Note: This speech has been revised and is published in Antitrust, vol. 11, Spring 1997, pp. 21-26.
Thomas Kuhn's recent death reminded me of his book The Structure of Scientific Revolutions, which I read while in college. As I am sure many of you also recall, the book explores how a discipline's paradigms can change. Our own discipline, antitrust, underwent its own Copernican revolution within the professional experience of all but the most recent antitrust practitioners. I am speaking, of course, of the rise of the Chicago school approach.
My subject today is a less dramatic paradigm shift within antitrust, but one of great importance in merger analysis. It is the growing significance of unilateral theories of adverse competitive effect of mergers. I will describe in detail one example, based on an auction model applied at the Federal Trade Commission to analyze the Rite-Aid/Revco merger.(2) As always, my views are not necessarily those of the Federal Trade Commission or any individual commissioner.
Unilateral Theories of Potentially Adverse Competitive Effects of Mergers
The 1992 Horizontal Merger Guidelines distinguish between anticompetitive mergers that make it more likely or more effective for firms to coordinate their actions, and anticompetitive mergers that make it profitable for the merging firms to reduce output and raise price unilaterally.(3) Unilateral theories are now by far the most common, at least in the memoranda Associate Director Gary Roberts and I have written to the Commission.
This was not always the case. The first Chicago-school era merger guidelines, issued by the Justice Department in 1982, highlighted factors facilitating collusion;(4) that discussion was the predecessor to the current Guidelines' discussion of coordinated competitive effects. Although the 1982 Guidelines also contained a "leading firm proviso" that dealt with the creation or enhancement of the market power of a dominant firm,(5) the application of this unilateral theory of potential adverse competitive effects of mergers was very narrow. As late as 1986, the leading survey of antitrust policy issues raised by horizontal mergers, this Section's publication Horizontal Mergers: Law and Policy,(6) essentially ignored unilateral theories.(7)
Two developments in economics brought unilateral theories to the fore over the past decade. The first was a theoretical literature, initiated by Salant, Switzer and Reynolds, that investigated the conditions under which oligopolists would find merger profitable even if the industry members were not coordinating their actions.(8) The second was an empirical literature encouraged by the simultaneous development of new econometric tools and computerized point-of-sale scanner data (recording individual transactions at supermarkets and other retail outlets). These tools and data made it possible to identify in many cases the extent to which consumers consider individual products close substitutes; the extent to which, in consequence, individual products constrain the pricing of rivals; and the extent to which mergers encourage higher prices by removing those constraints.(9)
The 1992 Horizontal Merger Guidelines recognize these economic developments by setting forth several ways in which mergers may "less[en] competition through unilateral effects."(10) The settings in which this may occur include two in which competition is localized -- a spatial location model of competition among sellers of differentiated products,(11) and an auction model variant(12) -- and a third in which firms sell homogeneous products and are distinguished primarily by their capacities.(13)
Mergers in a Model of Auctions with Product Indivisibilities and Capacity Constraints
The Rite Aid/Revco merger illustrates one unilateral theory of adverse competitive effects from merger, an auction model that may be thought of as a variant of the homogeneous product story told in the 1992 Merger Guidelines.(14) The auction setting highlights that firms selling indivisible goods, and unable to expand output because of capacity constraints, may be able to induce price increases by making "all-or-nothing" offers -- a possibility not noted in the 1992 Guidelines.
One way to tell the story is with a numerical example, for which I am indebted to my former academic colleague, Professor Robert Hansen of Dartmouth's Amos Tuck School of Business Administration. The connection to drug stores or other real world industries may not be obvious initially, so bear with me. So you can follow the theoretical story more easily, let me tell you where it is going at the outset. I will assume that the goods are indivisible and that selling firms are capacity constrained, forcing a buyer to purchase from multiple sellers. If one seller gains control over what had been two seller's resources, the merged firm may be able to engineer a higher price by making the buyer an all-or-nothing offer for the output of its now-larger capacity. The reason is that if the buyer were to reject the all-or-nothing offer, it may find it necessary to go deeper into its list of alternative suppliers to replace the merged firm. If the buyer's alternatives grow less attractive because of differences among sellers (seller heterogeneity), the seller's bargaining leverage is enhanced and a higher price may result.
In particular, consider a market with ten sellers, where each produces exactly one unit. The goods are indivisible: a seller cannot produce a partial unit. This assumption is often reasonable: it is typically hard to sell part of a hospital or part of a brand name, for example.(15) Because the selling firms are capacity constrained, an individual buyer seeking multiple units must purchase from multiple sellers. All firms' products are identical, but the sellers differ -- in the example, they have different production costs.(16) The first seller produces its unit at a marginal cost of one, the second seller at a marginal cost of two, and so forth, so the tenth seller has a cost of ten.
Now suppose that a buyer needs seven units of the good, but has little or no use for more.(17) The buyer asks all sellers to bid to supply any or all of the seven units it needs, and pays the same price to all sellers whose bid it accepts. Although the setting is an auction, it differs from the auction model mentioned in the Merger Guidelines because the products are homogeneous and each seller's capacity is constrained.(18) In the most plausible auction models,(19) the buyer will purchase one unit from each of the first seven firms, who are lowest cost suppliers, and pay a price of eight (or just under) for each unit. No seller can induce the buyer to pay more than eight, because the buyer has the option of turning to the eighth firm, who is just willing to produce and sell for a price of eight.(20) The total paid by the buyer is 56.
The potential for firms to engineer a price rise by making an all-or-nothing offer is demonstrated when any two of the seven lowest-cost sellers merge. To make the example concrete, I will assume that the merger involves the first two sellers, who happen to be the two lowest cost producers. After the merger, the combined one/two firm offers the buyer two units at a price of (just under) eighteen for the package. This is equivalent to demanding that buyer acquire two units and pay a price (just under) nine for each. Buyer's best option for purchasing the seven units it needs is to accept that offer and obtain the other five units it requires from firms three through seven. Buyer obtains the seven units it demands from the seven lowest cost sellers, which include the merged firm, and pays a price of (just under) nine for each. As all sellers receive the higher price, the per unit price rises from eight to nine and the total paid by the buyer increases from 56 to 63.
The reason the price goes up is that the buyer's alternatives, if it refuses an all-or-nothing offer from the merged firm, are less attractive then they were before the merger.(21) After the merger, if the buyer turns down the all-or-nothing offer from the combined firms one and two, it can no longer merely turn to a seller with a cost of eight; the buyer's best alternative is now to buy from both the eighth and ninth firms. Because one of these producers has a cost of nine, the buyer would have to pay a price of nine to every seller.(22) Accordingly, the merged firm can extract a per unit price of (just under) nine through an all-or-nothing offer to the buyer.(23)
I will highlight two features of this example that are critical to generating the price increase. First, the available units exceed what the buyer needs. In consequence, what the firms produce are substitutes to the buyer,(24) making the transaction a horizontal merger among the sellers of substitutes. I will return to this point later, after discussing the Rite Aid/Revco merger. Second, the ninth firm's output is less attractive to the buyer than the eighth firm's; in the example, it costs more to produce. This makes the buyer's alternatives worse when the merged firm makes an all-or-nothing offer, compared to the buyer's alternatives to dealing with the first firm before the merger. If instead, for example, the ninth firm were identical to the eighth firm and could also produce at a cost of eight, the merged firm could not induce the buyer to sell to it at a price above eight by making an all-or-nothing offer.(25)
In the numerical example -- characterized by product indivisibilities, seller capacity constraints, and a rigid buyer "need" requirement -- the surplus that the seller extracts through all-or-nothing offer is a pure transfer from the buyer. However, in settings where seller has some ability to restrict output or lower quality, and buyer has some ability to reduce its purchases in response to a price increase, the merger could also be expected to generate an efficiency loss. On the other hand, if the merged firm can credibly make all-or-nothing offers, it may also be able to expand output through volume discounts, multipart pricing or other complex pricing schemes, and thereby potentially create an efficiency gain to weigh against that efficiency loss.
As with other unilateral theories of adverse competitive effect from the merger of firms selling demand substitutes, this story shows how the loss of seller competition makes buyer's best alternative to dealing with a seller less attractive, leading buyer to pay a higher price. The novelty here comes from marrying product indivisibilities and seller capacity constraints with the auction model. In the auction model mentioned in the Merger Guidelines, only mergers involving the marginal excluded firm, firm eight in the example, would raise price by making buyer's alternatives less attractive. In this story, in contrast, mergers among inframarginal sellers, such as firms one and two, can create market power. And, again as with other unilateral theories,(26) even small increases in market concentration can generate higher prices.(27)
The Rite Aid/Revco Merger
Let me now describe, in terms of this theory, why the Rite Aid/Revco merger appeared to create a competitive problem. The merger involved two large operators of retail pharmacies in the United States, each with a substantial market share in numerous metropolitan areas. The Commission was concerned that the combined firm would be able to exercise market power in the sale of retail pharmacy services to managed care providers offering pharmacy benefits and to their enrollees. Previous drug store mergers investigated by the Commission had not involved consolidations of this scope and had not found problems in the sale of pharmacy services to managed care; the earlier investigations had uncovered a potential for raising drug prescription prices to cash customers only.
I will explain the competitive problem through a hypothetical example, involving a Blue Cross plan that seeks to put together a pharmacy network in a major metropolitan area. I will spend a few minutes setting forth a simplified description of the institutional setting before describing the way a merger among pharmacy chains could raise price to Blue Cross.
Blue Cross needs to contract with a pharmacy network in order to provide drug benefits to those covered by its health care plan. It wants to market the health plan to major employers in the local area, and needs to tell plan enrollees what pharmacies they can go to in order to fill their prescription. Blue Cross must pay the pharmacies by the prescription to perform that service.
Both Blue Cross and the pharmacies see a tradeoff between a large network and a small one. This tradeoff is fundamentally between convenience and price. A large network, containing most of the pharmacies in town, will allow virtually all plan enrollees -- individual consumers -- to obtain their prescriptions at a convenient location. This makes the plan attractive to enrollees and employers and, for that reason, profitable for Blue Cross and the pharmacies in its network.(28) But a small network, one containing only a fraction of the pharmacies in the metropolitan area, gives each member pharmacy more business. In exchange for that benefit, the pharmacy may be willing to accept a lower per-prescription payment from Blue Cross, thereby reducing Blue Cross's costs and allowing Blue Cross to lower the price of the drug benefits plan to employers and enrollees.
In my hypothetical drug plan example, Blue Cross has found that most employers, looking at the price reductions made possible by reducing the scope of pharmacy networks, resolve the tradeoff largely in favor of convenience. Accordingly, Blue Cross believes it must include, let us say, at least 60 percent of the drug stores in town, distributed across all neighborhoods, in order to offer a drug plan it can successfully market to employers.(29) All pharmacy counters are not equally attractive to Blue Cross. On average, Blue Cross favors pharmacy counters located in stores that are part of a large chain to pharmacies in small chains, and it prefers chain store pharmacies to independents.(30) Some of the independent stores may be as attractive to Blue Cross as some of the chain stores, but many of the independent stores are unattractive.
Blue Cross effectively conducts an auction, seeking bids from pharmacies for the payment they would require to join Blue Cross's network. The setting reasonably tracks the auction model we have been discussing. An individual drug store is limited to a single location, thus creating both an indivisibility and the equivalent of a capacity constraint from the point of view of Blue Cross seeking to put together a network of pharmacies at multiple locations. Seller heterogeniety creates the equivalent of marginal cost variation across pharmacy counters: chain store pharmacies tend to be like the low cost sellers in the numerical example by virtue of their greater attractiveness to Blue Cross, where cost is conceived of in terms of providing retail pharmacy services with a given level of quality to Blue Cross.
Blue Cross does not negotiate with pharmacy chains for store-by-store participation in its pharmacy network; it negotiates chain-by-chain. So let us suppose, again hypothetically, that in some metropolitan area Rite Aid accounts for 25 percent of the pharmacy counters and Revco accounts for 10 percent. Eight other chains each control 5 percent of the pharmacy counters, and the remaining 25 percent are found in individually-owned stores.(31)
Blue Cross's ability to negotiate a low price with Rite Aid turns on the nature of its alternatives. Before the merger, when Blue Cross informally seeks bids from drug store chains for joining its network, it views Rite Aid as making an all-or-nothing offer for 25 percent of the pharmacy counters in town. Without Rite Aid, it can put together a network with 60 percent coverage by contracting with Revco, the eight small chains, and some of the better independents. It needs less than half the independents to do so. After the merger, Rite Aid would make an all-or-nothing offer involving 35 percent of the area's pharmacy counters. Blue Cross's alternatives for doing without Rite Aid would then be significantly worse; to avoid dealing with the merged firm, Blue Cross must go deeper into the group of sellers it would previously have rejected. To reach 60 percent coverage without Rite Aid, it must include most of the independent pharmacies, including some of the least attractive ones. Recognizing this, the Rite Aid can successfully hold out for a higher price than it would have received before the merger, and the price Blue Cross must pay rises.
This is not a theory that automatically makes any drug chain merger illegal. Some of the counter-arguments Rite Aid might make, assuming factual support, would apply to protect any merger from antitrust challenge by government enforcers, regardless of the competitive effects theory. These include proof that post-merger concentration is within Merger Guidelines safe harbors based on market shares, that entry would solve the competitive problem, that efficiencies would lead the merged firm not to raise price, or that efficiency benefits from the deal outweigh any harm to competition.
Other possible arguments for the parties come from the framework of the auction model.(32) The parties might prove that Blue Cross's alternatives do not get worse following this merger. For example, they might show that most of the independent pharmacies are about as attractive to Blue Cross as the small chains. If so, the merged firm could not obtain more bargaining leverage in dealing with Blue Cross by framing a larger all-or-nothing offer than Rite Aid had made pre-merger.(33) Or the parties might demonstrate that in response to the anticompetitive potential of this merger, enough of the least attractive independents would find it worthwhile to make themselves more desirable to Blue Cross -- and thus prove that repositioning by non-party rivals would solve the competitive problem. Or the parties might show, contrary to what the hypothetical example assumed, that Blue Cross and its enrollees are indifferent between a network with 60 percent coverage and a network with, let us say, 30 percent coverage (after consideration of the lower price for prescription drugs that a smaller network could make possible).(34) If so, Blue Cross's alternatives to dealing with Rite Aid do not get worse with the merger. Blue Cross retains the post-merger alternative of putting together a smaller network composed only of chain stores, all from the smaller chains, without need for including any of the unattractive independents.
I promised several minutes ago, when discussing the theoretical example of auctions with product indivisibilities and capacity constraints, that I would return to one point: the comment that when the available units exceed what the buyer needs, what the firms produce are substitutes to the buyer and the transaction is a horizontal merger among the sellers of demand substitutes. With respect to the hypothetical drug chain merger, the point is that individual Rite Aid and Revco stores in different parts of town can be substitutes from the perspective of Blue Cross even if they are not substitutes from the point of view of individual consumers. The stores compete to fill out Blue Cross's network, which requires 60 percent coverage, even if they do not compete to serve individual customers seeking to fill prescriptions.
This phenomenon -- that products can be substitutes from the point of view of a distributor even if they are not substitutes from the perspective of view of end use consumers -- has come up in other merger investigations.(35) We might want to understand, to choose an arbitrary example, why two cable television channels, like a news channel and a movie channel, could be substitutes from the point of view of a cable system operator even if it were to turn out that most viewers did not consider the two channels substitutes. While I am raising this example as a hypothetical, I should note that the Federal Trade Commission has as yet taken no action on a proposed settlement in our ongoing cable matter.
One possible explanation why a news channel and a movie channel could be substitutes for the cable system is suggested by the theory of mergers in a model of auctions with product indivisibilities and capacity constraints. Suppose there are 10 "marquee" cable networks, and a cable system needs to carry at least 7 out of the 10 in order to attract a significant number of subscribers. Then the marquee networks would be substitutes in filling the cable system's "must have" slots.(36)
Let me conclude by noting another way in which the development of science, as that field was described by Thomas Kuhn, parallels the development of antitrust. In much the way that "normal science" proceeds, antitrust practitioners, enforcers, and courts must creatively work out, over time, the implications of a new paradigm like that of the unilateral competitive effects of mergers. This is an exciting challenge, and one that we should welcome.
1. Director, Bureau of Economics, Federal Trade Commission. The author is grateful to Bob Hansen, Gary Roberts, and Bobby Willig.
2. FTC Will Seek to Block Rite Aid/Revco Merger, FTC News, Federal Trade Commission, April 17, 1996; Rite Aid Abandons Proposed Acquisition of Revco After FTC Sought to Block Transaction, FTC News, Federal Trade Commission, April 24, 1996.
3. U.S. Department of Justice and Federal Trade Commission, Horizontal Merger Guidelines, April 2, 1992, 2.
4. U.S. Department of Justice, Merger Guidelines, June 14, 1982, III.C.
5. U.S. Department of Justice, Merger Guidelines, June 14, 1982, III.A.2.
6. ABA Antitrust Section, Monograph No. 12, Horizontal Mergers: Law and Policy (1986).
7. See id. at 254 n.1300.
8. Stephen W. Salant, Sheldon Switzer & Robert Reynolds, Losses from Horizontal Merger: The Effects of an Exogenous Change in Industry Structure on Cournot-Nash Equilibrium, 98 Q. J. Econ. 185 (1983). Later contributions include Raymond Deneckere & Carl Davidson, Incentives to Form Coalitions with Bertrand Competition, 16 RAND J. Econ. 473 (1985); Martin K. Perry & Robert H. Porter, Oligopoly and the Incentive for Horizontal Merger, 75 Am. Econ. Rev. 219 (1985); and Joseph Farrell & Carl Shapiro, Horizontal Mergers: An Equilibrium Analysis, 80 Am. Econ. Rev. 107 (1990).
9.E.g. Jonathan B. Baker & Timothy F. Bresnahan, The Gains from Merger or Collusion in Product-Differentiated Industries, 33 J. Ind. Econ. 427 (1985); Steven Berry & Ariel Pakes, Some Applications and Limitations of Recent Advances in Empirical Industrial Organization: Merger Analysis, 83 Am. Econ. Rev. 247 (Papers & Proceedings, May 1993); Jerry Hausman, Gregory Leonard, J. Douglas Zona, Competitive Analysis with Differenciated Products, 34 Annales D'Econ. Stat. 159 (1994); Gregory J. Werden & Luke M. Froeb, The Effects of Mergers in Differentiated Products Industries: Logit Demand and Merger Policy, 10 J. L. Econ. & Org. 407 (1994).
10. U.S. Department of Justice and Federal Trade Commission, Horizontal Merger Guidelines, April 2, 1992, 2.2.
11. U.S. Department of Justice and Federal Trade Commission, Horizontal Merger Guidelines, April 2, 1992, 2.21.
12. U.S. Department of Justice and Federal Trade Commission, Horizontal Merger Guidelines, April 2, 1992, 2.21 n.21.
13. U.S. Department of Justice and Federal Trade Commission, Horizontal Merger Guidelines, April 2, 1992, 2.22.
14. The most closely related theoretical analysis may be Justice Department economist Gregory Vistnes' examination of how an alliance among multiple hospitals within the same market can profitably raise prices charged health insurers. Gregory Vistnes, Strategic Alliances and Multi-Firm Systems, unpublished manuscript (October 1995). Deneckere & Davidson, note 8 supra, and Perry and Porter, note 8 supra, had previously highlighted the importance of constraints on the ability of non-merging rivals to expand, such as capacity constraints or product differentiation, in making it profitable for merging firms to reduce output and raise price unilaterally.
15. In some cases quality reduction could be understood as a form of doing so.
16. The analysis would be similar if the products cost the same to produce, but vary in attractiveness to the buyer. Then, from the point of view of the buyer, the quality-adjusted cost would effectively vary in a way similar to what is postulated in the example in the text.
17. The buyer needs 7 units in the sense of having a very high willingness to pay for each. Matters would not change much if buyer would be willing to buy more than 7 units so long as the price is no higher than that charged for the first seven.
18. The example assumes the auction is "simultaneous" rather than sequential, that sellers' costs are public information, and that all sellers receive the identical price per unit. These assumptions highlight the logic, which remains important if the auction is sequential, sellers' costs are private, or buyers can pay sellers discriminatory prices.
19. This result would arise, for example, in a sealed-bid, second-price model.
20. None of the low cost sellers can extract more from buyer (nor divert to itself any of the payments buyers make to other sellers) by announcing that unless buyer pays it more than eight, it will not join the auction. Pre-merger, such an announcement would not be credible in the postulated auction setting.
21. Before the merger, each seller made an all-or-nothing offer to the buyer for its sole, indivisible unit in a non-trivial sense. That is, the merger could generate a price increase if capacity-constrained sellers capable of producing multiple units make all-or-nothing offers pre-merger -- if the sellers credibly present buyers with a supply schedule giving a price for various combinations of units sold -- and if the merged firm can make a larger all-or-nothing offer than could any firm pre-merger.
22. If the buyer were to conduct an auction among the remaining firms (three through ten) to buy the seven units it needs, ignoring the merged firm rather than using the alternative of buying from firms eight and nine to limit the bargaining leverage of the merged firm, the buyer would pay an even higher price per unit, of ten.
23. In the example, the merged firm would not find it profitable to raise price unless it can forbid the buyer from choosing to purchase one but not two units from it. The reason is that the merged seller would lose more from not selling the second unit than it would gain by raising price on the unit it does sell. Were the merging firms instead high cost sellers, so that the foregone contribution margin on the unit not sold is small, it could be profitable for the merged firm to permit the buyer from choosing to purchase only one unit, though not as profitable as insisting on an all-or-nothing offer.
24. Once the buyer has chosen its first six suppliers, all remaining suppliers are competing to sell the seventh unit the buyer requires. If the buyer needed all ten units available in the market, in contrast, the output of the sellers would be complements rather than substitutes from the perspective of the buyer.
25. The steeper the slope of the supply curve buyer faces in the neighborhood of the number of units it needs -- a possible consequence of seller heterogeneity -- the greater the price increase likely to result from a given merger.
26. See Jonathan B. Baker, Product Differentiation Through Space and Time: Some Antitrust Policy Issues, The Antitrust Bulletin, forthcoming.
27. In the example, a merger creating a firm owning 20% of the market, and increasing the HHI from 1000 to 1200, led to price increase of 12.5%.
28. If the Blue Cross plan has many enrollees, each pharmacy in the network will be able to fill more prescriptions. Not only that, people walking into the drug store to fill prescriptions tend to buy other products. Health plans generally do not insist on exclusivity, so the pharmacies would not be forced to drop other networks in order to participate in Blue Cross's network.
29. Nothing significant in the story changes if, after making a deal with 60 percent of the stores, Blue Cross takes on any other pharmacies that choose to join the network on the same terms. When a stock exchange specialist "cleans up the book," it effectively conducts an auction in a similar way.
30. Chain stores tend to have features that make them attractive to consumers -- better service, broader merchandise stock, lower prices, convenient location, better parking, and the like -- and thus make them attractive to Blue Cross. Chain stores also tend to have features that make them attractive to Blue Cross directly, such as computerized drug inventory management.
31. All the stores of any given level of attractiveness to Blue Cross -- whether large chain stores, small chain stores, the better independents or the less desirable stand-alone stores -- are distributed throughout the area in roughly the way the population is distributed. This assumption, and the assumption that each pharmacy counter is capable of filling any prescription, make outlet counts a reasonable measure of competitive significance in the hypothetical example.
32. In addition to the arguments below, Rite Aid might highlight drug chain uncertainty about rival chain bids to Blue Cross. Such uncertainty probably reduces the competitive harm from merger without eliminating it, however.
33. In terms of the numerical auction example set forth previously, this is like demonstrating that the eight and ninth firms both have costs of eight.
34. In terms of the auction example set forth previously, it is as if the government claims that Blue Cross needs 7 units while Rite Aid insists that Blue Cross needs only 3 units. If health plan enrollees are not indifferent, but prefer the network with 60 percent coverage (as the hypothetical example presumed), and if Blue Cross responds to the merger by shrinking its network to 30 percent coverage in order to keep drug prices from rising, that response would reduce enrollee convenience without lowering price, and thus constitute harm to competition. The possibility that Blue Cross might seek to evade anticompetitive price increases in retail markets for prescription drugs by substituting a mail order network for a retail network with 60 percent coverage would be analyzed similarly.
35. Another example of this phenomenon, not necessarily related to a merger investigation, comes from a shopping mall in one of the Washington, D.C. suburbs. The mall recently lost the I. Magnin department store, and brought in Borders Books to fill the space. These stores do not in general sell substitutes from the point of view of most shoppers. To the extent they represent alternative ways for the shopping mall owner to generate customer traffic by putting together a package of stores, they are nevertheless substitutes from the mall owner's point of view.
36. A second possibility, suggested by Princeton economist Robert Willig, is that a news channel and a movie channel might be very attractive to the same group of viewers. The two channels might then be alternatives for inducing those viewers to become cable system subscribers, even if the channels are not substitutes to the viewers. To provide an analogy: suppose that economists like Kung Pao chicken and hot-and-sour soup independently, but are not particularly enthusiastic about other Chinese cooking. Economists then might start to patronize a Chinese restaurant that added either dish to its fixed-price lunch buffet. Thus, even if the economists do not view these dishes as substitutes, they would be substitutes from the point of view of the restaurant owner seeking to attract economist business. A third possibility is that viewers might on average desire variety in cable formats (while placing a lower value on further variety once offered some choice). For example, once viewers have available a leading news channel, they may next desire access to a leading channel in another key niche such as movies or sports. If so, a news channel and a movie channel might represent alternative ways of adding variety to the cable package, making them substitutes from the cable system operator's perspective.