Are judges leading economic theory? Sunk costs, the threat of entry and the competitive process.

Coate, Malcolm B.

I. Introduction

Since 1984, antitrust law has shown a growing respect for the threat of entry as a condition to immunize from legal challenge a merger that otherwise would be believed to significantly increase the probability of collusion. Yet the economics literature, strictly interpreted, would imply that a collusive agreement should not fear entry in the presence of positive sunk costs. This suggests that either the judiciary is leading economic theory or that the importance of entry conditions in merger analysis is likely to decline in the future as courts incorporate the latest economic learning into the case law.

In this paper, we attempt to bring theoretical form to a particular entry argument that has found favor in the courts. We suggest that sunk costs may not be a major impediment to entry when a group of customers can commit to an entry enhancing strategy. Buyers have strong incentives to adopt such strategies, because they benefit directly from the resulting lower prices. The simplest example involves a large buyer that is able to guarantee an entrant a market for its product. Long term contracts or even informal purchase commitments (backed by customer reputation) may also allow the entrant to obtain a guarantee of sufficient business to make the entry profitable. Once entry is thought likely to occur, the existing competitors may be unwilling to attempt a price increase of any magnitude. Thus, the threat of entry can maintain competitive prices even in the presence of sunk costs.

In section II, we motivate the analysis by discussing the current controversy over the role entry conditions play in an antitrust merger review. Then we discuss the necessary assumptions for a model of how the threat of entry can deter any price increase in the presence of sunk costs. We note that these conditions are different from those commonly used in the economic literature. Section III contains the details of our general sunk cost model. We illustrate how large buyers, uncertainty, the cost of collusion and market growth can interact to maintain competitive pricing. We also discuss how economies of scale can affect the threat of entry. The general applicability of the model is discussed in section IV with empirical evidence from recent federal court antitrust litigation.

II. Entry in Antitrust - And Economics

Starting with U.S. v. Waste Management, 743 F.2d 976 (2d Cir. 1984), courts have consistently found against the government in highly concentrated industries with few or no apparent barriers to entry.(1) In a recent merger case, the Department of Justice (DOJ) attempted to reverse this trend by advancing a restrictive legal approach to entry. The DOJ argued that the only way to defeat a presumption of an anticompetitive effect based on high concentration is "by a clear showing that entry into the market by competitors would be quick and effective." Judge (now Supreme Court Justice) Clarence Thomas, writing for the Court of Appeals for the District of Columbia Circuit, rejected the government's approach, concluding that it is unrealistic to expect such strong proof in the context of a merger case and even if a firm never enters a market, the threat of entry can stimulate competition.(2) Thus, one could conclude that some showing of entry barriers is now a necessary condition for a merger to violate the antitrust laws.

During the 1980s, the official government policy on entry, written into the DOJ Guidelines [25, 12], was somewhat unclear, stating both

If entry into a market is so easy that existing competitors could not succeed in raising price for a significant period of time, the Department is unlikely to challenge mergers in that market.

and

In assessing the ease of entry into a market, the Department will consider the likelihood and probable magnitude of entry in response to a "small but significant and nontransitory" increase in price.

The Guidelines then defined the magnitude of "small but significant" as a five percent price increase (although this figure was to be adjusted for special industry conditions) and interpreted the time frame for "nontransitory" as generally two years.

The first quote suggested that the DOJ would focus on how entry conditions affected the expected profitability of a nontransitory price increase. The threat of entry within a two year time period could render a price increase unprofitable, if the expected expansion of output lowered the long run price sufficiently to reduce the profits of the colluding firms below the level associated with competitive behavior. This point was implicitly recognized in a footnote to the Guidelines that posited that the prospect of entry may have a deterrent effect on the exercise of market power.

The second quote highlighted the more tangible effects of entry. By focusing on the likelihood and magnitude of entry, the Guidelines suggested that it is the magnitude of entry that would occur during a two year period that should be considered. As entry became more likely to prevent or eliminate an anticompetitive price increase within two years, the government indicated that it was less likely to challenge the merger.

The two analyses illustrated different approaches to the entry question. The first approach relied on information that suggests that the threat of entry would deter a price increase, while the second suggested sufficient entry should occur in two years to return the market to competitive equilibrium. The first method was compatible with the classical microeconomic theory of markets in which profits attract entry. In contrast, the second approach appeared to consider the idea that sunk costs may prevent entrants from investing in a market in response to supracompetitive prices, because entry could depress the price below the competitive level.

The government appears to have made an attempt to integrate the two approaches in the 1992 revision of the Guidelines. Although the direct focus of the 1992 Guidelines [26] is on the likelihood (in combination with the timeliness and sufficiency) of entry, the discussion is general enough such that the threat of entry could be addressed in the analysis.(3) In particular, the 1992 Guidelines attempt to determine if the profitability of entry is undermined by the scale necessary for efficient entry. In determining if the minimum viable scale of entry would depress price below the competitive level, the Guidelines require consideration of market growth, vertical integration or forward contracting (large buyers) and anticipated accommodation by the incumbents.(4) To the extent that the ease of forward contracting impacts on the threat of entry, the Guidelines appear to recognize a defense based on the threat of entry defeating an anticompetitive price increase.

One theoretical justification for the court's (and potentially the 1992 Guidelines') position can be found in the contestability literature. Baumol, Panzar, and Willig [3] show how the threat of entry into a perfectly contestable market can be sufficient to deter price from rising above competitive levels, regardless of the level of concentration in the market. As Schmalensee [22, 42] observes, however, the contestable market result may describe an "empty box". In practice, the zero (or extremely small) sunk cost requirement for contestability appears unlikely to be met. Without this condition, contestability theory seems to have little to tell us about the threat of entry for antitrust policy.

Consider, for example, a competitive market with a few competing firms facing potential entrants with small but significant sunk costs. Should the firms in that market decide to collude and raise price, they have no apparent reason to fear entry. While a prospective entrant may observe prices that would make entry profitable, those prices are, from the point of view of the entrant, merely a mirage. As soon as a new firm enters the market, the collusive agreement dissolves and prices will be driven below the pre-entry competitive level. Thus, the entrant will not capture profits from supra-competitive prices, instead the entrant will lose money on its sunk costs, even though it has a cost structure identical to its entrenched competitors. Entry in such circumstances would appear illogical. Knowing this, incumbent firms would appear able to raise their prices collusively without the worry of entry. [24, 886, 890; 12, 386-9].

Our purpose here is to present a model that shows circumstances when the threat of entry can deter such a price rise in the presence of sunk costs.(5) We note that a model of entry in this context should have several features. First, it should have positive sunk costs. Second, it is desirable that the model have a marginal cost pricing equilibrium with more than one firm. Previous models in the literature, such as Gelman and Salop [11] and Scheffman and Spiller [20] have assumed an industry with constant (horizontal) marginal cost curves and positive sunk costs. In such an industry, Bertrand competition generates marginal cost pricing and firms are unable to recoup their sunk costs. Thus, the first firm in the market may expect to retain a monopoly position, because entry is unprofitable. If entry should occur, Bertrand competition should continue until only one firm remains in the market. Either case may be of limited interest to antitrust policy, which implicitly adopts the preservation of marginal cost pricing as a policy goal.

Third, entry should threaten to lower the long run returns to the colluding firms. If entry only reduces incumbents' returns back to their previous level of zero economic profits, they will have little to lose by colluding. They will choose to collude, taking the chance of gaining positive profits and the "risk" of zero profits, rather than face the certitude of gaining zero profits. Thus, entry must pose a risk to incumbents' quasi-rents to have a possibility to deter a price increase.

Fourth, it is important to model buyers as strategic players. As Sexton and Sexton [23] and Scheffman and Spiller [20] point out, buyers can take actions to protect themselves from anticompetitive prices above a certain limit price.(6) In particular, as Demsetz [10] and Yu [29] suggest, they can use long-term contracts (of various forms, including complete vertical integration, the ultimate long-term contract) to either enter the market themselves or induce another firm to enter. What we are presenting here can thus be thought of as a generalization of Demsetz's original model.

Fifth, a model of entry should take account of uncertainty in the market, as has become common in the industrial organization literature over the past decade. The collusive firms are unlikely to know how high they can price without inducing entry. One thing, however, is known for certain: that the potential entrant has not yet entered. This will be shown to put a floor on the costs of entry.

Finally, a model of entry should focus on the threat, rather than the probability, of entry. This is contrary to the general focus of the economics literature (starting with Bain [1]) which evaluates what deters entry, given that a price increase has already taken place. Our model considers the decision one step earlier in the economic game, the decision to raise price above competitive levels. As the discussion in section III will show, this approach can change the interpretation of scale economies as a barrier to entry.

Before we proceed further, we note that it is theoretically possible for a monopolist to deter entry by offering a large buyer lower prices than its other customers [14].(7) There are, however, several problems with this argument. First, antitrust merger cases often involve hypothetical analysis of tacitly collusive agreement, not dominant firms. In this context, firms could have serious difficulties in deciding how to supply the potential entrant with low-cost product. Second, there may be a number of potential entrant buyers in the market such that offering all of them discount prices would dissipate the available supracompetitive profit. Finally, were such price discrimination to occur, it could constitute a violation of the Robinson-Patman act.

Rules of the Game

We present a simple one period entry game where the threat of entry can deter, at least under some conditions, any collusive price increase. There are N identically sized producers in this industry, each producing with the same commonly available technology with a known marginal cost structure. Each of the incumbent producers has already paid the sunk costs necessary to enter the industry. To create the required quasi-rents, we assume that the available technology generates for each firm an upward sloping supply curve. This allows for a competitive equilibrium in the presence of sunk costs, and places some of these firms' quasi-rents at risk in the event of entry.

For ease of presentation, following on Scheffman and Spiller [20], buyers have a perfectly inelastic demand for the industry's product. Buyers, however, are allowed to enter the upstream market themselves. Entry can take the form of one firm vertically integrating, a group of firms entering into a joint venture, or buyers using long term contracts to induce entry. The lowest cost potential entrant faces some sunk costs [S.sup.E], the distribution of which we will discuss below. Once the entrant has incurred the necessary sunk costs, it produces with the same marginal cost schedule as incumbents.

In the first stage of this game, producers decide what price they will charge. We assume that they are able to effectively agree on some price to be charged, absent entry, at some cost C. We initially assume C = 0.

In the second stage, buyers have a choice. They can either accept the industry price or induce entry. Entry is induced by payments from a coalition of buyers. This coalition may take the form of the largest buyer in the industry. Or the coalition may consist of two or more buyers creating a joint venture that produces the relevant product. Or it can take the form of any number of firms entering into contracts with the entrant that compensate the entrant for its losses in a post-entry competitive market. For simplicity, we assume that the transaction costs of forming the relevant coalition are zero.(8) We also assume that entry is induced when there is an arrangement available that is pareto optimal for all members of the buyers' coalition as well as for the potential entrant. The buyer coalition has market share [lambda]. Once entry occurs, the collusive agreement is assumed to dissolve and firms price at marginal cost.(9)

The Base Case Solution

In the mathematical model, demand is perfectly inelastic at Q.(10) P equals price. Incumbent firms price at marginal cost and have aggregate supply curve

[Q.sup.I](P) = (P - A)/B A,B > 0. (1)

Let N equal the number of equally-sized minimum efficient scale firms in the industry. (Thus N = 1/MES, where MES is interpreted as a measure of minimum efficient scale.) A potential new entrant is willing, after paying its sunk costs, to supply the market with

[q.sup.E](P) = (1/N)(P - A)/B. (2)

Let K = N/(N + 1), 0 < K < 1. The total supply curve after entry is

[Q.sup.E](P) = (P - A)/KB. (3)

A graphical representation of the model is presented in Figure 1. The two industry supply curves ([S.sup.I] and [S.sup.E]) both originate at point A and determine the market price by their intersection with the (perfectly inelastic) demand curve. Given the additional capacity, the post-entry price [P.sup.E] is below the pre-entry price [P.sup.I]. Industry profits, both before and after entry, are defined by the area between the supply curve and the relevant market price.

Returning to the model, price and industry profits after sunk costs (quasi-rents) before and after entry are

 [P.sup.I] = BQ + A (4a)
 [pi.sup.I] = [Q.sup.2]B/2 (4b)
 [P.sup.E] = KBQ + A (5a)

[pi.sup.E] = [Q.sup.2]BK/2. (5b)

We assume that industry profits at least cover sunk costs in the pre-entry equilibrium. The existence of sunk costs insures lumpy entry, which in tum creates the possibility that the existing firms will earn supranormal rents as long as the available returns do not trigger entry. To examine the entry question, we focus on profits after entry for the new entrant and all previous incumbents.

[pi.sup.E] = [Q.sup.2]KB/2(N + 1) (6)

[pi.sup.E.sub.i] = [Q.sup.2][K.sup.2]B/2. (7)

We know that the potential entrant has not yet entered. We also know from (6) what profits (quasirents) it would make, and what the buyer coalition with market share [lambda] would be willing to pay it to enter, [lambda]([P.sup.I] - [P.sup.E])Q.(11) For this situation to be an equilibrium the amount of sunk costs facing the new entrant cannot be less than the sum of the post-entry quasi-rents and the available buyer payments. Thus, the minimum amount of sunk costs SL the entrant can face is

[Mathematical Expression Omitted] (8)

Since (1 - K) = 1/(1 + N) = K/N

[Mathematical Expression Omitted] (8a)

[S.sub.m], the maximum possible amount of sunk costs that the entrant may be facing, remains unknown. Without loss of generality, let [S.sub.m] = (1 + R)[S.sub.L], R > 0, and [S.sub.m] - [S.sub.L] = [RS.sub.L], where R is unknown.

Assume that the cartel raises price above the competitive level by [P.sub.m]. It will be worth an additional [p.sun.m][lambda]Q to the buyer coalition to induce entry. Similar to Crawford and Sobel [9, 1440] and McAfee and McMillan [17, 109], we assume that the actual sunk costs of the most favorable entrant [S.sub.E] are distributed uniformly in the range [S.sub.L, S.sub.M ].(12) This implies the probability of entry in response to a price rise [p.sub.m] is equal to the probability that [S.sub.L] + [p.sub.m lambda]Q > [S.sub.E], or

[Mathematical Expression Omitted] (9)

If the cartel raises price without inducing entry it gains profits

[Mathematical Expression Omitted] (10)

The cartel thus maximizes profits over [p.sub.m], given the probability of entry

[Mathematical Expression Omitted] (11)

Taking derivatives and setting equal to zero yields

[Mathematical Expression Omitted] (12)

Dividing (12) by Q and rearranging generates

[Mathematical Expression Omitted] (13)

Equation (13) implies that the cartel will either raise price, with [p.sub.m*] > 0 and face the threat of entry, or it will not raise price at all. If it chooses not to raise price, it does so because the expected profits from [p.sub.m*] > 0 are less than expected losses from a decline in price due to new entry. We now assume no price rise ([p..sub.m*][less than or equal to] 0, which imphesp.[p.sub.m*] = 0 because in this circumstance a cartel will have no reason to set price below the equilibrium competitive level). Since [RS.sub.L]/2,[lambda]Q > 0, (13) implies

[Mathematical Expression Omitted] (14)

[Mathematical Expression Omitted] (15)

[Mathematical Expression Omitted] (16)

Let N = 5(13) (MES = 0.2 and K = 5/6), [lambda] = 0.2 and let [R.sup.c] denote the critical level of R, above which supracompetitive pricing will occur. Equation (16) can then be evaluated as R [less than or equal to] 11/37 = 0.2973. Thus, if the sunk costs to enter will not be greater than ([R.sup.c] + 1 =) 1.2973 times the minimum sunk cost to enter, no price rise will occur. On the other hand, if R > 1.2973, a cartel will find it ex ante profitable to risk the threat of new entry and will engage in a form of limit pricing.

One issue that often arises in antitrust is whether economies of scale constitute an entry barrier [21, 424]. In the model here N represents the inverse of economies of scale. If economies of scale are a barrier, [dR.sup.c]/dN < 0. Differentiating (16) yields

[Mathematical Expression Omitted] (17)

A priori, (17) indicates that it is ambiguous whether economies of scale reduce the threat of entry. Thus, contrary to the conclusion of Bain [1] and others, it is uncertain whether economies of scale represent a barrier to entry from a merger enforcement perspective. Evaluating (17) given N = 5 implies that increasing economies of scale act to reduce the threat of entry if [lambda] < 0. 5.

The intuition behind this result is straightforward. Large economies of scale increase the costs of entering, because such entry will have a larger impact on price. This larger impact on price, however, constitutes a greater threat to incumbents.(14) Further, it increases the available level of payments from a buyer coalition. Thus, increased economies of scale decrease the probability of entry, but increase the costs to incumbents of that entry should it occur.

Growing Markets, Entry and "Predation"

We can adjust our model slightly to accommodate the dynamics of growing markets. Assume that before the game starts, but after incumbents have sunk their costs, the market grows by some factor g > 0.(15) Demand for the industry's product now equals (1 + g)Q. In this circumstance

 [P.sup.G.sup.I] = (1 + g)BQ + A (18a)
 [[pi].sup.G.sup.I] = (1 + g) [Q.sup.2.sup.2]B/2 (18b)
 [P.sup.G.sup.E] = (1 + g)KBQ + A (19a)
 [[pi].sup.G.sup.E] = (1 + g)[Q.sup.2.sup.2] BK/2 (19b)
 [[pi].sup.G.sup.E.sub.e] = (1 + g)[Q.sup.2.sup.2]KB/2(N + 1) (20)
 [[pi].sup.G.sup.E.sub.i] = (1 + g)[Q.sup.2.sup.2][K.sup.2]B/2. (21)

An entering firm will capture ([(1 + g).sup.2)] - 1)[Q.sup.2]KB/2(N + 1) additional profits in inframarginal rents. Further, the buyer coalition is now willing to pay an additional A [lambda] ([(1 + g).sup.2)] - 1) (1 - K)[Q.sup.2]B to induce entry. Thus, entry becomes more profitable to the entrant and the buyer coalition in this scenario by ([(1 + g).sup.2] - 1)S.sub.L and even without a price increase, entry will occur with probability ([(1 + g).sup.2] - 1)/R. (We assume that R > [(1 + g).sup.2] - 1.)

If entry does not occur absent conclusion, the actual sunk costs of the most favorable entrant [S.sup.E] are distributed uniformly in the range [[(1 + g).sup.2][S.sub.L], (R + I)[S.sub.L]]. Thus, the probability of entry in response to a price rise is [[sigma].sup.G] ([p.sub.m]) = ([p.sub.m.][lambda](1 + g)Q/[(R + 1 - [(1 + g).sup.2])[S.sub.L]] for [p.sub.m] in the relevant ranges. The cartel thus maximizes profits over [p.sub.m], given the probability of entry

Max [pi] = [[sigma].sup.G][(p.sub.m).sup.G][[pi].sub.i.sup.E] + ([1 - .sup.G][sigma]

[(p.sub.m).sup.G)][[pi].sup.C] ([p.sub.m]) (22) where [[pi].sup.G.sup.C] ([p.sub.m]) = .sup.G [pi].sup.I + [p.sub.m] + (1 + g)Q. Taking derivatives and setting equal to zero yields

[lambda][(1 + g).sup.3][Q.sup.3][K.sup.2]B/2(R - [(1 + g).sup.2] + 1)[S.sub.L] + [(1 + g).sup.2]

+ 1)[S.sub.L] + (1 + g)Q

- 2([lambda]).p.sub.m (1 +g).G.sup.2/(R - (1 + g).sup.2) + 1).S.sub.L

- [lambda] [(1 + g).sup.3][Q.sup.3]B/2)R - [(1 + g).sup.2] + 1)[S.sub.L] = 0

(23)

Dividing (23) by (1 + g)Q and rearranging terms yields

[p.sup.*.sub.m] = [(R - [(1 + g).sup.2] + 1)[S.sub.L]/2[lambda]Q]

[1 - [lambda](1 - [K.sup.2])[(1 + g).sup.2][Q.sup.2]B/2(R - [(1 + g).sup.2] + [S.sub.L]. (24)

No price rise implies

2(R - [(1 + g).sup.2] + 1)[S.sub.L] [less than or equal to][lambda]([1 - K.sup.2])

[(1 + g).sup.2] [Q.sup.2]B

2(R - [(1 + g).sup.2] + 1)[Q.sup.2] BKH [less than or equal to] [lambda] (1 - [K.sup.2])

[(1 + g).sup.2][Q.sup.2]B.

(R - [(1 + g).sup.2] + 1) [less than or equal to][(1 + g).sup.2][[lambda](2N + 1)/[N

+ 2[lambda]N + 2[lambda]]

(R [less than or equal to] [(1 + g).sup.2][[lambda](2N + 1)/[N + 2[lambda]N + 2[lambda]]

+ [(1 + g).sup.2] - 1. (25)

Letting N = 5, [lambda] = 0.2, and g = 0.05 implies R [less than or equal to] 0.4304. Thus, in this example an increase in the market demand by 5 percent yields a 44.7 percent increase in the range of entry costs that deters an anticompetitive price rise.(16)

We note that in the case of a growing market, the equilibrium number of firms (absent strategic action by incumbents) may not yet be achieved. In this case, incumbent firms may actually seek to collude to lower price, engaging in a form of predatory pricing. Such collusion, while dropping price below marginal cost (but above average cost) may be profitable, because it serve to protect incumbent firms' quasi-rents by detering entry.

Costs of Collusion and Entry

The model can also be generalized to allow for positive costs of collusion. Along these lines, we assume that cartel coordination requires firms to incur both fixed and price-related costs.

C([p.sub.m]) = Z + [p.sup.d.sub.m]Q Z > 0, 1 > d > 0. (26)

The initial expenditure (Z) proxies the costs necessary to organize the cartel and the variable expenditure (d) represents the costs of policing the agreement which are a function of the supra-competitive profits at risk.(17)

Given these costs must be incurred to increase price regardless of whether entry occurs, the cartel's profit maximizing problem over [p.sub.m] becomes

Max [pi] = [sigma]([p.sub.m])[[pi].sub.i.sup.E] + (1 - [sigma]-([p.sub.m])).[[pi].sup.C]

([p.sub.m]) - (Z + [p.sup.d.sub.m.Q). (27)

Again taking derivatives and equating to zero

[lambda][Q.sup.3][K.sup.2]B/[2RS.sub.L] + Q - 2[lambda][p.sub.m.][Q.sup.2]/[RS.sub.L] - [lambda]

[Q.sup.3]B/[2RS.sub.L] - dQ = 0. (28)

Rearranging (28) generates a new equation for price

[p.sub.m.sup.*] = [RS.sub.L]/2[lambda]Q][(1 - d) - [lambda](1 - [K.sup.2])

[Q.sup.2]B/[2RS.sub.L]]. (29)

The only difference between (29) and (13) is the use of the term (1 - d) instead of 1. Thus, the remaining analysis is identical except for the additional of (1 - d) on the left-hand side of the analysis. This implies that (16) can be written in the more general form of

R [less than or equal to] [[lambda](2N + 1)/[(N + 2[lambda](N +1))(1-d)].

(30)

Integrating both the costs of collusion and the growth model yields

R [less than or equal to][(1 + g).sup.2] [[lambda](2N + 1)/)N + 2[lambda])(N +1)) (1 - d)]

+ [(1 + g).sup.2] -1. (31)

Retaining the assumptions that N = 5, [lambda] = 0. 2, and defining d = 0.2 generates a value of R of 0.372. If growth is considered (i.e. as in equation (31) with a value of 0.05) the value is 0.512. This value is 72 percent higher than the initial value for a stagnant market with no costs of collusion.

Even if the threat of entry by itself will not deter a price increase, it is possible that the optimal price increase will be so small that the cartel will be unable to cover the fixed costs of collusion. Thus, collusive activity will not occur, unless the firms can impose a sufficient price increase to cover the fixed costs of cartelization. Thus, as the fixed costs of collusion increase from zero, small price changes optimal under (28) generate a loss for the cartel, so no price change will occur.

IV. The Judicial Response to Large Buyers

Large Buyer Arguments in the Case Law

Empirical evidence on the threat of buyer-induced entry is inherently difficult to obtain, because it is the threat of entry, not entry itself, that defeats an anticompetitive price increase. It is possible, however, to gather evidence on the ability of buyer coalitions to induce entry by reviewing litigated cases where actual or potential buyer coalitions played a role. Perhaps the best example of a buyer coalition actually defeating anticompetitive pricing is described in Sewell Plastics Inc. v. Coca Cola Co. 720 F. Supp. 1186 (W.D.N.C. 1988), aff'd 912 F.2d 463 (1990), cert. denied III S. Ct. 1019 (1991). Sewell Plastics, one of the innovators of plastic soft drink bottles, attempted to maintain high prices into the 1980's. In 1981, a group of Coca Cola bottlers approached Sewell and tried to negotiate lower prices by threatening entry. When Sewell failed to offer a sufficient discount, the bottlers created a joint venture to enter the market. (The District Court decision at 1208 indicates that such cooperatives are common in this industry.) The entrant, SouthEastern Container, grew to obtain a 33.5 percent share of the plastic bottle market and prices fell from $220 per thousand in 1982 to $146 in 1986. Thus, when induced by Sewell to actually enter, the bottlers succeeded at pushing price down dramatically.(18)

Additional support for the concept of buyer-induced entry can be gleaned from examining the merger challenges litigated by the government between 1982 and 1991 listed in Table I.(19) A review of litigated cases presents examples of when courts believed that the threat of entry was sufficient to maintain competition or when parties induced actual entry to create competition at vertically-related levels of production.(20)

One of the first observations from reviewing the list of recent merger cases in Table I is that it is relatively rare to have a merger litigated in federal court involving direct sales to atomistic consumers. Even consumer goods can be sold through retailers that are large enough to create new entry. Other consumer goods are sold through mixed systems with some consumers purchasing directly and others through large buyer groups. Finally, when atomistic consumers face monopolistic sellers in retailing, buyer coalition strategies can easily be inverted and applied to input suppliers. Thus, for almost any class of mergers, buyer strategy arguments, or their equivalent, can be entertained.

Table I. Products in Merger Challenges: Government Cases in
Federal Court, 1982-1991 (Number of Cases in
Parentheses)

Banking Services (2) Gasoline Distribution
Carbon Black for Tires Rigid Wall Containers
Pre-recorded Music Industrial Dry Corn
Commercial Trash Collection Automatic Railroad Tampers
Carburetor Kits Supermarkets
Corrugating Medium Night Vision Tubes
Sprayers and Dispensers First Run Movie Releases
Milling of Paddy Rice Fluid Milk
Plastic Feed Stocks (3) Hardrock Hydraulic Mining Equipment
Carbonated Soft Drinks Printing Services
Aircraft Transparencies Schmidt-Cassegrain Telescopes
Hospital Services (5) Movie Laboratory Service Agreements
Race Track Equipment Gas Cabinets

Source: Various Federal Court Merger Decisions

The simplest buyer coalition argument involves large buyers purchasing from a concentrated group of sellers. Examples from litigated cases include carbon black (a key input into tires), aircraft transparencies (a vital component for aircraft), and 25 mm second generation night vision tubes (the crucial input for a class of night vision devices used for military purposes). Although the potential for buyer-induced entry in these examples appears clear, the relevant decisions imply that a buyer coalition would have had difficulty inducing entry, because entry was technically a long and difficult process. In other cases, entry appeared to be much easier. For example, in U.S. v. County Lake et al. at 117, District Court Judge Renner found that the large milk distributors, who accounted for over 90 percent of the customers in the relevant market, could and would seek suppliers outside the local area or vertically integrate in response to anticompetitive pricing by local milk processors. This was considered sufficient to maintain competition. Similarly, in U.S. v. Calmar at 1304, District Court Judge Debevoise found that buyers of pump dispensers and sprayers would react to an anticompetitive price increases by either vertically integrating or entering into joint ventures to make the products.

Another example of the potential for buyer strategies can be found in the 1990 Baker Hughes decision. District Court Judge Gesell noted that the major customers for hardrock hydraulic mining equipment would insist on receiving competitive bids and were likely to have contacts with mining equipment manufacturers in Canada. Thus, Judge Gesell found that buyer strategies would facilitate successful entry into the U.S. market were the merger to induce collusion. A key point in the decision appears to relate to the sophistication of buyers rather than their absolute size. Thus, even when the buyers do not have large market shares, it has been concluded that they may be able to contribute to maintaining competition. Similarly in Echlin Manufacturing Co., 105 F.T.C. 410 (1985), the FTC observed that resellers which purchased carburetor kits from assemblers had some power to maintain competitive prices. Moreover, the Commission noted that resellers could either have another firm package the carburetor kits for them or do it themselves. Thus, buyers were considered potential entrants whose existence was believed to keep the market competitive. While buyer-induced entry has had significant success as a defense tactic in merger litigations, it by no means represents a panacea for defendants. For instance, buyer strategy arguments have been given little weight in hospital merger decisions where third party payers (Blue Cross and the government) appear to be very large buyers. In Hospital Corp. of Am. v. F.T.C., 106 F.T.C. 361, 509 (1985), 807 F.2d 1381 (7th Cir. 1986), cert. denied 107 S.Ct. 1975 (1987), the Commission rejected the buyer strategy argument, observing Blue Cross could not switch its business to hospitals outside the geographic market. In reviewing the case, Judge Posner developed this idea further by noting third party payers are not completely analogous to large buyers. Insurance companies are obligated to pay the contracted portion of the medical charges for their customers as opposed to large buyers that could strategically reduce purchases of product. It is not clear how the third party payer could threaten to move large amounts of business away from the oligopolists, although conceivably they could help to develop HMOs. Similarly, in U.S. v. Rockford Memorial Corp., 898 F.2d 1278 1285 (7th Cir. 1990), Judge Posner observed that the overall effect on competition of buyer threats was unclear.

We note that both hospital cases involved relatively large barriers to entry (such as certificate of need regulation), with no potential entrant apparently being "close" to entering. Thus, the third party payers were unlikely to defeat the price increase by motivating new entrants to come into the market. In U.S. v. Carillon Health Systems, 707 F. Supp 840, 849 (W.D. Vir. 1989), however, Judge Turk held that the ability of other hospitals to expand was sufficient to outweigh the concerns caused by the increased concentration level. Although the concept of buyer induced-entry was not explicitly mentioned in this decision, it could have easily been integrated into the analysis. Overall, buyer strategy arguments may be applied to hospital mergers, but only in limited fact situations.

Finally, the idea of buyer-induced entry can easily be inverted into supplier strategy arguments. If a retailer attempts to monopolize a geographic area, both suppliers and consumers may suffer injury. The supplier can respond by inducing entry to defeat the anticompetitive overcharge. An example of this behavior comes from the movie theater industry. In U.S. v. Syufy Enterprises, 712 F. Supp. (N.D. Cal. 1989), aff'd 903 F.2d 659 (9th Cir. 1990), the court decisions report that Orion Releasing Group shifted its business to a small second run theater after a contract dispute with defendant Syufy. This action meant that a new competitor entered the first run market in Las Vegas where businessman Raymond Syufy had acquired a (short-run) market share of over 90 percent. District Court Judge Orrick and Judge Kozinski for the Appeals Court chronicled the success of the entrant and both concluded that Syufy's various movie theater mergers in Las Vegas had no anticompetitive effect.

Empirical Estimate of the Effect of Large Buyers on the Judiciary

To test our hypothesis that the presence of large buyers affects judicial decisions on entry, we use the litigated mergers cases where either the Department of Justice of the Federal Trade Commission was a plaintiff after the publication of the 1982 Merger Guidelines. After excluding three cases where barriers were not relevant to the decision, there remains a sample of 30 observations, 20 with court opinions reporting some form of entry barriers and 10 without.(21) Our dependent variable (Barriers) was equal to one if the court found barriers and zero if no barriers were identified. We measured the first independent variable, buyer power (BUYER), with a binary variable taking on the value of one if the court decision recognized some from of buyer power and zero otherwise. A second independent variable (ECON) represents the economic sophistication of the court as illustrated by the number of structural variables identified as affecting competition. If the court wrote a detailed opinion that explained how various structural factors affected competition, it may be more likely for the court to search for sophisticated economic arguments to justify barriers to entry.(22) To define this variable we summed the number of factors mentioned in the decision as either compatible or incompatible with noncompetitive behavior. Finally, there may be differences between the two antitrust agencies in the cases chosen to be litigated or in each agency's litigation incentives or abilities. To capture this, a dummy variable (DOJ) was included for the DOJ cases.

The model was estimated with a probit methodology to account for the binary dependent variable and the results are presented below (with t-statistics in parentheses).

Barriers = 0.589 - 4.09 BUYER + 0.909 ECON - 1.64 DOJ (32)

(0.88) (-2.50) (2.38) (-1.93)

Likelihood Ratio test: 22.53; Pseudo R-square: 0.590.

The coefficients of the independent variables are significantly different from zero and the overall model clearly passes the Chi-square at the one percent level. Using a fifty percent predition criteria, the model predicts the entry finding correctly for 95 percent of the twenty high barrier cases and 70 percent of the ten low barriers cases. Overall prediction success is achieved in 26 of the 30 cases (86.7 percent).(23)

The model confirms the basic hypothesis concerning the judiciary's use of the concept of buyer power to support a finding of low barriers. In addition, both the dummy variable for the litigating antitrust agency and the number of structural conditions found to affect competition also affected the likelihood of a barrier finding. Assuming the sophistication variable (ECON) is fixed at its mean of 2.37, a finding of buyer power lowers the probability of finding barriers from almost 100 percent to 9 percent in FTC cases and from 86 percent to less than 1 percent in DOJ cases.(24)

V. Conclusion

This paper attempts to bridge the gap between judicial decisions and economic theory relating to entry. Our model shows how buyer strategies can be used, at least in some circumstances, to overcome the presence of sunk costs such that the threat of entry is able to deter price increases. Such threats are likely to be most effective when sophisticated buyers (sellers) make up a large portion of the relevant output (input) market and when the lowest cost potential entrant is relatively "close" to entering the market absent an anticompetitive price increase. One, though not the only, way of thinking of this concept is asking how near to contestability a particular market is.

Our model also has other implications. We suggest analyzing the threat of entry at the time incumbents consider raising prices, rather than the probability of entry once a price increase occurs. We show how such a change could affect the definition of economies of scale as an entry barrier. Our model also implies that antitrust law should be more lenient towards joint ventures in vertically related markets to allow a wider array of buyer strategies. Recent court decisions clearly show that the judiciary recognizes many of these concepts. This paper is an attempt to formally present this phenomena to economists, as well as to allow jurists to give more structure to their decisions.

References

[1.] Bain, Joe. Barriers to New Competition. Cambridge: Harvard University Press, 1956. [2.] Baker, Jonathan B., "Identifying Cartel Policing Under Uncertainty: The U.S. Steel Industry, 1933-1939." Journal of Law and Economics, October 1989, S47-S76. [3.] Baumol, William J., John C. Jr. Panzar, and Robert D. Willig. Contestable Markets and the Theory of Industry Structure. New York: Harcourt Brace Jovanovich, 1982. [4.] Calkins, Steven, "Developments in Merger Litigation: The Government Doesn't Always Win." Antitrust Law Journal, Fall 1988, 855-900. [5.] Caves, Richard, and Michael Porter, "Market Structure, Oligopoly, and Stability of Market Shares." Journal of Industrial Economics, June 1978, 289-313. [6.] Coate, Malcolm B. and James Langenfeld. "Entry Under the Merger Guidelines, 1982-1992." Mimeo, Federal Trade Commission, 1992. [7.] ________, "Economics, the Guidelines and the Evolution of Merger Policy." Antitrust Bulletin, Winter 1992. [8.] ________ and Andrew N. Kleit, "Antitrust Policy for Declining Industries." Journal of Institutional and Theoretical Economics, September 1991, 477-98. [9.] Crawford, Vincent P. and Joel Sobel, "Strategic Information Transmission." Econometrica, November 1982, 1431-51. [10.] Demsetz, Harold, "Why Regulate Utilities?" Journal of Law and Economics, April 1968, 55-65. [11.] Gelman, Judith R. and Steven C. Salop, "Judo Economics: Capacity Limitation and Coupon Competition." Rand Journal of Economics, Autumn 1983, 315-23. [12.] Harrington, Joseph E., Jr., "Collusion and Predation Under (Almost) Free Entry." International Journal of Industrial Organization, September 1989, 381-401. [13.] Jacobson, Jonathan M. and Gary J. Dorman, "Joint Purchasing, Monopsony, and Antitrust." Antitrust Bulletin, Spring 1991, 1-80. [14.] Katz, Michael, "The Welfare Effects of Third Degree Price Discrimination." American Economic Review, March 1987, 154-67. [15.] Kovacic, William E., "Reagan's Judicial Appointments and Antitrust in the 1990's." Fordham Law Review, October 1991, 49-124. [16.] Lanning, Steven G., "Costs of Maintaining a Cartel." Journal of Industrial Economics, December 1987, 157-74. [17.] McAfee, R. Preston and John McMillan, "Auctions and Bidding." Journal of Economic Literature, June 1987, 699-738. [18.] Salop, Steven C., "Strategic Entry Deterrence." American Economic Review, May 1979, 335-38. [19.] ________ , "Measuring Ease of Entry." Antitrust Bulletin, Summer 1986, 551-70. [20.] Scheffman, David T. and Pablo T. Spiller, "Buyers' Strategies, Entry Barriers and Competition." Economic Inquiry, July 1992, 418-36. [21.] Scherer, F. M. and David Ross. Industrial Market Structure and Economics Performance, 3rd edition. Boston: Houghlin Mifflin Company, 1989. [22]. Schmalensee, Richard, "Ease of Entry: Has the Concept Been Applied Too Readily?" Antitrust Law Journal, Spring 1987, 41-51. [23.] Sexton, Richard J. and Terri A. Sexton, "Cooperatives as Entrants." Rand Journal of Economics, Winter 1987, 581-96. [24.] Stiglitz, Joseph E. "Do Entry Conditions Vary Across Markets?", in Brookings Papers on Economic Activity (Microeconomics), edited by Martin N. Bailey and Clifford Winston. Washington: The Brookings Institution, 1987, 883-947. [25.] U.S. Department of Justice. "Merger Guidelines." Antitrust and Trade Regulation Report, No. 1169, Special Supplement, June, 1984. [26.] ________ . "Department of Justice and Federal Trade Commission Merger Guidelines." Antitrust and Trade Regulation Report, No. 1559, April, 1992. [27.] Uri, Noel and Malcolm B. Coate, "The Department of Justice Merger Guidelines: The Search for Empirical Support." International Review of Law and Economics, June 1987, 113-20. [28.] Willig, Robert. "Merger Analysis, Industrial Organization Theory and the Merger Guidelines," in Brookings Papers on Economic Activity (Microeconomics), edited by Martin N. Bailey and Clifford Winston. Washington: The Brookings Institution, 1991, pp. 281-312. [29.] Yu, Ben T., "Potential Competition and Contracting for Innovation." Journal of Law and Economics, October 1981, 215-38.

(1.) Other examples include U.S. v. Calmar Inc., 612 F. Supp. 1298 (D.N.J. 1985), F.T.C. v. Promodes, 1989-2 Trade Cas. (CCH) [paragraph]68,688 (N.D. Ga. April 14, 1989) and U.S. v. Country Lake Foods, Inc., 1990-2 Trade Cas. (CCH) [paragraph]69,113 (D. Minn. June 1, 1990). (2.) U.S. v. Baker Hughes, 731 F. Supp. 3 (D.D.C. 1990), aff'd. 908 F.2d 981 (D.C. Cir., 1990). In his decision, Judge Thomas pointed out that "[s]ection 7 [of the Clayton Act] involves probabilities, not certainties or possibilities" (emphasis original). (3.) Coate and Langenfeld [6] provide a more detailed discussion of entry under the 1992 Guidelines. (4.) Salop [19] and Willig [28] provide a more detailed discussion of the minimum viable scale concept. (5.) Although we explicitly model the ability of entry to deter a post-merger price increase, the same analysis would apply to situations of price fixing dealt with under a legal rule of reason. Explicit price fixing, however, is per se illegal in the United States and therefore the threat of entry is not a relevant legal issue in such contexts. (6.) Sexton and Sexton [23] model a cooperative under certainty and conclude limit pricing is a possible outcome, while Scheffman and Spiller [20] model a market with a large buyer where the supplier must invest in customer-specific sunk costs. They also find limit pricing can occur. (7.) As section IV discusses, there have been situations in which noncompetitive pricing has induced entry. Thus, the argument clearly does not apply to all cases. (8.) This will be most applicable when the buyer coalition consists solely of the largest relevant buyer. Should there be costs to forming the coalition, those costs should be subtracted from the right hand side of equation (8) below. This simplifying assumption does not affect the conclusions of the model. We also assume away the free rider problem for both a supplier cartel and a buyer coalition. While a more general model would find buyers less likely to induce entry, it would also find oligopolists less likely to be able to raise prices. We note that buyers could use strategies other than inducing entry to defeat a price increase. Moreover, a buyers' coalition could also conceptually serve to facilitate of monopsony power [13]. (9.) Given that collusive schemes are inherently unstable, we assume that the disruption resulting from entry breaks up the collusion. The additional output of the new entrant will then drive prices below the pre-cartel level. Theoretically, entry increases the returns to cheating on collusion, because it is uncertain to members of the cartel whether they are losing sales to the new entrant or to an incumbent who is cutting price [5; 2]. (10.) This is done for the sake of simplicity. The inelastic demand curve assumption overestimates the return to supracompetitive pricing, the harm to buyers from such pricing, the loss in profits to incumbents due to entry, and the cost to buyers of inducing entry. (11.) Note that the larger the market share of the buyer group, the closer the market equilibrium moves to the contestability result even in the presence of sunk costs. (12.) [S.sub.L] can be interpreted as the minimum possible sunk costs in an industry and [S.sub.m] as the maximum possible sunk costs. If the actual sunk costs for the most likely potential entrant were less than [S.sub.L] then entry will occur and generate a new equilibrium. Then one could continue the analysis with one more incumbent and the next lowest cost potential entrant. (13.) We choose N = 5 to proxy the competitive conditions associated with the marginal antitrust case. The DOJ Guidelines view markets as "highly concentrated" if the relevant Herfindahl-Hirschman Index (HHI) is greater than 1800, which can be interpreted as 10000/1800 = 5.56 equal sized firms. This implies that a market with five equally sized firms (HHI=2000) creates a highly concentrated market structure, which if not offset by other market characteristics (such as the threat of entry or market characteristics that discourage collusion) may generate an anticompetitive effect. As Uri and Coate [27] point out, however, there is no empirical reason to believe that any particular cutoff level of market concentration such as an HHI of 1800 is related to the likelihood of anticompetitive behavior. (14.) The derivative of the right hand term of equation (9) with respect to N is positive, implying that increased economies of scale decrease the probability of entry. The derivative of the loss to the incumbent producers as a result of entry (which can be calculated by subtracting equation (7) from equation (4a)) with respect to N is negative, implying that should entry occur, the increased economies of scale will lead to greater losses to incumbent producers. (15.) Note that growth in this context can mean either a positive shock to demand or a decrease in available supply due to depreciation of existing capacity. (16.) This model does not imply that "entry" cannot happen in a declining industry. In a declining industry, collusion may be defeated if fringe firms see a higher price and choose not to exit. In effect, "not exiting" becomes entry. See Coate and Kleit [8, 489]. The threat of "not exiting" played an important role in a recent Canadian merger decision, DIR v. Hillsdowne Holdings Ltd. (1992) 41 C. P. R. [30] 289. (17.) Lanning [16, 167] points out that the higher the collusive price, the higher the gains from cheating, therefore requiring higher expenditures by the cartel for enforcement purposes. (18.) Sewell's response to the successful entry was to sue in federal court alleging a litany of antitrust violations. After three years of pre-trial arguments and analysis, the defendants prevailed on a motion for summary judgement as Judge McMillan concluded that the plaintiff had raised no issue of material fact for a jury to decide even though 13 feet of paper (over three million pages of documents) was filed with the court. The court observed that "the volume of paper which a modern law firm can produce is often greater than a busy district judge can read and evaluate with care." (19.) For merger policy, June 14, 1982 is a watershed date, marking the revision of the merger guidelines. Further revisions followed in 1984 and 1992, but most of the changes only clarified the 1982 guidelines. By limiting the merger sample to cases after June 14, 1982, we maximize the chances of finding decisions based on the improved economic model of competition in the 1982 Guidelines. (20.) We note that our model implies that in some circumstances the threat of entry will deter a price increase, in others a price rise will be ex ante profitable but not ex post as it induces entry, and in others entry occurs solely due to market growth. (21.) Given barriers to entry are an important input into the overall evaluation of the merger, all cases were clear as to whether the court considered barriers to exist. We note that the courts found for the defendant in all cases where no entry barrier was found. Almost identical results exist if the model focuses on the 24 Federal court decisions in the sample. More details about the data are available in Coate [7]. (22.) Sophisticated jurists are not necessarily more likely to find for the government, because the detailed economic analysis could be used to explain why collusion is unlikely even in a concentrated market with barriers to entry. One could claim these sophisticated economic findings are only made when barriers are high (otherwise the case is dismissed). Courts, however, generally make alternative findings to support their decision in the event of one finding (i.e., that entry barriers are low) being reversed on appeal. (23.) Exclusion of the economic sophistication variable does not markedly affect the magnitude or significance of the buyer power or DOJ variables. The new model (with t-statistics in parentheses) is 1.56(3.14) - 1.69(-2.41)BUYER - 1.47(-2.43) DOJ with a Chi-square statistic of 12.65. While the limited model predicts 90 percent of the high barrier outcomes, it is only able to identify half of the no barrier findings. Thus, the economic sophistication variable appears to add to the power of the model. (24.) Although it has previously been recognized that the DOJ has a worse record than the FTC [4], our sample of merger cases over the 1980's highlights the difference. The DOJ won only 29 percent of its cases (4 of 14), while the FTC won 69 percent (9 of 13) of its Federal court cases. The FTC has won slightly less than 50 percent of its completed administrative complaint cases, with a number of others currently on review. Calkins [4, 874] hypothesizes that the DOJ's failure may have been due to its litigation tactics, particularly with respect to market definition. Although Coate [7] presents some evidence that the Commission is more successful at establishing relevant markets than the DOJ, our results suggest a difference also exists with respect to barriers to entry. See, Baker-Hughes, supra note 2 for the most obvious example. Kovacic [15] hypothesizes that judicial ideology drives many antitrust decisions. To test this theory, we estimated another model that included a variable equal to the percentage of judges nominated by either President Reagan or Bush on the deciding court. The judge's variable had a coefficient with a positive sign (the opposite of that predicted by Kovacic's theory), but it was statistically insignificant. All the other variables retained some statistical significance.