Definition and Types of Price Discrimination
In the last section, online price discrimination was defined as differentiating the online price for identical products or services partly based on information a company has about a potential customer. Simple examples, which many people feel comfortable with, are a reduced conference fee for academics or doctoral students as opposed to participants from commercial entities, or a reduced fee for senior citizens or children when booking theatre tickets online. However, this definition fails to account for differences in cost which serving different customers may entail and which may justify price differences. Transportation or delivery costs can differ between customers, for instance, and including such costs in the total charge would generally not be considered price discrimination. Similarly, differences in legal regimes can translate into diverse cost levels for the same service between different EU member states. Other more personal characteristics may also give rise to cost differences resulting in price differences, for instance in insurance and credit markets. Based on demographic data or a person’s track record, he or she may have a higher probability to cause a traffic accident, fall ill, become unemployed, or default on a loan. By consequence, the cost of providing insurance or credit will differ. These cost differences will generally justify price differences that most authors would not consider price discrimination.Footnote 2
An economically more proper definition of price discrimination by Stigler is “the sale of two or more similar goods at prices that are in different ratios to marginal costs” (Stigler 1987, p. 210). Indeed, under this definition, the examples above of price differences that purely stem from cost differences would not qualify as price discrimination.
Stigler’s definition of price discrimination can be criticized for being vague by using the word “similar.” However, an advantage of this vagueness is that the definition also covers versioning of products. A classic example of versioning is selling a book in a paperback and a hardcover edition. Publishers usually make more profit on the hardcover. A notorious example of versioning concerns IBM, which added a microchip to its popular professional LaserPrinter to reduce the printing speed and subsequently sell it for half the price on the consumer market (e.g., Preston McAfee 2008). Other examples are versions of the same software that have slightly different features, cars with different engine powers, etc. All these product sets can be considered similar, and in such examples, the version that has superior features or quality is typically sold with a higher profit margin.
For price discrimination to work, three conditions need to be satisfied: A seller must be able to distinguish between customers to know which price to charge to whom, he must have enough market power to be able to set prices above marginal costs, and resale must be impractical, costly, or forbidden to prevent arbitrage between customers (e.g., Varian 1989). For online sales, these conditions are often met: Distinguishing customers is possible with great accuracy. Various internet sellers have very high market shares in their relevant market which are likely to give them at least some market power, while market power may also derive from switching costs or lack of transparency in the market. And, resale is often impossible (in the case of airline tickets or hotel rooms for instance) or relatively costly. Combined with the ease of adapting prices unnoticed, this suggests that a lot of price discrimination should occur online.
A classic distinction is between first-, second-, and third-degree price discrimination (Pigou 1932). First-degree price discrimination refers to a situation in which each consumer is charged an individual price equal to his or her maximum willingness to pay. For first-degree price discrimination, the seller needs precise information about the buyer’s willingness to pay (the reservation price). First-degree price discrimination enables the sellers to extract all consumer surpluses. In practice, such an extreme form of price discrimination will never occur, as sellers cannot learn the buyer’s exact reservation price. First-degree price discrimination serves as a stylized benchmark to evaluate other pricing schemes.Footnote 3
Second-degree price discrimination refers to pricing schemes in which the price of a good or service does not depend on characteristics of the customer but on the quantity bought. Such schemes are also called “non-linear pricing” and may involve a quantity discount, or a two-part tariff with a fixed fee and a variable fee. For example, in the cinema, popcorn is often cheaper (per gram) if you buy a larger box. For second-degree price discrimination, the seller does not need information about the buyer, as buyers self-select: They choose a different price by choosing a different quantity.
Certain types of loyalty schemes are sometimes also characterized as second-degree price discrimination, namely when a customer gains credits or discounts with past purchases. If a loyalty scheme only amounts to a quantity discount over time, it can indeed be seen as a type of second-degree price discrimination. However, loyalty schemes are also used for other sales strategies such as cross-selling or up-selling, i.e., to sell additional or more profitable products and services to existing customers. As a loyalty scheme is an excellent tool for making customer profiles, it can also be used for personalized pricing. This implies that loyalty schemes should also be considered within the broad scope of third-degree price discrimination.
In third-degree price discrimination, prices (or more properly price-to-marginal-cost-ratios) differ between groups or types of buyers. This type of price discrimination is widely used and often uncontroversial: Many companies offer discounts to children, students, or elderly. A reduced conference fee for academics is an example of third-degree price discrimination too. A company could also charge people from different geographical areas different prices. For instance, medicines or college textbooks could be sold at lower prices in developing countries.
For third-degree price discrimination, it is not necessary to recognize individual buyers: Sellers only need to know the characteristic of the buyer that is used to discriminate prices. However, to distinguish types of buyers, sellers often use unique identifiers such as a student card with a student number and photo or even a formal ID card. Uniquely identifying customers helps to satisfy two of the key conditions for price discrimination to work: distinguishing between buyers and preventing ineligible customers to obtain a discount by arbitrage.Footnote 4 Nevertheless, such practices often lead to over-identification: More precise identification takes place than third-degree price discrimination requires.
Online price discrimination will typically work similarly: A cookie, IP address, or user log-in information enables an online company to identify a customer.Footnote 5 Like the student ID, this unique identification will generally not be the purpose but a means to the end of third-degree price discrimination, e.g., to distinguish between broader categories, for instance high and low spenders. However, compared to selecting students on the basis of a student ID card, an online profile can be much more detailed and can allow for much more refined price discrimination. By doing so, online third-degree price discrimination can, at least in theory, be pushed towards a seller’s holy grail of perfect or first-degree price discrimination, under which all consumer surplus is extracted to the benefit of the seller.
Versioning is outside the scope of this paper. If sellers cannot easily identify groups of customers for price discrimination, they can tempt customers to self-select by using versioning.Footnote 6 For example, customers with a higher willingness to pay could be tempted to buy hardcover books instead of paperbacks. Airlines aim to separate price-insensitive business travellers from price-sensitive leisure travellers by requiring a stay-over on Saturday night for cheaper tickets. Versioning could also be used to offer customers with a certain profile only the more expensive versions of a product. Such practices generally do not require the seller to have specific information about the buyer.
Price steering also falls outside the scope of this paper, as it does not involve price differentials between customers. Price steering is “personalising search results to place more or less expensive products at the top of the list” (Hannak et al. 2014). Hence, a website suggests a certain product and price to a consumer, but the consumer can still buy another product. For instance, the travel site Orbitz showed Apple users more expensive hotels than it showed to PC users (Mattioli 2012). Note, however, that since the use of specific information about potential buyers will also be required for price steering, conclusions about the applicability of data protection law may transfer to it.
Welfare Effects of Price Discrimination
From a welfare economic perspective, there can be good arguments in favour of price discrimination, as it can benefit both buyers and sellers. To illustrate this, image a monopolist who supplies a good with constant marginal production cost of €5 to a market of 100 customers (Fig. 1). Half of these customers, the high spenders, are willing to pay an amount between €15 and €20 for this product; the low spenders between €8 and €10.Footnote 7 Under uniform pricing, the monopolist would consider two price points. At P
low = €8, all customers are buying. The monopolist sells 100 units (Q
low = 100), making a profit of €3 per unit, €300 in total. Low spenders have a consumer surplus (the difference between their willingness to pay and what they pay) of between €0 and €2, and high spenders between €7 and €12. Total consumer surplus at this price is 50 × (€9–8) + 50 × (€17.5–8) = €525. Total welfare, the sum of consumer surplus and profits, equals €825.
At a price P
high = €15, the monopolist sells only 50 units to high spenders but makes a profit of €10 per unit, yielding a profit of €500. Hence, without price discrimination, the monopolist will set a price of €15, sell 50 units and leave low spenders unsupplied, even though they are willing to pay more than the marginal costs of production. Consumer surplus in this scenario is 50 × (€0 + 5)/2 = €125. Total welfare equals €625, which is less than in the previous scenario.
What will happen if the monopolist can price discriminate and charge €15 to high spenders and €8 to low spenders? Total profits become 50 × €10 + 50 × €3 = €650, making the monopolist better off. Consumer surplus becomes 50 × (€17.5–15) + 50 × (€9–8) = €175. Total welfare becomes €825, leaving both the monopolist and the consumers better off in comparison to the uniform price of €15 that is optimal for the monopolist.
Total welfare under price discrimination (€825) is equal to total welfare at the lower uniform price of €8. This is no coincidence. In a simple case such as this, welfare is created by the transactions that take place in the market, while the price paid only affects the distribution of welfare between the consumers and the producer. Therefore, any price between €5 and €8 generates the same (maximum) amount of total welfare. When a non-discriminating price is raised above €8, welfare is destroyed because customers who have a willingness to pay above marginal costs are not served by the market which entails so-called dead-weight-losses.Footnote 8
The example illustrates that price discrimination can reduce deadweight losses that result from consumer heterogeneity: Uniform prices often leave some consumers unserved, even though they have a willingness to pay which exceeds marginal costs. This causes a loss of welfare both for consumers and sellers, relative to a situation in which it would be possible to charge these consumers a price that is only slightly above marginal costs. The potential for price discrimination is particularly high in markets with high fixed costs and low marginal costs: Price discrimination can help the seller to regain its fixed costs without (substantial) dead-weight-losses.
On the other hand, for some customers, price discrimination will lead to higher prices than a uniform price. Hence, price discrimination deprives some consumer groups of some of their consumer surplus. The more refined the price discrimination scheme that the seller uses, the more he can deprive consumers of their consumer surplus. For instance, one could envisage the monopolist in Figure 1 discriminating further within the group of high spenders, to extract more surplus. This is not so much a concern for the aggregate welfare created in a market, but it is for the distribution of the welfare between producers and consumers.
Moreover, price discrimination can enhance market power: In the example above, the uniform price that was optimal for the monopolist would leave half the market unsupplied, thereby leaving room for competitors to serve the low end of the market. By using price discrimination, the monopolist can serve the entire market, with the result that it can increase its economies of scale and network effects. Thus, price discrimination can help to monopolize a market and to make market entry unattractive for competitors.
There is a large literature on the outcomes and welfare effects of price discrimination in various different competitive settings and under various assumptions about consumer demand, information that consumers and producers have, producer’s ability to commit to prices, etc. For an overview, see for instance Varian (1989) and Armstrong (2006). These welfare effects turn out to be ambiguous. Generally, for price discrimination to be welfare enhancing, it must lead to a substantial increase in total output by serving markets that were previously unserved. For instance, in the stylized case above, price discrimination can lead to a welfare improvement for both consumers and sellers. But even then, price discrimination will lead to a loss of welfare for some consumers.
When price discrimination does not lead to substantial market expansion, it often reduces total consumer surplus to the benefit of producer surplus. Price discrimination may even lead to a net welfare loss, when producers gain but consumers lose more. And sometimes, even sellers can suffer a welfare loss, due to intensified competition. The closer personalized pricing approaches first-degree price discrimination, the more it will extract welfare away from consumers and towards producers. Finally, price discrimination often introduces transaction costs for consumers and producers.
In conclusion, a welfare economic stance towards price discrimination is ambiguous. Under the right circumstances, price discrimination can increase total welfare and can even be averagely beneficial for consumers, provided it leads to a substantial increase in total output. On the other hand, price discrimination can lead to a transfer of welfare from consumers to sellers or even to a reduction of total welfare. In any case, consumers with a high willingness to pay will most probably be worse off under price discrimination.