Available in Adobe Acrobat PDF format

I. Introduction

State-owned enterprises (SOEs), also known as public enterprises, are owned by governments rather than by private investors. SOEs compete directly with private, profit-maximizing enterprises in many important markets. For example, government postal firms often offer overnight mail and package shipping services in direct competition with private delivery companies. Many public hospitals and educational institutions compete directly with private suppliers of similar services. Production by public enterprises can be particularly widespread in developing countries. During the 1980s, for example, public enterprises accounted for approximately 14 percent of gross domestic product in African nations, and approximately 11 percent in developing countries as a whole.[1]

SOEs are typically instructed to pursue goals other than profit maximization. Therefore, one might suspect that SOEs would act less aggressively toward their competitors than would private, profit-maximizing firms. We will demonstrate, however, that the opposite often is the case. Even though they may be less concerned with generating profit, SOEs may have stronger incentives than profit-maximizing firms to pursue activities that disadvantage competitors. Furthermore, these incentives can become more pronounced as the SOE's concern with profit becomes less pronounced.[2] These activities include setting prices below cost, misstating costs and choosing inefficient technologies to circumvent restrictions on below-cost pricing, raising the operating costs of existing rivals, and erecting entry barriers to preclude the operation of new competitors.

Incentives to act aggressively toward competitors can be created by governmental policy objectives that induce SOEs to value an expanded operating scale. To illustrate, SOEs are often instructed to increase local employment and/or to ensure that affordable service is provided ubiquitously to low-income families. Such directives can blunt incentives for profit maximization and thereby introduce a system in which the success of the manager of an SOE is measured more by the scale and scope of his operations than by the profit that his operations generate. Under such an explicit or implicit reward structure, SOEs may act as if they value expanded scale and scope--as proxied by revenue, for example--as well as, or instead of, profit. The enhanced valuation of increased revenue or expanded output leads the SOE to be less averse to the higher costs associated with expanded output and revenue. In aggressively pursuing expanded scale and enhanced revenues, SOEs may find it advantageous to engage in anticompetitive behavior against private, profit-maximizing enterprises.

For more than a century after the passage of the Sherman Act,[3] the United States led the world in developing a body of legal and economic principles for analyzing anticompetitive behavior by private enterprises. The U.S. Constitution, however, is thought to immunize much of the anticompetitive behavior by SOEs from U.S. antitrust law. Within the American federalist system, the Supreme Court has long addressed whether states may impose and supervise policies that reduce competition. These cases articulate the state action immunity in U.S. antitrust law, [4] which generously immunizes states (and, less generously, municipalities) from antitrust claims as long as they actively supervise the suppression of competition.[5] The crude rule of thumb is that private plaintiffs suing states for anticompetitive behavior generally lose.

Far less developed is the body of law on federal government activities that impair competition. If a federal SOE cloaks itself with the claim of sovereign immunity and if Congress has not consented to claims against the sovereign, including the sovereign's economic enterprises, a plaintiff has little chance to prevail in an antitrust proceeding against the SOE.[6] So it is not surprising that the antitrust jurisprudence on SOEs pales in comparison to American antitrust precedent on practically every conceivable business practice.

The other force that has contributed to the stunted growth of American case law on SOEs is capitalism itself. Unlike Europe, Australia, New Zealand, or even Canada, the United States has never embraced government ownership of enterprise. Railroads, telephone companies, electric utilities, banks, airlines, steel mills, automobile factories, and aircraft plants were routinely owned and operated by the state in Europe and much of the world.[7] Other than during wartime, the U.S. government generally has refrained from nationalizing and from directly managing private industries.[8] The 1927 nationalization of the radio spectrum used for wireless communications is a notable exception,[9] and one that is often criticized.[10] The most familiar state-owned enterprise operated by the federal government is probably the U.S. Postal Service.

Times have changed. The United States now feels the growing influence of the European Commission (EC) and the various national enforcement agencies around the world, as General Electric's failed acquisition of Honeywell in 2001 attests.[11] Less noticed than the defeat of the GE/Honeywell merger, but equally important for its long-term implications for the developments of competition law on all continents, was the EC's decision in 2001 regarding Deutsche Post AG, the German postal monopoly now undergoing privatization. The EC found that Deutsche Post had used profits from its state-granted monopoly in letter mail services to subsidize efforts to dominate the parcel delivery business in Germany by pricing below cost and undercutting competitors.[12] The EC ordered Deutsche Post to divest its parcel delivery business and to engage the new owner only on an arm's-length basis for any continuing commercial relationships.[13]

The European Union has generally outlawed state aid within its common market. Under Article 87 of the EC Treaty, "any aid granted by a Member State or through State resources in any form whatsoever which distorts or threatens to distort competition by favouring certain undertakings or the production of certain goods shall, insofar as it affects trade between Member States, be incompatible with the common market." Consolidated Version of the Treaty Establishing the European Community, 1997 O.J. (C 340) 173, available at http://europa.eu.int/eurlex/en/treaties/dat/amsterdam.html#0173010078. Article 87 is generally applicable in the liberalized postal services sector. See Notice from the Commission on the application of competition rules to the postal sector and the assessment of certain State measures to postal services, 1998 O.J. (C 39) 2, P7(a). The underlying objective of the prohibition against state aid is to prevent trade from being affected by advantages granted by public authorities, which, in various forms, distort or threaten to distort competition by favoring certain undertakings or certain products. See, e.g., Case C-39/94, SFEI v. La Poste (1996), E.C.R. I-3547, P58.

The Deutsche Post case could soon become instructive to SOEs owned by the U.S. government, as they become subject to various forms of competition law. In 2002, the U.S. Court of Appeals for the Ninth Circuit held in the Flamingo Industries case that the Postal Service was subject to federal antitrust law because "Congress has withdrawn the cloak of sovereign immunity from the Postal Service and given it the status of a private corporation."[14] The Ninth Circuit found that the Postal Service lost its sovereign status upon enactment of the Postal Reorganization Act of 1970, [15] which provided: "The Postal Service shall have the . . . power to sue and be sued in its official name." [16] The Supreme Court granted certiorari in the case for the October 2003 Term. [17]

Another significant development concerning competition law for SOEs is the complaint filed by United Parcel Service in 2000 against Canada Post [18] under Chapter 11 of the North American Free Trade Agreement (NAFTA).[19] Chapter 11 permits an investor of one signatory nation to initiate arbitration against another signatory nation for its failure to comply with NAFTA's obligations concerning foreign investment and regulation of monopolies.[20] Among other things, Chapter 11 enables a foreign firm to sue for damages caused by a nation's preferential treatment of its SOE, even though sovereign immunity might block an analytically identical case brought by a citizen of that same nation and styled as a violation of its domestic law. The applicable law is not necessarily that of any NAFTA country.[21]

The Flamingo Industries decision and the pending Canada Post NAFTA arbitration illustrate how American SOEs, such as the U.S. Postal Service, the Tennessee Valley Authority, and Federal Prison Industries, all could become the targets of analogous NAFTA complaints filed by Canadian or Mexican parties under NAFTA, as well as targets of antitrust suits filed by American plaintiffs under American law.[22] The purpose of this article is to begin to fill the void in the American law concerning anticompetitive behavior by SOEs. In particular, we develop a framework for identifying appropriate price floors for the products that SOEs offer in non-reserved markets, in competition with private producers.[23] In doing so, we explain why SOEs may be more likely to engage in anticompetitive activities than are private, profit-maximizing firms.[24]

We do not provide a comprehensive assessment of the benefits and costs of SOEs. In particular, we do not explain why the operation of SOEs may be preferred to operation by private, profit-maximizing firms. We also abstract from any innate cost differences between public and private enterprises, and we take as given the objective of the SOE. Therefore, this article is not designed to deliver broad policy prescriptions regarding the proper scope of SOEs.

We begin in Part II by examining the EC's decision in the Deutsche Post case. The EC found that pricing below long-run average-incremental cost (LRAIC) is inappropriate for both profit-maximizing firms and SOEs. We argue in Parts III through V that a higher price floor may be appropriate for SOEs. Part II also reviews standard multiproduct cost concepts, including LRAIC.

Part III examines the objectives of an SOE and the prices it will set when it pursues the identified objectives and faces no pricing restriction other than the restrictions imposed by competition in non-reserved markets. We identify conditions under which an SOE will choose to set prices below marginal production costs, even though such prices generally are considered to be predatory, and thus anticompetitive.[25] In addition, we discuss the methods that an SOE might employ to relax a binding prohibition against below-cost pricing. We examine an SOE's incentives to raise the costs of existing rivals and to erect barriers to keep potential rivals from entering relevant markets. Finally, we examine the implications of an SOE's ability to achieve cost advantages in a non-reserved market by virtue of its statutory monopoly in a reserved market. These advantages can result from economies of scope between the reserved market and the non-reserved market--economies of scope that the SOE's rivals are denied the opportunity to achieve.

Part IV explains why SOEs may have greater ability than private firms to act anticompetitively. This enhanced ability arises in part from the expanded powers and special privileges that often are extended to SOEs. These powers and privileges can help to ensure that an SOE, unlike its private competitors, does not need to recoup the costs of its anticompetitive behavior by subsequently raising prices in non-reserved markets.

Part V concludes that, in light of an SOE's greater incentive and ability to price below cost, the same cost-based standard that is employed to determine whether the prices set by a profit-maximizing firm are anticompetitive is not appropriate for SOEs.[26] Instead, the price floor for an SOE typically should be set higher than the price floor for a profit-maximizing firm. Part V identifies and analyzes the key factors that should inform case-specific guidelines regarding the extent to which price floors should be set higher for an SOE than for a profit-maximizing firm.

II. The European Commission's Deutsche Post Decision

The EC's 2001 decision in the Deutsche Post case raised novel legal questions concerning the application of pricing floors to firms in network industries, including SOEs. The EC declined to apply to SOEs in network industries a lower price floor than would be applied to privately owned firms in industries that do not exhibit significant network effects. In addition, the EC clarified, for purposes of analysis of cross-subsidization or predatory pricing, the difference between a firm's costs of supplying network capacity and its costs of supplying network usage.

A. Extending the Akzo Test to Network Industries 

The established predatory pricing rule at the time of the Deutsche Post case was from a 1991 decision of the European Court of Justice, AKZO Chemie BV v. Commission.[27] Under AKZO, a dominant firm that prices below average variable cost is presumed to have done so to eliminate its competitor and thus to have abused its dominant position.[28] In other words, when a dominant firm has so priced its products, it is unnecessary to prove an anticompetitive intent to establish that an abuse of dominant position has occurred. This "first branch" of the AKZO decision is distinct from the "second branch," which may apply when the dominant firm prices above its average variable cost but below its average total cost.[29]

The critical insight in the first branch of the AKZO test is that a firm that persistently fails to set a price for the single product it produces above the average variable cost of its operations will not be financially viable. Economic rationality will prevent a profit-maximizing firm from persistently pricing below average variable cost.[30] When prices are too low to generate profit--that is, when total cost exceeds total revenue at the established prices--a firm must confront the prospect of shutting down operations. In particular, the firm should continue to operate in the short run if and only if the loss incurred when the firm stays in business (that is, total cost less total revenue) is less than the loss incurred when the firm shuts down (that is, total cost less total variable cost). Hence, the economic decision to remain in operation distills to the following simple rule: remain in operation if and only if total variable cost is less than total revenue. Because total variable cost and total revenue are both divisible by the quantity produced, the rule can be restated as: A profit-maximizing firm should remain in operation if its average variable cost of producing a product is less than the price at which it sells the product.[31] If this condition is not met, the firm should discontinue its operations because doing so would reduce cost by more than it would reduce revenue and would thereby increase profit. The first branch of the AKZO rule captures this economic insight.

The need to articulate the AKZO rule more precisely for the case of a multiproduct firm was a subtlety of the Deutsche Post case that deserves emphasis. Strictly speaking, the EC's recognition of the multiproduct nature of the AKZO test did not require breaking new legal ground, for a close reading of AKZO reveals that it too involved a multiproduct firm. In an earlier stage of the AKZO litigation before the EC in 1985, Engineering and Chemical Supplies (Epsom and Gloucester) Ltd. (ECS), a small producer of organic benzoyl peroxide in the United Kingdom, alleged that AKZO Chemie BV, part of the large multinational group AKZO, had abused its dominant position in the European Economic Community (EEC) organic peroxides market.[32] ECS alleged that AKZO implemented "a policy of selective and below-cost price-cutting designed to damage the business of ECS and exclude it as a competitor."[33] ECS also claimed that AKZO's tactics had been concentrated in a relatively specialized submarket (the flour additives sector) in the United Kingdom and Ireland, which then accounted for the majority of ECS's sales. Such targeted behavior was allegedly intended to prevent ECS from expanding to the much broader EEC market for organic peroxides for the plastics industry. According to ECS, the actual price-cutting behavior was preceded by "threatened reprisals in the flour additives sector unless ECS agreed to abandon the polymer (or "plastics") market."[34]

When formulating its predatory pricing rule, the EC relied exclusively on generic cost attributes--fixed costs and variable costs--that typically apply to firms that supply a single product.[35] Nonetheless, the multiproduct nature of AKZO was implicit in the EC's formulation of its pricing rule, for it observed: "Besides being one of several facilities in the EEC where AKZO produces organic peroxides for the polymer industry, AKZO UK also manufactures benzoyl peroxide compounds for use as a bleaching agent in the commercial baking of bread together with other associated flour or milling additives."[36] Consequently, the EC's task in Deutsche Post of explaining the application of AKZO to a multiproduct firm in a network industry was more a challenge of economic explication than of doctrinal legal analysis.[37] The EC concluded that the appropriate multiproduct counterpart to variable cost is long-run average incremental cost (LRAIC).

B. Cost Definitions

To clarify the implications of using LRAIC as a predatory pricing floor, a brief review of this and other standard multiproduct cost concepts is useful. Most of the cost concepts discussed here have acquired clear meanings in economic theory and regulatory practice.[38] The distinction between capacity costs and usage costs, however, is less well understood. It therefore deserves clarification, particularly in light of its relevance to network industries in which multiproduct SOEs often operate.

Incremental cost is a generic concept that refers to the increase in the firm's total cost when it expands its output of a particular product or products by some specified increment, holding constant the amount of other products that the firm produces. Often, the increment in question is the entire output of the relevant product. Average incremental cost is incremental cost per unit of the output in question. Marginal cost generally differs from average incremental cost. The marginal cost of product X refers to the increase in the firm's total outlays that result from a small increase in the output of X. Marginal cost can be approximated by average incremental cost if the increment in question is small. But if the increment is large, marginal cost and average incremental cost can differ substantially.

The incremental cost of product X is computed by taking the difference between (1) the total cost of the firm if it were to produce all of its products and (2) the total cost of the firm if it were to produce all of its products except product X. This difference in these total costs is the incremental cost of producing X. If the resulting incremental cost measure is divided by the number of units of product X that the firm produces, then the result will be an average, or per-unit, estimate of that product's incremental cost--namely, the average-incremental cost of X, or AIC[X]. Although such calculations are relatively fact-intensive, they are routinely generated for regulatory proceedings in telecommunications, energy, and other network industries.

Average-incremental cost generally is the long-run figure obtained after plant and equipment are adjusted so as to minimize the average cost of the pertinent output. It is therefore often called long-run average-incremental cost. The average-incremental cost of X includes any fixed cost that must be incurred to produce that product alone.[39] A price floor equal to the average-incremental cost of product X would require the producer to recover in its revenue from the sale of X both the fixed costs and the variable costs attributable only to product X.

The stand-alone cost of service X (SAC[x]) is the outlay that would be required for a firm to produce service X and no other service. The concept of stand-alone cost also applies to combinations of services or products. The stand-alone cost of products Y and Z (SAC[Y,Z]), for example, is the cost incurred by a firm producing only products Y and Z. Stand-alone cost differs from incremental cost in part because the stand-alone cost of producing multiple products can include costs that are common to the group of products in question, even if the costs are not incremental to the production of any one of the products individually.

A cross-subsidy is present when the extra revenue derived from the sale of a product of a firm is less than the incremental cost of the product, but the firm nevertheless earns sufficient revenue from all of its products to cover all of its costs. When total revenue exceeds total cost, some products other than the cross-subsidized product must be generating the revenue required to offset the shortfall of the revenues of the cross-subsidized product. Thus, one can say that product X receives a subsidy if its revenues are inadequate to cover the costs caused by the firm's supply of X, but the firm's overall operations are financially viable.

As noted above, average-incremental cost includes any fixed cost incurred exclusively for the service in question. But the average-incremental cost of service X does not include any contribution toward any fixed costs incurred in common for X and some other service Y supplied by the same firm. In this simple two-product example, a particular common cost constitutes no part of the incremental cost of either X or Y by itself. But it is distinctly included in the incremental cost of X and Y in combination. Thus, to ensure the absence of cross-subsidies in this example, the following three conditions must all be satisfied: (1) the price of X must not be below the AIC of X; (2) the price of Y must not be below the AIC of Y; and (3) the prices of X and Y must be such that their combined incremental revenue is not less than the combined incremental cost of X and Y together. This case is the simplest version of the combinatorial incremental-cost test, introduced by Gerald Faulhaber.[40] A profit-maximizing firm will set prices that satisfy the combinatorial incremental-cost test. If it persistently failed to do so, the firm could increase its profits by setting different prices. The more general version of the combinatorial test states that the prices of the firm must be such that the resulting revenue of every product by itself, and the combined revenue of every combination of the firm's products, must at least equal the corresponding average-incremental costs of production.

The proper implementation of the combinatorial incremental cost test can help to avoid erroneous conclusions about appropriate price floors. To illustrate this point, suppose a firm produces three products, X, Y, and Z. Suppose further that when it is producing Z, the firm's incremental cost of producing X and Y is relatively high, but its incremental cost of producing X is low and its incremental cost of producing Y is also low. In this case, if price floors simply reflected product-specific incremental costs, the appropriate price floors for X and Y would both be low. However, if prices were set at these (low) floors, the firm's incremental revenue from X and Y would fall short of the incremental cost of producing X and Y. Under such circumstances, higher price floors for X and/or Y would be appropriate to ensure that the firm does not drive a more efficient supplier of goods X and Y from the market. Products X and Y may not be entirely unrelated products or services; instead, they may be variants of what usually would be seen as a common family of services. Although throughout the remainder of this article we speak of a test of the LRAIC of a product X, in fact the discussion should be read as applying equally to the LRAIC of a group of products.[41]

Prices that are set at or above product-specific average incremental-costs and that satisfy the combinatorial incremental-cost test preclude cross-subsidies. This is the case because a product or group of products is being cross-subsidized only if it generates incremental revenue below the relevant incremental cost of production.

C. Sunk Costs, Capacity Costs, and the Relationship of Lraic to Average Avoidable Cost 

The level of sunk costs and the relevant production decision affect the magnitude of LRAIC. To see why, notice that one might construct the incremental cost of product X by using either of two different measures of the cost of producing all but product X. The first measure would be the cost of producing all but X for a firm that had never produced X. The second measure would use the cost of producing all but X for a firm that initially produced all products and subsequently ceased production of X. Any sunk, irreversible costs that the firm must incur to produce X would be part of the incremental cost of X using the first measure, but not the second measure.

In an influential article, William Baumol advocates that "average avoided cost" (AAC), rather than average variable cost, be the price floor used to judge predation.[42] In essence, Baumol recommends the use of the second measure of LRAIC in which the firm initially produces all products and subsequently ceases production of the product in question. Baumol defines AAC as "the decremental rather than the incremental cost to firm B if it decides to exit"--that is, "the cost that B can escape or avoid by leaving." [43] The critical distinction between Baumol's measure of AAC and AVC is that the former includes "all pertinent portions of the product-specific fixed but avoidable costs, that is, all portions of such costs that can be escaped in the pertinent period of time." [44] Baumol advocates that the traditional Areeda-Turner rule for predatory pricing [45] be construed to use AAC instead of AVC as the price floor. Baumol's AAC standard remains the topic of continuing commentary.[46]

The question of the proper test for whether prices charged by profit-maximizing firms are predatory is the subject of a large and growing literature. We do not examine that literature in detail because our primary conclusions are not sensitive to the resolution of the questions addressed in that literature. Our focus is on the behavior of SOEs and the implications of this behavior for setting the floor for SOE prices relative to the floor for profit-maximizing firms. We argue in Part V that the floor for SOEs should be set above the corresponding floor for profit-maximizing firms. Our arguments apply whether the floor for profit-maximizing firms is set at long-run average-incremental cost, average avoidable cost, or some other measure of cost.

For expositional ease, we will focus on one measure of LRAIC in the ensuing discussion. We will focus on the first measure (in which LRAIC reflects the costs of producing a product for a firm that has never produced the product), in part, because we believe this is the more conventional interpretation of LRAIC.[47] This focus, though, does not imply that this measure of LRAIC, which includes product-specific sunk costs, is the only relevant measure. For the present purpose, it is simply the particular measure of LRAIC on which we focus for expositional simplicity. Our central qualitative conclusions regarding the appropriate price floor for an SOE relative to the corresponding floor for a profit-maximizing firm are not sensitive to the particular measure of LRAIC that we adopt.

Regardless of the measure of LRAIC that is employed, a universal service obligation (USO) can affect significantly a firm's LRAIC. To understand why, consider, first, the example of an independent, profit-maximizing barge owner who faces no obligation to provide service. Such an entrepreneur can design the barge's capacity however he likes, even if he is a dominant supplier. He can incur the fixed costs of constructing an enormous barge that is capable of accommodating the shipments of the highest-volume shippers. If he does so, he will need to forecast the likely demand of such shippers, and then he will bear the risk (or must contract around the risk) that his enormous barge will sometimes leave him with excess capacity. Alternatively, the barge owner may incur lower fixed costs to build a smaller barge, which will have a higher likelihood of operating at full capacity. If the barge owner faces excess demand with the smaller barge, he can raise his price and thus ration the scarce capacity of his barge to those customers who value the service at least as highly as the specified price.

Now, in contrast, consider how a state-owned postal operator might determine the characteristics of its delivery network. Governments often announce social policies of universal access, geographically averaged prices, and minimum service-quality standards. Such policies imply for the SOE a particular capacity requirement for its letter delivery network, as well as an "obligation to serve" all customers as the "carrier of last resort."[48] Once the SOE has built its delivery network according to these mandates, its "barge" must sail every specified day, regardless of whether it is nearly empty or entirely full. Unlike the owner of a real barge who bears no obligation to serve, the SOE may not raise prices (or offer particular customers inferior service quality) to ration scarce capacity on its delivery network when faced with excess demand. Similarly, a firm with a USO does not lawfully have the discretion to limit its output by restraining the capacity of its network. The firm's USO generally compels it to supply a level of network capacity that exceeds the level of network capacity that would be supplied in a competitive market by firms that do not have USOs. "Mandating that the [regulated incumbent firm] alone act as the carrier of last resort forces the firm to hold capacity in reserve to meet demand at peak load." [49] The SOE must therefore build its network with "reserve" or "standby" capacity that will accommodate peak demand.

To provide customers the option of sending letters on any given day to any given destination in the nation, the SOE must incur many kinds of costs that do not vary with the number of letters ultimately transported that day. One cannot attribute such network costs to any particular customer. The cost of standing ready to provide service on demand cannot be attributed to Customer A, rather than Customers B and C, because the firm is required to offer the same service to all of them. Consequently, "when a regulated firm has special-service obligations imposed upon it," "these obligations are appropriately treated as sources of common fixed costs for the firm . . . ." [50] The SOE's USO implicitly requires it simultaneously to offer both network access and network usage. Even if a customer sends no letters on a given day, he will nonetheless have enjoyed the option of doing so.[51] The delivery infrastructure was there to be used, if the customer had wanted to use it.

The option value of network access has implications for network capacity and for the distinction between network capacity costs and network usage costs. To ensure that consumers will be able to use the delivery network with relative ease, if they so desire, the network must be designed with capacity sufficient to accommodate expected demand, including peak demand during particular times of the day, month, or year. Such capacity costs are, by definition, fixed and sunk--they have already been made when a customer ships a parcel. Stated differently, if an existing customer stopped using the SOE's delivery network, these fixed costs of network capacity could not be avoided. The fixed costs of the network neither increase nor decrease when a particular customer, large or small, actually uses the network for parcel delivery service. Such fixed costs may include labor as well as capital.

The distinction between network capacity costs and network usage costs is intimately related to the scale of the network required to ensure a reasonable quality level for core operations. Absent strategic behavior, it would seem reasonable to count as overhead costs all costs associated with the network infrastructure required to ensure the established level of core service quality, even if they also support the delivery of non-core services. At the same time, such a rule for classifying costs provides strong incentives for the SOE that wishes to expand the scale and scope of its operations to set too high a quality standard for its core services. A high quality standard can justify expanded network investment (to ensure ample reserve capacity, for example), which can lower the SOE's incremental cost of providing non-core services. We discuss later an SOE's incentives and opportunity to make such strategic choices.[52]

D. The Deutsche Post Decision

Deutsche Post AG (DPAG) is a public limited company that in 1995 succeeded to Deutsche Bundespost Postdienst, which was a section of a special federal fund. Deutsche Post functions much like the U.S. Postal Service and has a profitable letter mail monopoly in Germany.[53] Deutsche Post also has a business parcel delivery operation that faces competition from United Parcel Service (UPS), Federal Express, and others. Deutsche Post's expansion from monopoly letter mail services into non-reserved parcel delivery services raises the issue not of a competitors' access to services controlled by Deutsche Post, but rather the use of infrastructure and cash flows available to Deutsche Post as a regulated public monopoly to distort competition in an adjacent market.

In the early 1990s, UPS complained to the EC that Deutsche Post was using letter mail monopoly profits to subsidize the sale of its parcel delivery services at below-cost prices. In July 1994, UPS asked the EC to stop the allegedly predatory pricing and to separate Deutsche Post's letter mail business from its parcel delivery businesses. In March 2001, the EC found that for five years Deutsche Post failed to cover incremental costs in its pricing of parcel delivery service, violating Article 82 of the EC Treaty.[54]

1. Cross-Subsidization in Non-Reserved Markets

The EC concluded that Deutsche Post's overall revenues in the reserved area (letter mail) exceeded its stand-alone costs, whereas revenue from its mail-order parcel service was below the incremental cost of providing that specific service.[55] Such mail-order parcels are not processed through the postal counter system but are collected by Deutsche Post directly at the customers' premises.[56] After collecting these parcels, Deutsche Post used the same infrastructure to process mail-order services that it used to process three other commercial parcel services.[57] These other commercial parcel services are (1) business-to-business parcels (B-to-B), (2) parcels that are handed in at post office counters and sent from one private person to another (P-to-P-services), and (3) mail-order returns (P-to-B-services). During the years investigated by the EC, mail-order services accounted for approximately 71 percent of the total volume of all commercial postal services every year.[58]

The EC defined cross-subsidization to occur when "the earnings from a given service do not suffice to cover the incremental cost of providing that service and where there is another service or bundle of services the earnings from which exceed the stand-alone costs."[59] The EC stated that "when establishing whether the incremental costs incurred in providing mail-order parcel services are covered, the additional costs of producing that service, incurred solely as a result of providing the service, must be distinguished from the common fixed costs, which are not incurred solely as a result of this service."[60] This passage indicates that the EC's interpretation of "incremental" is consistent with the interpretation employed here.

2. Universal Service and the Appropriate Price Floor

In making its calculation of the incremental cost of mail-order parcel services, the EC understood and considered the impact of Deutsche Post's universal-service obligation. Under the German Postal Universal Service Ordinance, Deutsche Post must maintain a capacity reserve large enough to cover any peak demands that may arise in over-the-counter parcel services while meeting statutory service-quality standards for those services.[61] Parcel services up to a weight of twenty kilograms are included in Deutsche Post's USO.[62] At least 80 percent of the parcels covered by the universal-service obligation must be delivered within two business days (known as D+2).[63] To comply with universal-service obligations, Deutsche Post must also maintain a certain minimum number of facilities, including at least one in every city with more than 2,000 inhabitants.[64]

As a consequence of these obligations, the EC concluded that even "if Deutsche Post were to stop offering a specific parcel service, it could not, unlike a private firm like UPS, cut back on staff and equipment in perfect proportion to the reduction in volume," because "some staff and equipment are also needed to provide over-the-counter-services that meet statutory quality standards."[65] Therefore, the EC held that "the costs of maintaining capacity arise independently of the services provided and the volume of parcels processed only as a consequence of maintaining capacity to allow everyone the standard option of having their parcels sent over-the-counter in the normal way."[66] These capacity costs consequently were found not to be attributable to any specific service, but rather to constitute common fixed costs. [67] Only where services other than over-the-counter parcel services are provided can costs be attributable to a specific service. Such costs cease to exist if the service at issue is stopped. The EC held that, to avoid subsidizing mail-order parcel services, Deutsche Post must earn revenue from this parcel service that at least covers the cost attributable (incremental) to producing the specific service, but not revenue sufficient to cover the network capacity costs that the SOE incurs as a result of its statutory USO.[68]

To determine the particular incremental costs of providing mail-order parcel services, the EC examined every stage in mail-order parcel processing (collection, sorting, long-distance transport, regional and local transport, and delivery) and asked whether these costs would be saved if mail-order services were discontinued.[69] The EC concluded that all costs of collection were fully attributable to the mail-order service because the parcels (unlike over-counter-services) are collected by Deutsche Post from the customer's premises and transported directly to one of the thirty-three outward and inward freight centers.[70] Regarding sorting, the EC found that capital cost of creating freight centers and delivery points could not be attributed to a particular service because these costs would be incurred as long as the USO for over-the-counter parcels applies to the postal service, but the staffing and equipment cost of sorting depend entirely on traffic volume and can be attributed in direct proportion to the mail-order services.[71] Costs of long-distance transport--staffing, equipment, and capital--were not attributed to a particular service because they could only be eliminated if universal-service obligations no longer applied to parcel services.[72] Half of the costs of regional and local transport (between freight centers and delivery points) were attributed to mail-order services.[73] Finally, delivery costs were divided into the cost of driving, which were not attributable to any specific service, and the cost of stopping and handing over the parcel, which was partially attributable to mail-order services.[74]

On the basis of this distribution between common costs and product-specific costs, the EC concluded that revenue generated from mail-order services did not suffice to recoup product-specific costs. Consequently, during the relevant period at issue in the case, every additional sale of mail-order services not only entailed the loss of at least part of the services' cost, but also made no contribution toward Deutsche Post's capacity maintenance costs.[75] By "remaining in the market without any foreseeable improvement in revenue," Deutsche Post was found to have "restricted the activities of competitors which are in a position to offer the service at a price that covers their costs."[76]

3. Remedy

In ordering a remedy for this violation of Article 82, the EC sought to "guarantee that competitors are not eliminated by offers which are not based on efficiency or superior performance, but solely on a basis of a price below the additional costs of providing the competitive service."[77] Deutsche Post was required to structurally separate its reserved letter mail business from its parcel delivery business, the latter to become a separate legal entity operating under a new name.[78] This new firm--Deutsche Post Euro Express--may procure necessary inputs (such as sorting, transport, and delivery) by purchasing them from Deutsche Post or its competitors, or by producing the inputs itself. If the new entity chooses to purchase inputs from Deutsche Post, Deutsche Post must provide them at market prices. In addition, all input services supplied by Deutsche Post to the new entity must be supplied to its competitors at the same price and on the same conditions. The EC intended that, with these structural and behavioral rules in place, Deutsche Post would have no incentive to treat the new firm differently from other parcel delivery services--in particular, to sell inputs to the affiliated firm at prices below incremental cost.[79]

III. Why State-Owened Enterprises May Have Stronger Incentives Than Private Firms to Act Anticompetitively

The EC's conclusion that Deutsche Post should not be permitted to set prices below LRAIC is undoubtedly correct. A question remains, though, as to whether SOEs should be required to set prices significantly above LRAIC in non-reserved markets. For the reasons presented in this Part and the next, we believe a price floor above LRAIC generally is appropriate for SOEs.

In this Part, we explain why SOEs may have stronger incentives than private firms to engage in anticompetitive activities. We demonstrate that when an SOE values an expanded scale of operation in addition to profit, it will be less concerned than its private, profit-maximizing counterpart with the extra costs associated with increased output. Consequently, even though an SOE may value the profit that its anticompetitive activities can generate less highly than does a private profit-maximizing firm, the SOE may still find it optimal to pursue aggressively anticompetitive activities that expand its own output and revenue. To illustrate, the SOE might set the price it charges for a product below its marginal cost of production, particularly if the product is one for which demand increases substantially as price declines. If prohibitions on below-cost pricing are in effect, an SOE may have a strong incentive to understate its marginal cost of production or to over-invest in fixed operating costs so as to reduce variable operating costs. A public enterprise may also often have stronger incentives than a private, profit-maximizing firm to raise its rivals' cost and to undertake activities designed to exclude competitors from the market because these activities can expand the scale and scope of the SOE's operations.

A. The Objective of An SOE

SOEs often have specific assigned missions and goals that go beyond profitability. For example, the U.S. Postal Service is charged with providing ubiquitous service throughout the United States at uniform rates across different geographic regions. Congress has mandated:
The Postal Service shall have as its basic function the obligation to provide postal services to bind the Nation together through the personal, educational, literary, and business correspondence of the people. It shall provide prompt, reliable, and efficient services to patrons in all areas and shall render postal services to all communities. The Postal Service shall provide a maximum degree of effective and regular postal services to rural areas, communities, and small towns where post offices are not self-sustaining. [80]

When it proposes rate increases (subject to Postal Rate Commission review), the Postal Service is required to consider the fairness, equity, and simplicity of its rate structure (across multiple services), as well as the relationships among prices, production costs, and the value of the service provided.[81] Such mandates indicate that, in contrast to the typical private firm in a capitalist society, SOEs seldom seek solely to maximize the profit they generate. The profit that SOEs are permitted to earn often is explicitly limited, and SOEs are commonly instructed to pursue goals that are distinct from, if not fundamentally incompatible with, profit maximization.[82]

In practice, an SOE is not a monolithic entity that faithfully executes its mandate. Rather, it is an organization comprised of many individuals, including managers who often have considerable discretion to pursue their own objectives. This discretion arises in part because SOEs are not subject to takeover threats and are generally less subject to the discipline of capital markets than are private enterprises. Even though the managers of private, profit-maximizing firms may have goals and interests similar to those of managers in SOEs, the discipline of capital markets will limit the freedom of private managers to pursue private interests that do not maximize shareholder value.[83] Managers of SOEs (and government officials who monitor them) often have considerable interest in expanding the scale or scope of their activities, in part, because a manager's abilities may be inferred from the size of the operations that he or she oversees.[84]

This preference for expanded scale and scope of operations suggests that SOEs might act as if they seek to maximize some combination of profit and a measure of operating scale.[85] In practice, revenue often serves as a convenient proxy for scale, because revenue provides a natural metric for comparing the outputs of multiple products.[86] Thus, one might view the SOE as seeking to maximize a weighted average of revenue and profit.[87] Different SOEs may value these two performance dimensions differently. To capture differences among SOEs, we employ the parameter w, which can range from 0 to 1, to denote the weight that an SOE places on revenue. We will also let 1 - w denote the corresponding weight on profit. By varying w, we can capture the objectives of different SOEs.

In the discussion that follows, we focus on this simple class of objective functions for the SOE. However, it is important to note that the key qualitative conclusions drawn in this discussion hold more generally. The conclusions hold, for example, if the SOE seeks to maximize a weighted average of output and profit, or if it seeks to maximize revenue (or output) subject to the constraint that its profit exceed some specified level. The key assumption is that the SOE values revenue or output as well as profit. Unlike a private firm that values only profit, an SOE may find value in increased revenue even when costs increase by as much or more. Its concern with revenue or output effectively induces the SOE to discount the cost of output expansion. Consequently, even though the SOE values the profit that its anticompetitive activities can generate less highly than does a private profit-maximizing firm, the SOE finds it optimal to pursue particularly aggressively anticompetitive activities that serve to expand its own output and revenue. In essence, the SOE's increased concern with output outweighs its reduced concern with profit in determining its interactions with competitors.[88]

Anecdotal evidence supports this theoretical analysis. For example, scholars and government analysts have noted that a series of specific investments and market-entry decisions of the U.S. Postal Service evince a proclivity for revenue maximization.[89] Moreover, that evident maximization of revenue can entail substantial risk and even financial losses. A March 2003 report filed by the Postal Service at the Postal Rate Commission noted the inherent riskiness of the agency's ventures into such non-postal services as prepaid calling cards and electronic bill payment:

Like any venture that depends on creating value and attracting revenue, the Postal Service needs the room to try new things, spread risk, stimulate innovation, and have flexible access to marketplace skills through partnerships. As with any new business initiative, it is reasonable to expect that some offerings will meet planned objectives while others will not. Undertaking new services requires a look forward and thus involves inherent risk.[90]

Some of the Postal Service's operations in non-postal markets have produced financial losses. The losses raise questions about whether concern with expanded scale and scope, rather than profit maximization, might have driven the decision to initiate these operations. In 1998, the General Accounting Office found that, from 1995 through 1997, the Postal Service lost more than $ 84 million on its development and marketing of non-postal products.[91] In addition to prepaid telephone calling card and electronic commerce services, those money-losing non-postal products included a remittance service, REMITCO, which the Postal Service ultimately discontinued.[92]

To analyze the implications of objectives other than profit maximization, it is convenient to introduce a formal statement of the class of objective functions under consideration. To do so, let n >/= 1 denote the number of products supplied by the SOE. Also let p[i] >/= 0 denote the price that the SOE charges for its i-th product, and let p [identical to] (p[l],...,p[n]) denote the various prices the SOE charges for its n products. In addition, let Q[j](p) denote the amount of product i that customers will buy when the SOE sets prices p. (Customers will buy more of any product the lower is its price.) Q [identical to] (Q[l](p),...,Q[n](p)) will denote all of the output produced and sold by the SOE. For simplicity, the ensuing discussion will focus on the setting in which customer demand for each of the SOE's products is independent of the prices charged for other products of the SOE. Unless otherwise noted, the discussion will also abstract from cost complementarities by assuming that the SOE's cost of supplying one product does not vary with the amounts of the other products it supplies. The function C(Q) will denote the SOE's cost of producing output Q.[93]

The first term in square brackets in expression (1) is the SOE's total revenue. Total revenue is the sum of the revenue derived from the sale of each of the SOE's n products. The revenue derived from the sale of any particular product (i) is simply the product of the number of units of the product sold (Q[j]) and the price (p[i]) at which each unit is sold. The last term in square brackets in expression (1) is the SOE's profit. Profit is the difference between total revenue and total operating cost. Thus, with the weight w applied to revenue and the weight [1 - w] applied to profit, expression (1) is simply the aforementioned weighted average of revenue and profit. Note that if w = 0, expression (1) collapses to the objective function of a private firm, which is to maximize profit only.

B. An Soe's Pricing

Before discussing the prices preferred by an SOE that seeks to maximize a weighted average of revenue and profit, we consider the prices that a private, profit-maximizing firm would set in the simple, static setting described above. It is well known that a firm will maximize profit in this setting by raising prices above marginal production costs by amounts that are inversely proportional to the sensitivity of customer demand to price.[94] In other words, the firm will set the price for a product close to its marginal cost of production when a higher price would cause many potential customers to decide not to purchase the product. In contrast, on products for which customer purchases do not decline much in response to price increases, the profit-maximizing firm will set prices well above marginal production costs.

This pricing strategy is summarized formally in Finding 1. The Finding refers to [SEE FORMULA IN ORIGINAL], which is the own-price elasticity of demand for product i. The price elasticity of demand for product i measures the rate at which customer purchases decline as the price of product i increases.[95] The larger is the price elasticity of demand for a product, the more pronounced is the decline in customer purchases as the price of the product increases.

Finding 1. The preferred prices of a profit-maximizing multiproduct firm are characterized by the following inverse-elasticity rule: [SEE FORMULA IN ORIGINAL]

 Expression (2) indicates that the profit-maximizing firm will always set the price of each of its products above its marginal cost of production.[96] In the simple, static setting considered here, reducing a price below marginal cost reduces profit, and so such pricing is not attractive to the profit-maximizing firm.

Now consider the prices preferred by a multiproduct SOE that seeks to maximize a weighted average of revenue and profit in the same setting. The prices that maximize expression (1) are characterized in Finding 2.

Finding 2. The SOE's preferred prices are characterized by the following modified inverse-elasticity rule: [SEE FORMULA IN ORIGINAL]

 Finding 2 reveals that the prices preferred by an SOE that seeks to maximize expression (1) follow a modified inverse-elasticity rule. To maximize a weighted average of revenue and profit, the SOE implements proportional markups of price over modified marginal cost, [SEE FORMULA IN ORIGINAL], that vary inversely with the price elasticity of demand. Prices are set further above modified marginal cost the more inelastic is the demand for the product.

Expressions (2) and (3) reveal that the SOE's pricing rule is the same rule that a profit-maximizing firm follows, except that marginal costs are scaled down by the factor 1 - w to reflect the SOE's reduced focus on profit. The greater is its focus on revenue rather than profit (that is, the larger is w), the more the SOE discounts marginal costs in the modified inverse-elasticity rule.[97]

This discounting of marginal costs reflects the fact that, as the SOE becomes more concerned with revenue relative to profit, it becomes less averse to the higher costs that arise from increased output.[98] Consequently, the SOE favors more highly the expanded output and revenue that result from low prices on those products for which competition from alternative suppliers is most pronounced. The rule suggests that when such competition exists, a reduced focus on profit can lead the SOE to set particularly low prices for the products on which it faces the most intense competition. This conclusion supports John Lott's observation that an SOE might set the price of a product below its marginal cost of production.[99]

Conclusion 1. Even in the absence of predatory intent, a SOE may set the prices for some of its products below their marginal costs of production. The SOE will tend to prefer below-cost prices when its focus on profit is more limited and when customer demand for its products is more sensitive to price.[100]

Conclusion 1 holds because, even though profit declines as the SOE reduces price below marginal cost, revenue can increase as price declines. Therefore, if the SOE's relative valuation of revenue is sufficiently pronounced, or if customer demand for some of its products is sufficiently sensitive to price (or both), then the SOE may choose to set some prices below marginal production costs even when demands are independent.

To illustrate this more general point, consider the simple case where the elasticity of demand for a product does not vary with the price of the product.[101] In this case, an SOE with the objective function specified in expression (1) will set a price below marginal cost on all products for which the price elasticity of demand exceeds 1/w. For example, if the SOE values revenue and profit equally (so w = 0.5), the SOE will set a price below marginal cost on all products for which the price elasticity of demand exceeds 2.[102]

Pricing below marginal cost is undesirable because it introduces three forms of inefficiency: allocative, productive, and dynamic.[103] Allocative inefficiency arises when too much of a service is consumed in the sense that the resources employed to produce the marginal units of the service exceed the value that consumers derive from those units. When the marginal cost of producing a service exceeds the marginal value that consumers derive from the service, society as a whole would benefit if less of the service were produced. The reduced production level would allow resources to be redeployed to produce other services that consumers value more highly. Pricing a service below its marginal cost of production results in allocative inefficiency because some consumers (those whose valuations exceed the established price but are less than the marginal cost of producing the service) choose to purchase the service even though their marginal valuation of the service is less than the marginal cost of producing the service. Although those who purchase the service that is priced below cost benefit from the low price, their benefit comes at the expense of consumers of other products. The costs of producing other products, and thus the prices of these products, could be reduced if production resources were redeployed to more highly valued uses. Most importantly, on balance, society as a whole would gain if the allocative inefficiency were eliminated by ensuring that prices do not fall below marginal production costs.[104]

Productive inefficiency arises when a service is produced by a firm that is not the least-cost provider of the service. Pricing below marginal cost can introduce productive inefficiency by rendering unprofitable the operation of the most efficient producers.[105] Industry costs increase, and thus the net benefits to society decline, when below-cost pricing limits or precludes production by the least-cost supplier.

Dynamic inefficiency arises when the level of investment in research and development that maximizes the net value to society is not undertaken. When one firm prices a service below the cost of producing the service, the profit that other firms anticipate from supplying the service is reduced. Consequently, firms may decide to devote research and development efforts to other services rather than to the service that is priced below cost. Although such a reallocation of investment in research and development will increase profit, it may reduce the net benefits that accrue to society as a whole.[106]

C. Avoiding Restrictions on Below-Cost Pricing

The foregoing analysis considers the prices preferred by an SOE when its pricing flexibility is unrestricted. In practice, an SOE may face restrictions on feasible prices. For example, an SOE might be prohibited from pricing below marginal cost by law, by regulatory fiat, or by the firm's charter. Profit-maximizing firms also may face regulatory or legal constraints on pricing below marginal cost. We now explain how firms might attempt to relax a binding prohibition against below-cost pricing, and why a public enterprise can have stronger incentives than a profit-maximizing firm to relax such a prohibition.

First, a firm might attempt to relax a binding constraint against pricing below marginal cost by manipulating accounting data so as to understate its actual marginal cost.[107] Such understatement might be achieved by classifying as overhead production costs some or all of the costs that truly vary as output varies. For example, the firm might count as central management some personnel whose daily efforts are devoted entirely to the delivery of the product in question. An alternate way for the firm to understate its true marginal cost is to record, as variable costs incurred in the provision of a different product, costs that are truly incurred in producing the product whose price the firm would like to set below marginal cost. For example, the firm might claim that materials and supplies employed to produce the product in question were employed to produce a different product.

Intentional understatement of marginal production costs is likely to entail personal risk. Laws against fraud carry severe financial penalties, and career prospects can be dimmed for managers who are suspected of knowingly reporting false information. But even if the SOE bears the full costs of the manipulation (and thus does not discount those costs by the factor 1 - w, as it implicitly discounts production costs), the associated benefits may outweigh the costs. Most importantly, when the SOE values more highly the expanded output and revenue that result from the lower price that the understatement facilitates than does a private, profit-maximizing firm, the SOE will be more likely than its private counterpart to understate its costs.[108] For emphasis, we record this observation formally as Conclusion 2.

Conclusion 2. An SOE that values both revenue and profit typically will have stronger incentives than a profit-maximizing firm to understate its marginal cost of production in order to relax a binding prohibition against pricing below cost. The less profit-oriented is the SOE, the greater is this incentive, all else equal.[109]

Next consider a more subtle strategy that an SOE might pursue to relax a binding prohibition against pricing below cost. Rather than misstate its true marginal cost, an SOE might choose to operate with an inefficient technology that secures a relatively low marginal cost at the expense of a particularly high overhead cost. In practice, a firm might do so by installing costly general-purpose equipment on a large scale and thereby reduce the need for project-specific equipment. It might also do so by, for example, retaining a large on-site staff with broad legal, engineering, computing, and/or marketing expertise that can substitute for specific expertise on individual products.

More generally, suppose that the SOE has a choice among production technologies and suppose that this choice is indexed by the amount of an overhead resource that the firm employs. For expositional convenience, refer to this resource as "capital."[110] The more capital the firm installs, the lower are its variable and marginal costs of production. In this setting, the following conclusion is readily verified.

Conclusion 3. An SOE that values both revenue and profit typically will have stronger incentives than a profit-maximizing firm to overinvest in capital in order to relax a binding prohibition on pricing below cost. The less profit-oriented is the SOE, the greater is this incentive, all else equal.

The more highly the SOE values expanded scale and scope relative to profit, the more it benefits from the expanded scale that a lower price provides and the less concerned it is with the associated c