Economic and Antitrust Barriers to Entry

Transcription

1 Eonomi and Antitrust Barriers to Entry R. Preston MAfee, Hugo M. Mialon, and Mihael A. Williams 1 Deember 1, 2003 Abstrat We review the extensive literature on barriers to entry in law and eonomis; we introdue four onepts, namely eonomi, antitrust, primary, and anillary barriers to entry; we employ these onepts to lassify a set of well-known strutural harateristis of markets and ompetitive tatis by inumbents; and we apply the resulting insights to evaluate the verdits that were reahed in a set of landmark antitrust ourt ases in the US. Bain (1956) defined an entry barrier as anything that allows inumbent firms to earn above-normal profits without the threat of entry. To defend his ontention that large sale eonomies are an entry barrier, Bain argued that if inumbents at in onert and potential entrants expet inumbents to maintain their pre-entry output levels after entry has ourred, the neessity for firms to be large relative to the market in order to attain produtive effiieny allows inumbents to earn above-normal profits without the threat of entry. However, inumbents may find that their interests are best served by reduing their output levels one large sale entry has ourred, so that Bain's assumption that potential entrants expet inumbents to maintain their pre-entry output levels may not be realisti. Moreover, Stigler (1968) rejeted the basi notion that sale eonomies an reate an entry barrier. He defined entry barriers as osts that must be borne by a firm that seeks to enter an industry but is not borne by firms already in the industry. In any given industry, entrants and inumbents alike enjoy the same sale eonomies as they expand their output. Therefore, aording to Stigler's definition, sale eonomies are not an entry barrier. With respet to sale eonomies, and other market harateristis, the definitions of Bain and Stigler are at variane, whih has resulted in muh ontroversy among eonomists and antitrust lawyers over the definition of an entry barrier. The purpose of this artile is to lear up this onfusion by providing a thorough lassifiation of entry barriers. In setion 1, we trae the historial development, in eonomis and law, of the existing disarray of definitions of a barrier to entry. In setion 1 Preston MAfee and Hugo Mialon, Department of Eonomis, University of Texas at Austin, Austin, Texas, (mafee@eo.utexas.edu and mialon@eo.utexas.edu), Mihael Williams, Analysis Group, In., PM KeyPoint LLC, 2200 Powell Street, Suite 1080, Emeryville, CA (mwilliams@pmkeypoint.om).

2 2, we introdue four onepts, namely eonomi, antirust, primary, and anillary barriers to entry, and lassify a group of well-known strutural harateristis of markets and ompetitive tatis by inumbents aordingly. This lassifiation is original, and lears up most of the onfusion highlighted in setion 1. In setion 3, we onsider a set of antitrust ourt ases for whih the new lassifiation an be employed to evaluate the verdits that were reahed. Setion 4 summarizes and proposes avenues for further researh. I. HISTORY OF THE CONCEPT Many eonomists and legal sholars have attempted to define the onept of a barrier to entry, and this has produed a medley of definitions, several of whih address different issues, and several of whih lash. We begin by presenting, in hronologial order, the definitions that have been proposed in the eonomis literature. I.A. ECONOMICS Historially, the most ommon impediments to free entry into markets have been government monopoly grants and patents. Many governments have granted monopolies for the exlusive purpose of olleting government revenue. An example is the salt gabelle in China. At the turn of the entury, the right to manufature, transport, and sell salt was under strit governmental ontrol; the salt gabelle was levied at every stage from prodution to onsumption. Only liensed merhants ould deal in salt, and then only within limited presribed areas. The salt merhants of the ountry were monopolists for the benefit of the government, whih derived a very large revenue from this soure (Muhse, 1916). Governments have also granted monopolies to enourage soially benefiial invention. The earliest known English patent for invention was granted by Henry VI to Flemishborn John of Utynam in The patent gave John a 20-year monopoly for a method of making stained glass, required for the windows of Eton College, whih had not been previously known in England. In the time of the Tudors, the Crown ommonly granted monopolies to traders and manufaturers, inluding patents for invention, sometimes to royal favorites or for the purpose of replenishing royal offers (US Patent Offie, 2003). However, the use of the term "barrier," in relation to entry into a marketplae, originated in disussions of transit taxes, not government monopoly grants. In the Chinese Empire, for example, "liken" stations were physial barriers, guarded by loal offiials, where duties were olleted on partiular merhandise in transit from one part of the empire to another. In 1911, five liken stations barred the water route between Shanghain and Soohow, a distane of no more than eighty miles (Williams, 1912). The Imperial Maritime Customs managed similar stations at all the major ports in the empire, whih barred foreigners from entering the ountry with foreign goods without paying trade tariffs. Unlike liken stations, ustoms stations olleted tariffs at rates that were set by international treaties with Britain, Frane, and the United States. 2

3 While government patents and monopoly grants are barriers to the reation of new enterprises, transit and trade tariffs are barriers to the expansion of existing enterprises into new markets. Eonomists would eventually ome to refer to both types of barriers jointly under one heading, "barriers to entry," and inlude many impediments to ompetition other than government monopoly grants and trade tariffs under this heading. Donald H. Wallae has the distintion of being the first eonomist to use the term in an artile published in an aademi journal. In the Papers and Proeedings of the Fortyeighth Annual Meeting of the Amerian Eonomi Assoiation, published in Marh 1936, he explains that the key priniples of publi poliy that emerged from the literature on monopolisti ompetition, initiated by Chamberlin (1935), are that "Competitive measures whih did not truly measure effiieny should be eliminated; and, by impliation at least, any other barriers to free entry exept those inherent in differing personal qualities or ability to obtain apital should be removed" (p. 79). Wallae goes on to lament the neglet in the existing literature of other important barriers to entry: "Publi poliy seems to have overlooked suh important barriers to free entry as ontrol of sare resoures of raw materials,... and the impressive formidability of size and length or purse supplemented by industrial and finanial affiliations" (p.80). He expresses his belief, whih would be shared by many eonomists after him, that large apital requirements are also an important barrier to entry that warrants the srutiny of antitrust authorities. Wallae onludes his artile with a researh program that would prove to be visionary: "The nature and extent of barriers to free entry needs thorough study" (p.83). Fifteen years later, Joe S. Bain would publish a series of artiles ulminating in a book that would onstitute the first thorough study of barriers to entry. Bain (1949), seeking to explain the empirially observed tendeny in some ollusive oligopolisti industries (suh as those for igarettes and steel) to hold prie below the level that would maximize short run profits in the industry, introdued the limit-prie model of entry deterrene. The limit prie is the highest prie that inumbent firms an harge without induing at least one other firm to enter the market. Inumbents estimate the market share they would lose to an entrant, and also the onditions of ompetition that they would fae after entry. They ompare these estimates to the profits they would lose by setting the limit prie rather than the short run profit maximizing prie. Bain explains that inumbents might want to set eah short run prie (and hene long run average prie) at a lower level than the one that maximizes industry profit, in order to disourage entry, and keep the smaller (and non-maximized) profits all for themselves. These insights lead Bain (1950) to look for market onditions under whih firms would want to sarifie short-run profits by limit priing. The author noted that a ruial determinant of market onditions is freedom of entry. He hypothesized that if new firms annot easily enter the market, inumbents maximize short run profits, while if firms an easily enter the market, inumbents sarifie short run profits to deter entry. He then identified three broad strutural market harateristis that might restrit freedom of entry. Entry into the market might be diffiult beause (1) inumbents have patents on prodution proesses or ontrol of ruial resoures, (2) inumbents enjoy substantial ost advantages over potential entrants, whih inlude advantages in prodution osts as 3

4 well as established produt preferenes for going firms, or (3) the sale of an optimum firm is very large relative to the market and the eonomies of sale are great. Four years later, Bain (1954) published a paper in whih he eluidated the logi by whih he ame to believe that large sale eonomies are a barrier to entry. Suppose a firm must add signifiantly to industry output in order to be effiient, and inumbent firms are ommitted to maintain their output levels in the event of entry. If a firm enters this market at less than the effiient sale, it enters at a signifiant ost disadvantage relative to inumbent firms. If the firm enters at or above the effiient sale, then the ombined industry output would exeed industry demand ausing the industry selling prie to fall and dissipating all profits for the entrant. Therefore, firms in industries where the effiient sale is large relative to the market may be able to earn onsiderable profits without induing entry. Bain alled this effet of sale eonomies on barriers to entry the "perentage effet," beause it reflets the importane of the proportion of industry output supplied by a firm of effiient sale. He suggested that this is only one of two effets of sale eonomies on barriers to entry. Sale eonomies may be important to entry also beause large absolute amounts of apital are required for effiieny. That is, absolute apital requirements may so large that relatively few entrepreneurs ould seure the required apital, or that entrants ould seure it only at interest rates that plaed them at an important ost disadvantage to inumbents. In the proess of defending his view that sale eonomies and apital requirements pose important barriers to entry, Bain formulated the first general definition of a barrier to entry, whih he offered in the introdutory hapter of his 1956 book, "Barriers to New Competition." Definition 1 (Bain, 1956, p. 3). A barrier to entry is an advantage of established sellers in an industry over potential entrant sellers, whih is refleted in the extent to whih established sellers an persistently raise their pries above ompetitive levels without attrating new firms to enter the industry. Pries would settle down to their ompetitive levels if new firms were free to enter the industry. At their ompetitive levels, pries are equal to marginal ost. Aording to Bain, a barrier to entry is anything that allows inumbents to raise pries above marginal ost, whih usually entails above-normal profits, without induing entry of new firms. As Visusi et al (1992) point out, a problem with this definition is that it is tautologial. Bain defines a barrier to entry in terms of its outome, the extent to whih inumbents prie above marginal ost or earn above-normal profits without induing entry, whih he alled the "ondition of entry." The definition is true by virtue of the meaning of the ondition of entry alone, without referene to external fat, and its denial results in selfontradition. Although not theoretially sound, this definition might have been fashioned for the purpose of identifying barriers to entry empirially. If the ondition of entry were 4

5 observable, then Bain might have been able to identify the extent of barriers to entry aross industries. However, Bain ould find no immediate observable proxy for the ondition of entry. So he simply measured, for a ross-setion of twenty industries, the size and importane of the market harateristis that he believed to have an important effet on the ondition of entry: eonomies of sale, apital requirements, absolute ost advantages, and differentiation advantages. For example, to measure the perentage effet of eonomies of sale, he used Census data on the perentage of national industry apaity ontained in one plant of minimum effiient size. He found, for example, that this statisti was 5 to 6 perent in the igarette industry and 1 to 2.5 perent in the steel industry (Table 3, p. 72). He also measured absolute ost requirements by asking exeutives of various firms in the twenty industries questions related to the probable investment neessary to establish one plant of minimum effiient sale in eah of the twenty industries. He found that the absolute apital requirement for establishing one plant of minimum effiient size was 125 to 150 million dollars in the igarette industry, and 265 to 665 million dollars in the steel industry (Table 13, p.158). Relative to other industries, Bain found that apital requirements were high in the steel and igarette industries, and eonomies of sale were average in the steel industry, and low in the igarette industry (Table 14, p. 169). Whether sale eonomies and apital requirements atually had an effet on the ondition of entry in the igarette, steel, and other industries, and hene whether they atually were barriers to entry, Bain answered only in theory. While admiring Bain's important empirial ontributions, Nobel laureate George S. Stigler rejeted Bain's basi ontention that sale eonomies and apital requirements are barriers to entry, and developed a more useful definition in defending his point of view. Definition 2 (Stigler, 1968, p. 67). A barrier to entry is a ost of produing (at some or every rate of output) whih must be borne by firms whih seek to enter an industry but is not borne by firms already in the industry. Stigler's definition avoids tautology by identifying an entry barrier in terms of its fundamental harateristis, emphasizing the differential osts between inumbents and entrants. However, the present tense "is" in the definition is ause for onfusion. Suppose entrants have to bear a ost that inumbents do not have to bear today, but had to bear in the past (when they entered). Is this ost a barrier to entry? Stigler most likely would have answered in the affirmative, and one an safely assume that Stigler meant to define a barrier to entry as a ost that entrants have to bear, but inumbents do not, or have not had to, bear. Aording to Stigler's definition, a barrier to entry exists only if the potential entrant's long-run osts after entry are greater than those of the inumbent. Stigler's definition is narrower than Bain's definition, that is, some things are barriers to entry aording to Bain, and not aording to Stigler; but nothing is a barrier to entry aording to Stigler, 5

6 and not aording to Bain. In any given industry, entrants and inumbents enjoy the same sale eonomies as they expand their output. With equal aess to tehnology, eonomies of sale are not a barrier to entry aording to Stigler; but they are a barrier to entry aording to Bain (via their perentage effets). Absolute apital requirements are not a barrier to entry either, aording to Stigler, unless the inumbent never paid them; but they are a barrier to entry aording to Bain, for they seem to be positively orrelated with high profits. The spirit of Bain's definition did not fade after Stigler proposed an alternative definition. Ferguson (1974), who was mainly onerned with the question of whether advertising is a barrier to entry, proposed a definition that follows Bain's, but with the additional requirement that inumbents earn monopoly profits. Definition 3 (Ferguson, 1974, p.10). A barrier to entry is a fator that makes entry unprofitable while permitting established firms to set pries above marginal ost, and to persistently earn monopoly return. Ferguson pointed out that priing above marginal ost in the long run is not suffiient for inumbent firms to persistently earn above-normal profits. Inumbents only earn above-normal profits if pries exeeds average ost. Pries may not exeed average ost even though they exeed marginal ost beause of prie or non-prie ompetition among existing firms. For example, existing firms might ompete through advertising. Then potential entrants might be required to pay large fixed advertising osts to enter the industry. However, inumbents also pay these fixed advertising osts to ompete in the industry. These osts inrease the average ost urves of inumbents, as well as entrants (without affeting their marginal ost urves). As long as they are not a soure of sale eonomies, even if they allow inumbents to set pries above marginal ost, they are not a barrier to entry aording to Ferguson's definition, beause they inrease inumbents' average ost, thereby dissipating their above-normal profits, and hene reduing the inentives of potential entrants to enter the industry. In ontrast, they are a barrier to entry aording to Bain simply beause they allow inumbents to prie above marginal ost without induing entry. The definitions of Bain, Stigler, and Ferguson all fous attention on the different opportunities faing insiders and outsiders. Aording to Baumol (1982), these definitions divert attention away from other important barriers to entry, suh as legal restritions. For example, Baumol argues that the legal restrition that drivers must own an offiial medallion before supplying taxi servies is a barrier to entry into the taxi industry if the medallion is ostly, beause it redues the supply of taxi servies. 2 However, both inumbent and entrant had to bear the ost of the medallion, so Stigler's 2 Here, Baumol is using a more literal definition of a barrier to entry: a barrier to entry is anything that redues entry. 6

7 definition fails to identify the barrier to entry. Moreover, sine the prie of the medallion redues profits, the definitions of Bain and Ferguson also fail to identify the barrier. Moreover, a barrier to entry in the taxi industry ould even inrease soial welfare insofar as it redues exessive traffi. Baumol's main point of ontention with existing definitions is that they do not expliitly aount for the possibility that a barrier to entry may enhane soial welfare, and hene, given the negative onnotation generally attahed to the term "barrier to entry," taitly support the presumption that a barrier to entry neessarily redues welfare. By preserving monopoly profits, patents enourage researh and development of new produts and proesses, whih ould be soially benefiial. That firms in an industry with barriers to entry are earning high profits, or these firms have lower osts than outsiders, is not neessarily an indiation that soial welfare would be higher if the barriers to entry were absent. Fisher (1979), dissatisfied with existing definitions for muh the same reasons that Baumol was dissatisfied with them, proposed another definition, whih is in the spirit of Bain's and Ferguson's definition, but is normative rather than positive. Definition 4 (Fisher, 1979, p. 23) A barrier to entry is anything that prevents entry when entry is soially benefiial. Aording to Fisher, a barrier to entry exists if inumbents earn profits that are unneessarily high, in the sense that soiety would be better off if they were ompeted away, but firms do not enter to do this. To determine whether a potential barrier to entry auses profits to be unneessarily high, Fisher asks whether potential entrants make a alulation that is any different from the one that soiety would want them to make in order to deide whether to enter a market, given this barrier to entry. Consider, for example, an industry that firms an only enter if they make a large apital expenditure. A firm will not enter if the profits that it antiipates in the long run will not be suffiient to justify the initial apital requirement. But this is exatly the alulation that soiety would want the potential entrant to make. The apital expenditure would be soially wasteful if it did not guarantee a rate of return that exeeded the rate of return that it ould earn if it were invested elsewhere. Therefore, aording to Fisher's definition, an initial apital requirement, no matter how large, is not a barrier to entry. It is not a barrier to entry aording to Stigler's definition either, but only beause inumbent and entrant both had to pay it in the same amount. whih is an entirely different reason. Von Weizsaker (1980, 1) proposed a seond normative definition, whih is based on Stigler's rather than Bain's definition, in that it fouses on the differential osts between inumbents and entrants, rather than on the profits of inumbents. Definition 5 (Von Weizsaker, 1980, p. 400). A barrier to entry is a ost of produing that must be borne by a firm whih seeks to enter an industry but is not borne by firms 7

8 already in the industry and that implies a distortion in the alloation of resoures from the soial point of view. Von Weizsaker argues that a ost differential is a barrier to entry only if it results in a derease in welfare. His point is that the number of firms in a Cournot industry an be greater than the soially optimal number of firms. To prove his point, he develops a model of an industry with eonomies of sale, and shows that the number of ative firms in the Cournot equilibrium with free entry, defined as the largest number of firms suh that the Cournot equilibrium is still profitable, exeeds the number of ative firms that would maximize soial surplus, defined as the sum of onsumer surplus and market profit at the level of total industry output that arises when all firms set prie equal to marginal ost. In this model, eonomies of sale are not a suffiient barrier to entry. Welfare would inrease if the number of firms were limited to less than the free entry number. The ost savings that arise with fewer firms from taking advantage of eonomies of sale more than ompensate for the redution in total output from having fewer firms. In suh an industry, additional barriers to entry ould enhane welfare, by reduing the number of firms to their soially optimal level. However, industries where the number of firms is greater that the soially optimal number of firms are generally diffiult to identify. 3 The definitions of Stigler and von Weizsaker fous on the ost disadvantages of entrants relative to inumbents. Gilbert (1989) argues that suh definitions are unneessarily onfining, and proposes a new definition that fouses on the advantages of inumbents rather than the disadvantages of entrants. Definition 6 (Gilbert, 1989, p. 478). An entry barrier is a rent that is derived from inumbeny. 3 The real-estate industry might be one example. Aording to Hsieh and Moretti (2003), this industry has few barriers to entry and the brokerage ommission paid to real-estate agents is always a fixed 6 perent of the selling prie of the house. The prie of a typial house in Boston has long been muh higher than the prie of a typial house in Minneapolis. Sine the ommission rate is fixed, this implies that the brokerage fee from selling a typial house in Boston is muh higher than that in Minneapolis. However, beause the industry has few barriers to entry, there are more real-estate agents in Boston, even though the total number of homes sold eah year is higher in Minneapolis. Therefore, the average real-estate broker in Minneapolis is muh more produtive than the average broker in Boston. Even though the prie of a typial house is muh higher in Boston, real-estate agents are no better off in Boston than in Minneapolis. The higher ommissions in Boston are simply wasted through entry of real-estate agents seeking to earn higher ommissions, agents who ould be engaged in other profitable ativities. A larger number of agents in Boston, higher agent produtivity in Minneapolis, and real wages of agents that are no higher in Boston than in Minneapolis, may all be indiations that there are more than the soially optimal number of agents in the Boston real-estate industry. 8

9 Aording to Gilbert, a barrier to entry is the additional profit that a firm an earn as a sole onsequene of being established in the industry. An inumbent may be able to earn profit and exlude entry not only beause of ost advantages over entrants. Suppose the inumbent an ommit itself to produing the monopoly output, and this being the ase, no other firm an enter at a profit. 4 Then entry is exluded in this market even though the inumbent has no ost advantage over a new entrant, sine both had to pay the sunk osts. Sunk osts are a barrier to exit for the inumbent, whih allows it to ommit to a level of output, whih in turn deters entry, earning the inumbent a rent. Thus, exit barriers for inumbents reate entry barriers for entrants. Moreover, Gilbert argues that inumbents an use strategi behavior to exploit sunk osts to their advantage. Sunk osts inrease the entrant's loss in the event that entry fails, whih makes the inumbent's threats of strategi entry deterrene more effetive. Thus, exit barriers for entrants reate entry barriers for entrants. In these ways, sunk osts provide a rent to inumbents, and hene are a barrier to entry aording to Gilbert's definition. The legal restrition that drivers must buy an offiial medallion from ity authorities before supplying taxi servies may be a barrier to entry aording to Gilbert's definition for the same reason, while it is not a barrier to entry aording to the definitions of Bain, Stigler, Ferguson, Fisher, or von Weizsaker. Disagreement over the definition of a barrier to entry has persisted. Authors of modern textbooks in industrial organization openly doument the lak of onsensus (see for example, Visusi et al, 2000, p , and Churh and Ware, 1999, p ). In a popular textbook, Carlton and Perloff (1994) propose a literal definition of a barrier to entry. Definition 7 (Carlton and Perloff, 1994, p. 110). A barrier to entry is anything that prevents an entrepreneur from instantaneously reating a new firm in a market. A longrun barrier to entry is a ost that must be inurred by a new entrant that inumbents do not (or have not had to) bear. The authors argue that the first definition is rarely useful in pratie, for it implies that any apital requirement is a barrier to entry and that any industry in whih entry takes time has a barrier to entry. They note that the term "barrier to entry" is often used to refer to both osts of entering and the time required to enter. However, to our knowledge, they are the first to propose a definition that expliitly inludes a time dimension. 5 Unfortunately, they avoid the timing issue thereafter by onsidering only barriers to entry in the long run. Entry erodes profits in the long run. Therefore, if a firm earns profits in the long run, the industry must have long run barriers to entry. The authors argue that a firm an only earn profits in the long run if they have an advantage over potential entrants, whih leads them to adopt a modern version of Stigler's definition. Notie that 4 In this ase, Bain's assumption that entrants expet inumbents to maintain their preentry output levels even after entry has ourred is valid. 5 Shephard (1997) also distinguishes between the extent and speed of entry (p. 209), but does not expliitly inorporate speed into a definition of an entry barrier. 9

10 their version lears up the onfusion about the present tense "is" in Stigler's original definition. In another popular textbook, Churh and Ware (1999) distinguish between strutural and strategi barriers to entry, reserving the term "barrier to entry" only for the former. Definition 8 (Churh and Ware, 1999, p. 487). An entry barrier is a strutural harateristi of a market that protets the market power of inumbents by making entry unprofitable. Most definitions before this one were impliitly intended to apply mainly to strutural market harateristis anyway. That is not to say that these definitions ould not, in priniple, be applied to strategi behavior also. However, most strategi behavior involves sarifie by inumbents in order to inflit losses on entrants. Thus strategi behavior is never a barrier to entry aording to any definition that is inspired from Stigler's. To justify his fous on strutural barriers to entry, Von Weizsaker (1980, 2) onjetures that entry deterrene strategies are not even available to inumbents without strutural barriers to entry. This view is extreme. The musi animation industry has no signifiant strutural barriers to entry, yet Disney has suessfully and persistently deterred entry into the industry through a host of ompetitive tatis (need itation). More likely, ompetitive tatis by inumbents intertwine with the strutural harateristis of the industry to reate barriers to entry. We have seen that the onept of a barrier to entry has a rih and onfused heritage in eonomis. In attempting to define the term, eonomists have made at least eight fruitful distintions: (1) barriers to the expansion of existing firms versus barriers to the reation of new firms, (2) inumbents earning high profits versus entrants having greater osts than inumbents, (3) inumbents priing above marginal ost versus inumbents earning profits that are above normal, (4) entry being deterred versus soial welfare being redued, (5) advantages of inumbents versus disadvantages of entrants, (6) barriers to entry versus barriers to exit, (7) speed of entry versus extent of entry, and (8) strutural versus strategi barriers to entry. I.B. LAW The most basi antitrust laws were laid out in the Sherman At of 1890 and the Clayton At of By setion 1 of the Sherman At, ollusive ations suh as prie-fixing, market rigging, and sales-alloating shemes, and other praties in restraint of trade are illegal: 10

11 Sherman At, Setion 1. Every ontrat, ombination in the form of trust or otherwise, or onspiray, in restraint of trade or ommere among the several States, or with foreign nations, is delared illegal. By setion 2 of the Sherman At, monopolization or attempt to monopolize is illegal: Sherman At, Setion 2. Every person who shall monopolize, or attempt to monopolize, or ombine or onspire with any other person or persons, to monopolize any part of the trade or ommere among the several States, or with foreign nations, shall be deemed guilty of a felony. By setion 7 of the Clayton At, mergers that substantially redue ompetition are illegal: Clayton At, Setion 7. No person engaged in ommere or in any ativity affeting ommere shall aquire, diretly or indiretly, the whole or any part of the stok or other share apital and no person subjet to the jurisdition of the Federal Trade Commission shall aquire the whole or any part of the assets of another person engaged also in ommere or in any ativity affeting ommere, where in any line of ommere or in any ativity affeting ommere in any setion of the ountry, the effet of suh aquisition may be substantially to lessen ompetition, or to tend to reate a monopoly. II. ECONOMIC ANALYSIS As we have seen, the onept of an entry barrier has a rih and onfused heritage in eonomis. To lear up the onfusion, we offer a new lassifiation of entry barriers: Definition 8. An eonomi barrier to entry is a ost that must be inurred by a new entrant and that inumbents do not or have not had to inur. Definition 9. An antitrust barrier to entry is a ost that delays entry, and thereby redues soial welfare relative to immediate but equally ostly entry. Most eonomi entry barriers are antitrust entry barriers. However, many antitrust entry barriers are not eonomi. Indeed, the eonomi definition derives from Stigler s work, whih served to omplete the edifie of Chiago antitrust thought. The Chiago Shool has onsistently argued against the need for draonian measures against monopoly and ollusion, suh as those in the Sherman and Clayton Ats (see Posner, 1979). No surprise, then, that their definition of a barrier to entry is striter than those that the legal authorities had in mind when they enated the draonian measures. 11

12 When free entry leads to the effiient number of firms, if a market has no antitrust entry barriers, then it is effiient. If it has no eonomi entry barriers, then it is eventually effiient. An antitrust entry barrier in a market that is otherwise effiient redues welfare relative to what it would have been in the absene of that barrier. The presene of an antitrust entry barrier does not neessarily mean that a merger should be disallowed. The net hange in welfare resulting from the merger ould still be positive. Rather, the presene of the antitrust barrier means that welfare would be higher if that barrier did not exist. In our analysis, we also find it useful to distinguish between diret and reinforing barriers: Definition 10. A primary barrier to entry is a ost that onstitutes a barrier to entry on its own. Definition 11. An anillary barrier to entry is a ost that does not onstitute a barrier to entry by itself, but reinfores other barriers to entry if they are present. A group of small primary barriers may onstitute a signifiant entry barrier. A group of small anillary barriers do not ommonly onstitute a signifiant entry barrier unless other primary barriers are also present. However, in some ases, large anillary barriers an ombine, and reinfore eah other, to form a large primary entry barrier. A partiular anillary barrier may produe a primary entry barrier only when ombined with a restrited lass of other anillary barriers, or reinfore only a restrited lass of other primary entry barriers. If a market possesses no entry barrier from either lass, the anillary barrier in question does not deter entry. We now employ the onepts introdued in this and the previous setion to assess the nature of the barriers to entry posed by sale eonomies, apital requirements, and several of the other usual suspets. III. THE USUAL SUSPECTS In this setion, we provide a atalogue of industry harateristis and ompetitive tatis that inhibit or enourage the entry of new firms into the industry. III.A. SCALE ECONOMIES With aess to redit, an entrant ould easily build a plant of minimum effiient sale. The problem is that inumbents have already built plants of minimum effiient sale. If the added output of the entrant s plant of minimum effiient sale is large relative to industry demand and existing output, the produt prie would fall below the entrant s per unit ost, so that entry would be unprofitable. 12

13 However, this argument assumes that the new firm expets the inumbent to maintain its pre-entry output level even after entry has ourred. One the new firm has entered, the inumbent may want to redue its output from its pre-entry level, to prevent its profits from falling to zero. But then the entrant s profits might also be prevented from falling to zero, so that entry might be ex ante profitable. However, this requires some buyers to swith from the inumbent firm to the new entrant. Swithing from an IBM omputer system to that of a new rival may ause the business buyer to inur added osts for new software or for employee retraining. If suh swithing osts are high, then entry will not be profitable. On the other hand, the new firm ould enter and slightly underut the inumbent s prie. It would then get all of market demand, and entry would be profitable, provided the new firm indued all onsumers to swith to buying its produt by setting a slightly lower prie. Consumers may be loyal to existing brands, and for good reason. Rational onsumers who have had experiene with the existing brand may deide not even to try a new brand introdued at the same prie and of equal ex ante attrativeness, for one the brand has been used, ontinuing to buy it involves less risk than trying the new brand. In order to offset brand loyalty, a new firm would have to offer a onsiderable prie disount to lure onsumers away. But at this disount, entry might no longer be profitable. Therefore, sale eonomies are anillary barriers to entry that reinfore other barriers to entry, suh as ustomer swithing osts and brand loyalty. Whether sale eonomies are eonomi barriers to entry depends on whether swithing osts and brand loyalty are eonomi barriers to entry. The swithing osts borne by entrants today are usually omparable to those that were borne by inumbents bak when they entered the market, unless these inumbents were the pioneers. Thus, ustomer swithing osts are not usually eonomi barriers to entry. 6 On the other hand, brand loyalty seems to onfer a definite advantage to an inumbent over potential entrants, whih may lead one to onlude that it is an eonomi barrier to entry. However, this advantage may have been ostly for the inumbent to aquire, or it 6 Indeed, if sale eonomies are low, swithing osts may even be entry boosters rather than barriers. Farrell and Shapiro (1988) analyze an overlapping generations model, in whih two infinitely lived firms ompete over prie in the presene of buyer swithing osts, and buyers live for two periods. If swithing osts are greater than sale eonomies, the inumbent exploits his loked-in buyers and onedes the new buyers to the entrant (a Fat-Cat Effet). Although swithing osts make it harder for entrants to attrat attahed buyers, they atually enourage entry to serve unattahed ones. However, if sale eonomies are greater than swithing osts, the inumbent firm keeps the potential entrant out in equilibrium. The swithing osts protet the inumbent from the entrant s ompetition for attahed buyers, while the eonomies of sale make it unattrative for the entrant to enter and serve only the unattahed buyers. This suggests that sale eonomies and swithing osts might both be anillary barriers to entry that reinfore eah other, as well as other primary barriers to entry. 13

14 may be relatively easy for potential entrants to overome. Brand loyalty is an eonomi barrier to entry only if it provides the inumbent with an advantage that is more expensive for potential entrants to overome than it was for the inumbent to aquire. This test is more stringent. Consumers may view purhases of baby food as partiularly risky, so that brand loyalty in baby food may be partiularly diffiult for entrants to overome; but firms expenditures on advertising may also have to be partiularly large in order to aquire brand loyalty in the market for baby food, so that, on the whole, brand loyalty in this market may not be an eonomi barrier to entry. Even if brand loyalty is not more expensive for the inumbent to aquire than for potential entrants to overome, one might nevertheless argue that brand loyalty is an antitrust barrier to entry if it redues welfare by delaying entry. Consumer loyalty an redue onsumer welfare only if onsumers are ignorant of some underlying bioequivalene of the various brands in the market. Possibly, buyers may refuse to buy from a new entrant even though its brand is bio-equivalent to the inumbent s brand beause they are not informed of this bio-equivalene, whih is a market failure that might require intervention by the ourts. One way for the ourts to inform onsumers about the homogeneity of brands is mandatory trademark liensing, sine trademarks provide a good deal of information quikly to one who has experiene with it. With mandatory trademark liensing, the ourts an quikly inform onsumers without their knowing it. We now argue more formally that sale eonomies are anillary, antitrust barriers to entry. To do so, we present a simple model in whih (1) sale eonomies do not delay entry on their own, (2) brand loyalty delays entry on its own, and (3) brand loyalty delays entry even longer in the presene of sale eonomies. Consider a one-shot entry game. A potential entrant first hooses whether or not to enter a market. If it hooses not to enter, the sole inumbent ats as a monopolist. If it hooses to enter, the entrant and inumbent play a Cournot duopoly game. The entrant and inumbent both have the same ost funtion Cq ( ) q f, where is marginal ost, and f is fixed ost (the simplest expression of sale eonomies). Note that inumbent and entrant both have to bear the fixed ost f. Therefore, f is ertainly not an eonomi barrier to entry in this model. The inumbent s inverse demand funtion is given by Pq ( ) 1 Q, where Q is the total quantity produed by the industry, that is, Q qi qe if the potential entrant hooses to enter the market, and Q qi otherwise, where q I and q E are the inumbent s and entrant s quantity hoies, respetively. The potential entrant s inverse demand funtion, if it hooses to enter the market, is given by Pq ( ) 1 Q, where is a measure of onsumers loyalty to the inumbent s brand. Note that if deters entry, then it is an eonomi barrier to entry, sine in this model it is a ost to the entrant but not to the inumbent. Case 1. 0 and f 0 In the subgame that follows entry, the inumbent and entrant s maximization problem is 14

15 (1) max (1 Qq ) q f q i i i The equilibrium quantity hoies of the inumbent and entrant are (2) q I 1 qe 3 Therefore, the inumbent and entrant s equilibrium profits are given by (3) 2 (1 ) I E 9 f Hene, the potential entrant hooses to enter if and only if (4) f (1 ) 9 2 Note that if the marginal and fixed osts are small enough, inequality (4) is satisfied, so that the fixed ost never deters entry, in the absene of brand loyalty. In other words, for the parameter ranges defined by (4), sale eonomies are not primary barriers to entry. Case 2. 0 and f 0 In this ase, the entrant s maximization problem is (5) max (1 Q ) q q qe E E The first order ondition yields (6) q E 1 qi 2 The inumbent s maximization problem is (7) max (1 Qq ) q q E E E Here, the first order ondition yields (8) q I 1 qe 2 Solving (7) and (8) simultaneously yields 15

16 (9) q I 1 and 3 q E Therefore, the potential entrant s profits, if it hooses to enter the market, are given by (10) E (1 2 ) 9 2 The equation E 0 has the following root: 1 (11) 1 2 Therefore, the potential entrant hooses to enter if and only if (12) 1 2 where the quantity on the right hand side is the monopoly output. When brand loyalty is large enough, inequality (12) is not satisfied, and so entry is deterred, even in the absene of sale eonomies. In this ase, brand loyalty is a primary, eonomi barrier to entry. Case 3. 0 and f 0 In this ase, the entrant maximization problem is (13) max (1 Q ) q q f qe E E And the inumbent s maximization problem is (14) max (1 Qq ) q f qe E E The solutions to (13) and (14) are the same as the solutions to (5) and (7): (15) q I 1 and 3 q E Therefore, the potential entrant s profits, if it hooses to enter the market, are given by (16) E 2 (1 2 ) 9 f The equation E 0 now has the following two roots: 16

17 (17) 1 3 1,2 2 f Therefore, the potential entrant hooses to enter if and only if (18) 1 3 f Hene brand loyalty deters entry for a larger range of parameters with sale eonomies than without them. Can this imply that brand loyalty delays entry longer with sale eonomies than without them? The model does not have an expliit time dimension, but we an nevertheless address the issue of entry delay indiretly by onsidering how the model s parameters might hange over time. Suppose that tehnologial innovation in input markets will ontinuously redue the industry s marginal ost for all of its partiipants. Then, entry would eventually take plae, all else approximately onstant, for as dereases, the inequalities in (18) are more likely to be satisfied. But entry would take plae later with sale eonomies than without them, sine the first inequality in (18) is striter than the seond. Does the additional delay in entry oasioned by sale eonomies neessarily redue soial welfare? For an important lass of demand funtions (inluding linear demand), soial welfare under Cournot ompetition is higher than soial welfare under monopoly, beause the profit loss inurred by the inumbent is not large enough to offset the prie redution that benefits onsumers. In these ases, sale eonomies are anillary, antitrust barriers to entry, sine they delay entry by reinforing the entry deterrent effets of brand loyalty, and thereby redue soial welfare. III.B. SUNK COSTS The neessity for firms to be large relative to the market in order to attain produtive effiieny reinfores barriers to entry suh as brand loyalty and ustomer swithing osts. This is the perentage effets of sale eonomies on barriers to entry. Sale eonomies may also affet entry beause the absolute amount of apital required for effiieny may be so large that relatively few entrepreneurs ould seure the required apital, or that entrants ould seure it only at interest rates that plaed them at an important ost disadvantage relative to inumbents. However, many firms are apable of paying large apital osts, if the entry is worthwhile. Raising money for large projets is not neessarily more diffiult than raising money for small projets. If apital markets work properly, raising apital should 17

18 be no more diffiult for a profitable large-sale projet than for a profitable small-sale projet. Profitable projets should attrat many investors. If apital markets do not work properly, prospetive entrants may not be able to pay the large apital osts assoiated with entry even if entry is worthwhile, but inumbents may not be able to pay the large osts assoiated with replaing existing, depreiated apital either. Capital market imperfetions favor wealthier and more experiened firms over entrepreneurs without trak reords, but the former are not neessarily the inumbents. Some entrants are large, diversified firms that build new plants in a new industry. 7 Mirosoft entering the internet browser business and Sony entering the videogame business are instanes where the entrant was larger than the largest inumbent. Large apital requirements may be entry boosters rather than barriers for large diversifying firms, beause they an enter knowing that they won't be bothered by pesky smaller ompetition. Nevertheless, large apital requirements an indiretly disourage entry. Instead of being barriers to entry in their own right, apital requirements often reinfore other barriers to entry, by making the risks larger. Thus, when a solid reputation is neessary to enter an industry, large osts make it diffiult or impossible to test the market; instead, the entrant must ommit large resoures to enter. If large sunk osts are assoiated with entry and entry is unsuessful, the entrant s losses are large. In suh a setting, the threat of aggressive behavior by the inumbent may deter entry. The greater the potential loss, the more potent is the threat of aggressive behavior. By magnifying risks, apital requirements reinfore other barriers to entry. Therefore, apital requirements are anillary barriers to entry, espeially if a signifiant proportion of them are sunk. Capital osts are not anillary, eonomi barriers to entry, sine inumbents had to bear apital osts in the past similar in size to those that entrants have to bear today. Capital osts are analogous to an admission fee, whih must be borne by any firm that enters the industry. Although apital osts are not eonomi barriers to entry, they may nevertheless be antitrust barriers to entry. Sunk osts ause firms to delay entry beause of their option value. The option of entering is lost one the firm enters. With unertainty about market onditions, this option has value. Thus, dynami entry is delayed relative to a stati world. We now argue more formally that sunk osts are anillary, antitrust barriers to entry. To do so, we present a simple model in whih (1) sunk osts do not delay entry in the absene of unertainty, (2) unertainty does not delay entry in the absene of sunk osts, but (3) unertainty and sunk osts ombine to delay entry. 7 Although in most industries usually no more than 4 perent of entrants are large diversifying firms, these large entrants are typially the primary engines of industry growth. Baldwin and Gu (2003) find that in Canada almost all the ontribution of plant turnover to produtivity growth is due to more produtive new plants of multi-plant firms displaing existing plants of multiplant firms, suggesting that small independent singleplant firms have had little impat on aggregate produtivity. 18

19 Consider a two-period entry deterrene model in whih a prospetive entrant is unertain about the inumbent s type. The inumbent is either aggressive, with probability, or weak, with probability 1. The aggressive inumbent never aommodates. In period 1, the potential entrant hooses whether or not to enter, not knowing the inumbent s type. If the potential entrant enters, the weak inumbent hooses whether or not to aommodate. If the inumbent does not aommodate, its m payoff is 0, where is the disount fator, and the entrant s payoff is, where is a measure of the extent to whih the apital osts of entering the industry are sunk. If the weak inumbent aommodates, the weak inumbent and entrant both get the Cournot payoff,, in eah of the two periods, for a total payoff of (1 ). If the potential entrant does not enter in period 1, it hooses whether or not to enter in period 2. At the end of period 1, just before period 2, the entrant learns the inumbent s type (perhaps beause it has had time to observe the inumbent s reation to other entrants). If the potential entrant does not enter in either period, its payoff is 0, and the m m inumbent s payoff is (1 ), where is the monopoly profit. If the inumbent does m not aommodate in period 2, then its payoff is and the entrant s payoff is. If the weak inumbent aommodates in period 2, then its payoff is m and the m entrant s payoff is. Notie that if (1 ), the inumbent never aommodates, and hene the potential entrant never enters if it has to inur any positive m sunk entry ost. Heneforth, we assume that (1 ). Case 1. {0,1} and 0 Suppose 0. By bakwards indution, the inumbent aommodates in both periods, and hene the entrant enters in period 1, regardless of. Now suppose 1. In this ase, the entrant knows that the inumbent never aommodates, and therefore it never enters, whether is small or large. Thus, large sunk osts do not delay entry, or do not ause additional entry delay, in the absene of unertainty. In other words, sunk osts are not primary barriers to entry. Case 2. (0,1) and 0 By bakwards indution, the weak inumbent aommodates in both periods. Therefore, the potential entrant enters in period 2 if it has learned at the end of period 1 that the inumbent is weak, but does not enter if it has learned that the inumbent is aggressive. Now, the potential entrant s expeted payoff from not entering in period 1 is (1 ) (whih is a measure of the lost option value of entering), while its expeted payoff from entering in period 1 is (1 )(1 ). Thus, the potential entrant always enters in period 1. Thus, unertainty never deters entry, in the absene of sunk entry osts. In other words, unertainty is not a primary barrier to entry either. Case 3. (0,1) and 0 19

20 By bakward indution, we find, one again, that the weak inumbent aommodates in both periods, and therefore, the potential entrant enters in period 2 if it learns that the inumbent is weak, but does not enter if it learns that the inumbent is aggressive. The potential entrant s expeted payoff from not entering in period 1 is still (1 ), but now its expeted payoff from entering in period 1 is ( ) (1 )(1 ). Therefore, the potential entrant does not enter in period 1 if and only if (19) 1 Thus, large sunk osts (high ) and unertainty ( not too small) an ombine to delay entry until the realization of unertainty. For an important lass of demand funtions, effiient entry is in advane of the realization of unertainty. Hene, sunk osts and unertainty are anillary, antitrust barriers to entry that ombine, and reinfore eah other, to produe a primary, antitrust barrier to entry. III.C. ABSOLUTE COST ADVANTAGES Inumbents may have patents on superior prodution tehniques, learned through researh and development. And inumbents may have already established their operations in the most favorable loations, so that entrants may have to pay more for sare raw materials and other ruial inputs, and ship them a greater distane than inumbents. Patents and loational disadvantages diretly raise the average ost of entrants above those of inumbents at any ommon output level. Moreover, no entrant is apable of diretly overoming this ost. In the United states, patents are valid for twenty years, and firms that overtly breah other firms' patents are usually suessfully proseuted. These are osts that entrants have to bear but that inumbents have not had to bear. Therefore, they are diret eonomi (and hene antitrust) entry barriers. Do diret antitrust entry barriers, suh as absolute ost advantages, neessarily redue welfare? What about the soial benefits of entry barriers, suh as their tendeny to enourage innovation? To the extent that absolute ost advantages reate entry barriers and preserve monopoly profits, they supplement the inentives provided for development of new produts by patents, trademark, and opyright law. If imitation is easy, then firms will have less inentive to inur the fixed osts of new produt development sine ompetitors that imitate will prie to over only their prodution osts and the innovative firm will have trouble reovering its development osts. This obviously redues the inentives of firms to engage in the development of new produts and as a result soial welfare may be redued. Dempsetz (1982) observes that entry barriers into prodution redue entry barriers into innovation, and vie versa. Theorem: An inrease in diret antitrust entry barriers in markets that are otherwise effiient redues welfare. Capital market imperfetions seem to be another soure of absolute ost advantages for inumbents. Entrants may not be able to pay large apital osts even if entry is 20