfect competition has no scope to discriminate — any attempt to charge a buyer above the going price drives that buyer to a rival. Price discrimination therefore presupposes some downward-sloping firm-level demand: monopoly, monopolistic competition, oligopoly with differentiated products, or a temporary niche. The amount of discrimination possible scales with the steepness of that curve. Pure monopolies (patented drugs, utilities, local newspapers historically) can practice aggressive third-degree discrimination; brand-differentiated competitors (Apple, Coca-Cola) can practice a milder version; commodity producers (wheat, sugar at the world market) cannot discriminate at all.

2. Identification of types

The seller must be able to sort buyers by their willingness to pay or to elicit it through self-selection. Sorting devices vary by degree:

• First-degree: direct measurement of each buyer's reservation price (negotiation, detailed financial disclosure, behavioural data).
• Second-degree: design a menu so that buyer choice reveals type — the seller never needs to ask.
• Third-degree: a credential or observable that correlates with elasticity (student ID, age, ZIP code, time of day, device).

3. Prevention of arbitrage

If low-price buyers can resell to high-price buyers, the discrimination unwinds: the seller effectively faces a single price set by the lowest channel. The classical fences against arbitrage are:

• Geography. Shipping a 5-cent pill across an international border is expensive enough — and regulated enough — that few patients do it; the price gap persists.
• Time of consumption. A matinee ticket cannot be saved for the evening showing.
• Personal services. A haircut, a doctor's appointment, a flight — the buyer must be present, so resale is structurally impossible.
• Contractual restrictions. Site licences and ticket-name policies. "Non-transferable" is a fence.
• Bundling and metering. Two-part tariffs and proprietary consumables (printer toner, razor cartridges, game cartridges) tie the high-margin good to a non-resellable platform.

Welfare: ambiguous on total, redistributive in split

The natural reaction to price discrimination is that it must reduce welfare — it sounds unfair. The economics is more interesting.

Compared with a single-price monopolist, first-degree price discrimination is unambiguously efficient on total welfare grounds. The monopolist's deadweight-loss triangle disappears because the firm is happy to serve every buyer whose willingness to pay exceeds marginal cost. Output expands to the competitive level. But every dollar of that previous consumer surplus is now producer surplus — buyers as a group are no better off, often worse off, than under single-price monopoly. The total pie grows; the consumer slice vanishes.

Second- and third-degree discrimination produce ambiguous effects on total surplus. The classical result (Robinson 1933; Schmalensee 1981; Varian 1985) is that third-degree discrimination raises total surplus only if it opens up sales to markets that would otherwise not have been served at all — opening up new markets is welfare-improving; merely rearranging surplus across existing buyers is not. The intuition: discrimination changes the quantity-versus-price trade-off market by market. If output rises in aggregate, total surplus rises; if output is unchanged, the discrimination is purely redistributive.

The distributional effect is sharper. Under any form of price discrimination, the producer's share of surplus rises and the consumer's share falls relative to perfect competition. Compared with single-price monopoly, third-degree discrimination has winners and losers among consumers — the discounted group gains (matinee-goers, students, off-peak commuters), the full-price group loses (evening-goers, working-age adults, peak commuters). Whether the gains exceed the losses depends on the specific demand curves.

A simple monopoly worked example

Suppose a monopolist faces linear demand P = 100 − Q with constant marginal cost MC = 20. Three pricing strategies:

Single-price monopoly:
  MR = 100 − 2Q = MC = 20  →  Q = 40, P = 60
  Producer surplus = (60 − 20) × 40 = 1600
  Consumer surplus = ½ × (100 − 60) × 40 = 800
  Total surplus    = 2400
  Deadweight loss  = ½ × (60 − 20) × (80 − 40) = 800

Perfect price discrimination:
  Sells until P = MC at every unit
  Q = 80, every buyer charged their P(Q)
  Producer surplus = ½ × (100 − 20) × 80 = 3200
  Consumer surplus = 0
  Total surplus    = 3200
  Deadweight loss  = 0

Two-group third-degree (split into two groups with separate demands):
  Group A (price-inelastic): faces high price → close to monopoly
  Group B (price-elastic):   faces lower price → larger quantity served
  Total quantity sold rises, total surplus rises modestly,
  Producer share rises significantly.

The worked example makes the point: total surplus under perfect price discrimination (3200) strictly exceeds single-price monopoly (2400), and the entire gain plus the original consumer surplus accrues to the producer.

Real-world examples in detail

• Airlines. The most-studied second- and third-degree discriminator. Saturday-night-stay fences separate leisure from business; advance-purchase fences separate planners from last-minute buyers; cabin classes layer self-selection on top; yield-management software updates fares multiple times a day. The economics is extreme: marginal cost per additional passenger on a flight that's leaving anyway is roughly the cost of one extra meal, while willingness to pay ranges from a backpacker on a flexible date to an executive on a same-day return.
• Movie theatres. Matinee prices, student discounts, senior discounts, member discounts, child tickets — all third-degree fences. The marginal cost of seating one more person in a half-empty theatre is essentially zero.
• Pharmaceuticals. US prices for identical drugs are typically two to ten times higher than in Canada, the EU, and other developed countries — a textbook third-degree price discrimination across national markets, supported by import restrictions that prevent arbitrage. Patient assistance programmes within the US add a layer of first-degree discrimination by income.
• Uber surge pricing. When demand exceeds supply, the platform raises prices in real time. Riders who accept the surge reveal high willingness to pay; those who wait or substitute reveal low willingness to pay. The mechanism is closer to intertemporal market-clearing than to discrimination per se, but the practical effect — sorting riders by willingness to pay and capturing more of each group's surplus — is the same.
• College tuition. Sticker price is high; need-based financial aid reduces net tuition to roughly each family's ability to pay, inferred from a detailed financial-aid form. Merit aid layers a third-degree discount on high-ability students. Net tuition is therefore approximately first-degree across the income distribution and approximately third-degree across the ability distribution.
• Software-as-a-service. Almost universal second-degree versioning — basic/pro/team/enterprise tiers, per-seat pricing, usage-based add-ons. Marginal cost of a feature flag is zero; the entire pricing schedule is discrimination.
• Coupons. Old-fashioned but still effective second-degree self-selection — the cost of clipping is the fence, and price-sensitive shoppers self-select into the lower price.
• Amazon and personalised pricing. Millions of price changes per day, mostly responding to competitors and inventory, but a fraction respond to inferred buyer characteristics — closer to first-degree.
• Electricity time-of-use rates. Peak/off-peak pricing prices the elasticity of household demand across the day.
• Bundling. Cable packages, gym memberships, all-you-can-eat — a price-discrimination tool that exploits heterogeneous valuation across components.

Why economists are less hostile to it than the name suggests

The word "discrimination" carries negative connotations in everyday English, but the economic concept is neutral. Three reasons why economists generally do not regard price discrimination as a market failure to be corrected:

• Total surplus often rises. Compared with the alternative — a single price that leaves some willing-to-pay buyers unserved — discrimination expands output and total welfare. The harms are distributional, not efficiency-based.
• It enables markets that wouldn't otherwise exist. Many products have high fixed and low marginal cost (software, pharmaceuticals, broadband, broadcast media). A single price often cannot recover the fixed cost; price discrimination is what makes the product viable. Without discrimination, the entire market might disappear.
• The discount group benefits. Students, seniors, off-peak commuters, low-income buyers in poor countries — the discounted groups under third-degree discrimination are often the ones who could not afford the single-price equivalent. Discrimination is a redistribution from full-price buyers to discounted buyers, with the firm taking some of the surplus on the way.

The concerns that do attract regulatory attention tend to be inter-firm (Robinson-Patman, EU Article 102) where discrimination harms competition rather than consumers, and personalised pricing where individual-level price targeting may shade into deception or discrimination on protected characteristics.

Variants and extensions

• Two-part tariffs (Oi 1971). The disco-style "lump-sum entry fee plus per-unit charge" pricing scheme that captures most of the consumer surplus under the demand curve. Theme parks, mobile-phone plans, Costco, country clubs.
• Bundling (Adams-Yellen 1976; McAfee-McMillan-Whinston 1989). Selling two goods only as a bundle (pure bundling) or as both a bundle and separates (mixed bundling) can extract more surplus when buyers' valuations are negatively correlated.
• Versioning (Shapiro-Varian 1998). The information-economy version of second-degree discrimination — offer the same content as a free trial, a basic version, a premium version, and an enterprise version. Marginal cost of a copy is zero; pricing is purely about extracting heterogeneous valuations.
• Damaged goods (Deneckere-McAfee 1996). The seller deliberately produces a degraded version of the product to support a lower price for the price-sensitive segment — IBM laser printers slowed by an extra chip, Intel CPUs with disabled cores, software with disabled features.
• Personalised pricing. Algorithmic real-time pricing that approximates first-degree discrimination. Studied empirically since the early 2000s; regulatory frameworks (EU Digital Services Act, US state laws) still under development.
• Intertemporal price discrimination (Stokey 1979; Coase 1972 conjecture). The same product is offered at a high price today and a lower price tomorrow; consumers self-select on their willingness to wait. The hardback-then-paperback strategy, fashion seasons, durable goods.

Common pitfalls and misconceptions

• Equating it with price differences. Different prices justified by cost differences are not discrimination — they are cost-of-service pricing.
• Assuming it is always anti-consumer. The discount group under third-degree gains; total surplus often rises; the alternative may be that the market doesn't exist at all.
• Confusing discrimination with monopoly. Discrimination requires some market power but not full monopoly. Any firm with a downward-sloping demand curve can practice some discrimination.
• Ignoring the arbitrage constraint. The most common reason a beautiful pricing scheme fails in practice is that low-price buyers find a way to resell to high-price buyers — grey markets, parallel imports, ticket scalpers.
• Forgetting the fixed-cost story. In industries with high fixed and low marginal cost — pharmaceuticals, software, broadcasting — single-price markets often cannot recover fixed cost. Discrimination is then welfare-improving on a total-surplus basis because the alternative is "no product".
• Treating perfect discrimination as the benchmark. Pure first-degree is a theoretical limit; real-world discrimination is always partial. Welfare conclusions that rely on the "perfect" case may not survive in the partial case.