WHEN ARE SUNK COSTS BARRIERS TO ENTRY? ENTRY BARRIERS IN ECONOMIC AND ANTITRUST ANALYSIS†

What Is a Barrier to Entry?

By R. PRESTON MCAFEE,HUGO M. MIALON, AND MICHAEL A. WILLIAMS*

In the Papers and Proceedings of the Forty- ing the concept of barriers to entry. We begin by eighth Meeting of the American Economic As- contrasting the definitions of an entry barrier sociation, Donald H. Wallace (1936 p. 83) proposed in the economics literature. We then proposed a research program that proved vision- introduce a classification system to clear up the ary: “The nature and extent of barriers to free existing confusion, and we employ it to assess entry needs thorough study.” Fifteen years later, the nature of the barriers posed by scale econ- Joe S. Bain published a book that was the first omies and sunk costs. thorough study of entry barriers. In this book, Bain (1956) defined an entry I. History of the Concept barrier as anything that allows incumbents to earn above-normal profits without inducing en- In chronological order, the seven principal try. He believed that economies of scale and definitions of an entry barrier proposed in the capital requirements meet his definition because economics literature are as follows. they seem to be positively correlated with high profits. George J. Stigler (1968) later defined an Definition 1 (Bain, 1956 p. 3): A barrier to entry barrier as a cost advantage of incumbents entry is an advantage of established sellers in an over entrants. With equal access to technology, industry over potential entrant sellers, which is scale economies are not an entry barrier accord- reflected in the extent to which established sell- ing to this definition, and neither are capital ers can persistently raise their prices above requirements, unless incumbents never paid competitive levels without attracting new firms them. to enter the industry. With respect to scale economies and capital costs, the definitions of Bain and Stigler are at Definition 2 (Stigler, 1968 p. 67): A barrier to variance, which has resulted in controversy entry is a cost of producing (at some or every among economists and antitrust lawyers, both rate of output) that must be borne by firms over the definition of an entry barrier, and the seeking to enter an industry but is not borne by question of whether scale economies and capital firms already in the industry. costs each constitute one. The present article is an attempt to resolve the controversies concern- Definition 3 (James M. Ferguson, 1974 p. 10): A barrier to entry is a factor that makes entry unprofitable while permitting established firms to set prices above marginal cost, and to persis- † Discussants: Steve Berry, Yale University; Michael H. Riordan, Columbia University; Michael Whinston, North- tently earn monopoly return. western University. * McAfee: Humanities and Social Sciences, California Definition 4 (Franklin M. Fisher, 1979 p. 23): Institute of Technology, Pasadena, CA 91125 (e-mail: A barrier to entry is anything that prevents entry [email protected]); Mialon: Department of Economics, when entry is socially beneficial. Emory University, Atlanta, GA 30322-2240 (e-mail: [email protected]); Williams, ERS Group, 2000 Pow- ell Street, Suite 500, Emeryville, CA 94608 (e-mail: Definition 5 (C. C. von Weizsacker, 1980 p. [email protected]). We thank Sue Mialon for valu- 400): A barrier to entry is a cost of producing able comments. that must be borne by a firm seeking to enter an 461 462 AEA PAPERS AND PROCEEDINGS MAY 2004 industry but is not borne by firms already in the harmful only if potential entrants make a calcu- industry, and that implies a distortion in the lation that is different from the one that society allocation of resources from the social point of would want them to make in deciding whether view. to enter an industry that possesses the barrier in question. Definition 6 (R. Gilbert, 1989 p. 478): An entry Von Weizsacker’sdefinition is also norma- barrier is a rent that is derived from tive, but it follows Stigler’sdefinition. He ar- incumbency. gues that a cost differential is an entry barrier only if it reduces welfare. His point is that the Definition 7 (Dennis Carlton and Jeffrey Per- number of firms in a Cournot industry can be loff, 1994 p. 110): A barrier to entry is anything greater than the socially optimal number. In this that prevents an entrepreneur from instanta- case, entry barriers serve a socially desirable neously creating a new firm in a market. A purpose. long-run barrier to entry is a cost necessarily Gilbert’sdefinition focuses on the advantages incurred by a new entrant that incumbents do of incumbents rather than the disadvantages of not (or have not had to) bear. entrants. According to him, an entry barrier is the additional profit that firms can earn as a sole Bain’sdefinition is flawed in that it builds the consequence of being established in the indus- consequences of the definition into the defini- try. This definition has an immediate problem in tion itself. Moreover, one can imagine an indus- that a profit is not a barrier. try with competitive pricing due to the presence Carlton and Perloff offer two definitions. of many incumbents but with no possibility of They argue that the first is not practical because entry (e.g. by government fiat); such an industry it implies that any capital requirement is an has no entry barriers according to Bain’s entry barrier, and that any industry in which definition. entry takes time has an entry barrier. They note Stigler’sdefinition avoids tautology by identi- that the term “barrier to entry” is often used to fying an entry barrier in terms of its fundamental refer to both costs of entering and the time characteristics, emphasizing the differential costs required to enter. However, to our knowledge, between incumbents and entrants. However, the they are the first to propose a definition that present tense “is” in the definition is confusing. explicitly includes a time dimension. Literally, the definition implies that a cost that Unfortunately, they avoid the timing issue by only entrants (not incumbents) have to bear considering only entry barriers in the long run. today is an entry barrier, even if incumbents had They argue that a firm can only earn profits in to bear it in the past (when they entered the the long run if it has an advantage over potential market). entrants, which leads them to adopt a modern Stigler’sdefinition is narrower than Bain’s: version of Stigler’sdefinition. Notice that their some costs are barriers according to Bain and version clears up the confusion about the not according to Stigler; but all Stiglerian bar- present tense “is” in Stigler’sdefinition. riers meet Bain’sdefinition. Scale economies and capital requirements are Ferguson’sdefinition follows Bain’s, but entry barriers according to Bain’sdefinition be- with the additional requirement that incumbents cause they seem to be positively correlated with earn monopoly profits. Pricing above marginal profits. Scale economies are not an entry barrier cost is not sufficient for incumbents to persis- according to Stigler’sdefinition provided en- tently earn above-normal profits. Incumbents trants and incumbents have equal access to tech- only earn above-normal profits if prices exceed nology. Capital requirements are not Stiglerian average cost. Prices may not exceed average entry barriers either, unless the incumbent never cost even though they exceed marginal cost paid them. because of price or quality competition among Fisher also claims that capital requirements incumbents. are not entry barriers according to his definition. Fisher’sdefinition follows those of Bain and Consider, as Fisher does, an industry that firms Ferguson, but it is normative rather than posi- can only enter if they make a large capital tive. For Fisher, an entry barrier is socially expenditure. A firm will not enter if the profits VOL. 94 NO. 2 WHEN ARE SUNK COSTS BARRIERS TO ENTRY? 463 that it anticipates in the long run will not be definition, while they are not according to sufficient to justify the initial capital cost. But, Stiglerian definitions, since all entrants must argues Fisher, this is exactly the calculation that bear them equally. Similarly, Carlton and Per- society would want the potential entrant to loff exclude sunk costs as entry barriers (under make. The capital expenditure would be so- their second definition) since there are no sunk cially wasteful if it did not guarantee a rate of costs in the long run. return that exceeded that which it could earn elsewhere. Fisher therefore concludes that, ac- II. Economic Analysis cording to his definition, capital requirements, no matter how large, are not entry barriers. As we have shown, the concept of an entry Note, however, that Fisher’s argument ig- barrier has a rich and confused heritage in eco- nores consumer surplus, which enters into soci- nomics. To clear up the confusion, we offer the ety’s calculation, but not into the potential following new classification of entry barriers. entrant’s calculation. The addition of another firm to the industry could increase competition, Definition 8: An economic barrier to entry is a and hence consumer surplus, enough to com- cost that must be incurred by a new entrant and pensate for the entrant’s profit loss in society’s that incumbents do not or have not had to incur. calculation. Governments have operated firms on this theory, to create price competition where Definition 9: An antitrust barrier to entry is a there would otherwise be a natural monopoly. cost that delays entry and thereby reduces social Capital requirements or scale economies may welfare relative to immediate but equally costly not constitute entry barriers according to von entry. Weizsacker’sdefinition. To prove this, von Weizsacker models an industry with scale econ- All economic entry barriers are antitrust bar- omies and shows that the number of active firms riers. However, many antitrust barriers are not in the Cournot equilibrium with free entry ex- economic barriers. Antitrust is a larger category ceeds the number of active firms that would than economic. maximize social surplus. The cost savings that When free entry leads to the efficient number arise from firms taking greater advantage of of firms, if a market has no antitrust entry bar- scale economies more than compensate for the riers, then it is efficient. If it has no economic reduction in total output from having fewer entry barriers, then it is eventually efficient. An firms. In such an industry, additional entry bar- antitrust entry barrier in a market that is other- riers enhance welfare. wise efficient reduces welfare relative to what it Capital requirements can be entry barriers would have been in the absence of that barrier. according to Gilbert’sdefinition, especially if a The presence of an antitrust entry barrier does significant proportion of them are sunk. Sunk not necessarily mean that a merger should be costs generate earnings that would be lost if a disallowed. The net change in welfare resulting firm exits the market; in this sense, sunk costs from the merger could still be positive. Rather, are exit barriers. Exit barriers can affect entry the presence of the antitrust barrier means that by influencing the incentives of incumbents. If welfare would be higher if that barrier did not incumbents cannot exit without considerable exist. losses, then their threats of aggressive post- In our analysis, we also find it useful to entry behavior are more credible, which deters distinguish between direct and reinforcing bar- entry and earns them higher profit. Thus, exit riers, as follows: barriers for incumbents create entry barriers. Moreover, sunk costs increase an entrant’s Definition 10: A primary barrier to entry is a losses in the event that entry fails, which makes cost that constitutes a barrier to entry on its own. the incumbent’s threats of aggressive post-entry behavior more frightening. Thus, exit barriers Definition 11: An ancillary barrier to entry is a for entrants create entry barriers. In these ways, cost that does not constitute a barrier to entry by sunk costs provide rents to incumbents and, itself, but reinforces other barriers to entry if hence, are entry barriers according to Gilbert’s they are present. 464 AEA PAPERS AND PROCEEDINGS MAY 2004

A group of small primary barriers may con- experiencing growth, the fractional issue is stitute a significant entry barrier. A group of clearly temporary. Thus, scale economies are small ancillary barriers does not commonly not economic, but may be antitrust, entry bar- constitute a significant entry barrier unless other riers if they delay entry and thereby reduce primary barriers are also present. However, in social welfare. some cases, large ancillary barriers can combine In a technical appendix to this paper (available and reinforce each other to form a large primary from the authors upon request), we have con- entry barrier. structed a model in which new firms face Cournot A particular ancillary barrier may produce a competition with incumbents if they choose to primary entry barrier only when combined with enter, and in which (i) scale economies do not a restricted class of other ancillary barriers, or delay entry on their own, (ii) brand loyalty reinforce only a restricted class of other primary delays entry on its own, and (iii) brand loyalty entry barriers. If a market possesses no entry delays entry even longer in the presence of scale barrier from either class, the ancillary barrier in economies. Thus, according to the model, scale question does not deter entry. economies are ancillary barriers that exacerbate the entry delay caused by brand loyalty. III. Scale Economies Does the additional delay caused by scale economies necessarily reduce social welfare? Bain argued that scale economies are an entry For an important class of demand functions barrier. Incumbents may have already built (including linear demand), social welfare under plants of efficient scale. If the added output of Cournot competition is higher than social wel- the entrant’sefficient plant is large relative to fare under monopoly, because the profit loss industry demand and existing output, price incurred by the incumbent is not large enough to could fall below the entrant’s per-unit cost, and offset the price reduction that benefits consum- entry could be unprofitable. ers. In these cases, scale economies are ancillary This argument assumes that the entrant ex- antitrust entry barriers, since they delay entry by pects the incumbent to maintain its pre-entry reinforcing the entry-deterrent effects of brand output level even after entry has occurred. Once loyalty, and thereby reduce social welfare. the new firm has entered, the incumbent may want to reduce its output to prevent its profits IV. Sunk Costs from falling to zero. But then the entrant’s prof- its might also be prevented from falling to zero, Many firms are capable of paying large cap- and entry might be ex ante profitable. ital costs if entry is worthwhile. If capital mar- However, the incumbent’s output reduction kets are efficient, raising capital is no more would prevent the potential entrant’s post-entry difficult for profitable large-scale projects than profits from falling to zero only if it caused for profitable small-scale ones. And even if cap- some fraction of customers to switch from the ital markets are inefficient, perhaps because of incumbent to the entrant. Customers may be asymmetric information about industry prospects, loyal to an existing brand because continuing to they do not necessarily fail in larger measure buy it involves less risk than trying a new one. with large-scale projects than with small-scale Therefore, scale economies deter entry only if ones. customers are sufficiently loyal to the incum- Capital-market imperfections favor wealthier bent’s brand. Hence, scale economies are ancil- and more experienced firms over entrepreneurs lary entry barriers that reinforce primary entry without track records, but the former are not barriers such as brand loyalty. necessarily the incumbents. Some entrants are We now argue that scale economies are an- large diversified firms that build new plants in titrust, not economic, entry barriers. Bain’s no- an industry. Microsoft entering the internet tion, that entry may be deterred because the browser business is an instance where the en- industry only has room for a fraction of the trant was larger than the largest incumbent. In efficiently scaled plant, is generally a short-run industries where the primary potential entrants phenomenon that does not persist as existing are large diversified firms, large capital costs are plants are replaced. Moreover, in an industry not entry barriers. VOL. 94 NO. 2 WHEN ARE SUNK COSTS BARRIERS TO ENTRY? 465

Nevertheless, capital costs can indirectly dis- lawsuits. Usually, a merger in a particular courage entry. Instead of being entry barriers in industry cannot permanently reduce competi- their own right, they often reinforce other entry tion if new firms can easily enter the industry. barriers, by magnifying the risks. Thus, when a Therefore, to prove that mergers are socially solid reputation is necessary to enter an indus- harmful, antitrust authorities must usually, at try, large costs make it difficult or impossible to the very least, demonstrate the presence of test the market; instead, the entrant must com- entry barriers. To do so, they must rely on a mit large resources to enter. If large sunk costs definition, and for this, they have often turned are associated with entry and entry is unsuccess- to economists. ful, the entrant’s losses are large. In such a Unfortunately, economists have not yet been setting, the threat of aggressive behavior by the able to reach broad consensus over the defini- incumbent may deter entry. The greater the po- tion of an entry barrier, and this has probably tential loss, the more frightening is the threat of hindered the development of efficient antitrust aggressive behavior. By magnifying risks, cap- policy. In the hope of facilitating consensus, we ital requirements reinforce other entry barriers. have highlighted two aspects of entry: (i) the Therefore, capital requirements are ancillary effect of alleged entry barriers on the timing of barriers, especially if a significant proportion of entry and (ii) the effect on the timing of entry of them are sunk. the interaction between different alleged entry Capital requirements are not economic entry barriers. barriers, since incumbents had to bear capital costs in the past similar in size to those that REFERENCES entrants have to bear today. However, they may nevertheless be antitrust entry barriers. Sunk Bain, Joe S. Barriers to new competition. Cam- costs cause firms to delay entry because of their bridge, MA: Harvard University Press, 1956. option value. The option of entering is lost once Carlton, Dennis and Perloff, Jeffrey. Modern in- the firm enters. With uncertainty about market dustrial organization. New York: HarperCol- conditions, this option has value. Thus, dynamic lins College Publishers, 1994. entry is delayed relative to a static world. Ferguson, James M. Advertising and competi- In the technical appendix to this paper (avail- tion: Theory, measurement, fact. Cambridge, able from the authors upon request), we argue MA: Ballinger, 1974. formally that sunk costs, like scale economies, Fisher, Franklin M. “Diagnosing Monopoly.” are ancillary antitrust entry barriers. We present Quarterly Review of Economics and Busi- a model in which (i) sunk costs do not delay ness, Summer 1979, 19(2), pp. 7–33. entry in the absence of uncertainty and (ii) un- Gilbert, R. “Mobility Barriers and the Value of certainty does not delay entry in the absence of Incumbency,” in R. Schmalensee and R. D. sunk costs, but (iii) uncertainty and sunk costs Willig, eds., Handbook of industrial organi- combine to delay entry until the realization of zation. Amsterdam: North-Holland, 1989, pp. uncertainty. For an important class of demand 475–535. functions, efficient entry is in advance of the re- Stigler, George J. The organization of industry. alization of uncertainty. In these cases, sunk costs Chicago, IL: University of Chicago Press, and uncertainty are ancillary antitrust entry barri- 1968. ers that combine and reinforce each other to pro- von Weizsacker, C. C. “A Welfare Analysis of duce a primary antitrust entry barrier. Barriers to Entry.” Bell Journal of Econom- ics, Autumn 1980, 11(2), pp. 399–420. V. Conclusion Wallace, Donald H. “Monopolistic Competition and Public Policy.” American Economic Re- The presence of entry barriers is the central view, March 1936 (Papers and Proceedings), subject of contention in numerous antitrust 26(1), pp. 77–87.