Market Definition in the Computer Industry

January 1, 2003

Market definition is the cornerstone of EC competition law. In the computer sector it raises tricky conceptual and empirical issues because of the complexity of the products and services, the trading relationships, and the rapid rate of technical advance. In this article the principles of and difficulties raised by market definition in the computer industry are examined.

Market Definition Principles

The delineation of markets for the purposes of competition law is set out in the EC Commission’s Market Definition Notice published in 1997,[1] the UK Office of Fair Trading’s guidelines[2] issued under Competition Act 1998, and those of other Member States.

A “relevant product market” consists of products which are close substitutes in consumers’ eyes in conditions of effective competition. Under the EC Commission’s decisional practice, a multifaceted approach is used to determine substitution. The EC Market Definition Notice, and European caselaw, states that the relevant product “market comprises all those products and/or services which are regarded as interchangeable or substitutable by the consumer, by reason of the products’ characteristics, their prices and their intended use.” (para 9). It continues by stating that the relevant geographical “market comprises the area in which the undertakings concerned are involved in the supply and demand of products or services, in which the conditions of competition are sufficiently homogeneous and which can be distinguished from neighbouring areas because the conditions of competition are appreciably different in those area.” (para 8).

These definitions are imprecise, as they do not indicate the degree of substitution necessary to include or exclude products in a relevant product market. A more rigorous approach referred to in the EC Market Definition Notice is the hypothetical monopolist test. This looks at whether a hypothetical monopolist of a group of products or services can profitably raise relative prices by 5% to 10%. If it cannot, then the products are deemed part of the same relevant market and others are added to the product grouping until the point is reached where the hypothetical monopolist can profitably and permanently raise price. When this point is reached, there are no products outside the grouping which are sufficiently close substitutes to constrain the hypothetical monopolist from setting price above the competitive level.

There are several aspects of market definition in competition law that need to be clearly understood.

  • First, the relevant market is the narrowest grouping of products over which the hypothetical monopolist can raise price profitably. This is clear from the OFT’s Guidelines, which state that the market definition test must be applied to “the products (or group of products) at the centre of the investigation.” (para. 2.8), and “[T]he process starts by looking at a relatively narrow potential definition” (para. 3.1).

  • Second, the test seeks to identify consumers’ reactions which are fairly rapid. Under the OFT’s Guidelines rapid is defined as at “short notice” (para. 3.13) and “in the short run (for example, within one year)” (para. 3.15).

  • Third, the appropriate definition of the market depends on the product or complaint under the consideration, and varies on a case-by-case basis depending on the facts. For example, mainframe computers and PCs may at the same time be treated as one or two separate relevant product markets depending on the issue being considered. If one were examining market power in the mainframe sector, it may well be the case that PCs place a competitive constraint on the hypothetical monopolists’ ability to raise the price of mainframe computers. However, the opposite may not be the case since those who buy PCs and are subject to some alleged monopoly abuse by PC suppliers may not regard mainframe computers as close substitutes, and will therefore not be able to substitute away from PCs. In this case, PCs will be a separate market from mainframe computers. This is sometimes referred to as asymmetric market definition

  • Fourth, price discrimination for the same or a similar product can result in several markets being defined. If it is possible for a supplier to distinguish between classes of consumers (large corporate, SMEs, personal) and charge these different prices, then it may be possible to define different relevant markets for each. Switching costs in particular can generate a group of core captive customers which may form a separate market also. The hypothetical monopolist can exploit the fact that existing customers cannot switch easily to charge them a higher price than it charges new customers. These will form separate relevant markets if the switching costs prevent captive customers from substituting in sufficient numbers in response to a price increase.

  • Fifth, and related to the previous point, price differences between the same or similar products is not evidence that they are separate relevant markets. Different types of computers sold at different prices, even if these reflect different specifications and functionality, do not mean that they constitute separate markets. At the margin, the reactions of some consumers may place a competitive constraint on PC manufacturers from departing from price differentials which reflect quality differences. This is referred to as the “chain of substitution”. Thus markets with differentiated products sold at different prices may be horizontally linked by consumer reactions at the margin (see next point). Clearly, this implies that there are low switiching costs, at least for some consumers.

  • Sixth, and critically, in determining product substitutability it is the reaction of the marginal customers that counts. The hypothetical monopolist test only requires that sufficient customers at the margin react to a price increase, not all customers or the average customer. Thus, if in response to a hypothetical 5% price increase, a subset of existing customers reduce the quantity purchased of the product by more than 5%, then this will eliminate the monopolist’s ability and incentive to raise prices. The fact that the 95% of the remaining customers do not react is immaterial. It follows from this that survey evidence that many or most customers are price insensitive or regard the product as having no close substitutes does not establish that the product is in the relevant market.

  • Seventh, markets may be linked vertically where products are complements. This is of particular importance in the analysis of the computer sector because the primary product (hardware) will need complementary products such as software, consumables, peripherals and maintenance to function. It is here that considerable controversy and attention has recently centred; the decision as to whether the products constitute one or two markets has been critical for the assessment of market power and its abuse as shown in the US antitrust proceedings concerning Microsoft. This issue is discussed in more detail below.

Establishing Market Definition

Market definition should ideally be established on the basis of factual analysis. In one sense this is difficult because, in theory, determining whether two products are close substitutes is based on consumer and firm reactions at the competitive price. Since the prevailing price will often not be the competitive price, especially where the firm has market power, the data necessary to carry out the market definition test will not be available. Moreover, as is widely recognised, using the prevailing price may lead to the market being defined too broadly, a defect known as the Cellophane Fallacy after the celebrated 1956 [3] At a practical level while a variety of rigorous statistical techniques are available often these cannot be used because data are unavailable, incomplete and/or not in the correct form. Thus the evidence used will usually come from the industry, marketing literature, industry reports and market surveys. For example in ICL/Synstar, the OFT based its view that there was one market for computer hardware and hardware maintenance on survey evidence, industry views and simple statistics backed up by analysis, inferences and opinion.

Further, it is a common belief that rigorously testing for market definition is impossible – how does one test whether a hypothetical price increase is profitable? More rigour can, however, be given to the market definition using techniques which are not data hungry, and which can serve as a simple check on proposed or disputed market definitions.

One such technique is Critical Loss analysis.[4] Without going into the details, a Critical Loss is the percentage loss in output necessary to make a hypothetical price increase unprofitable.[5] All one needs to calculate the Critical Loss is data on price and the firm’s variable costs. While the Critical Loss does not estimate the actual lost sales expected from a 5% price increase – this would require estimating the price elasticities and involve considerable data and econometric analysis – it does give a figure whose plausibility can be tested to see whether the supply response would be below or exceed the Critical Loss. If the evidence shows that the reaction of consumers would exceed the Critical Loss, then the market is wider than proposed; and if it falls below the Critical Loss, the market has been correctly defined or is narrower.

This simple technique is increasingly used in US antitrust but not yet in EC law. For example, Critical Loss analysis was relied on in part in US v SunGard and Comdisco,[6] where the US District Court ruled against the US Department of Justice’s narrow market definition. The merging parties supplied disaster recovery services to business. The Department of Justice defined the relevant market as shared “hot-site” recovery services. The parties’ contended that the market was wider including hot site disaster recovery, cold-site disaster recovery, vaulting, high availability disaster recovery and mobile solutions. Using data on price and the firm’s variable costs, evidence was given that Sungard’s Critical Loss was approximately 5%. This meant that Sunguard could not profitably sustain a unilateral hypothetical price increase if it were to lose as little as 5% of its sales as a result of the price increase. The court agreed with the parties’ expert economists’ view that, given the large number of competitors and alternative solutions to which customers could turn, that the supply response would exceed 5%, and a broader market definition was appropriate. The merger was cleared.

Static v Dynamic Approaches

Another issue which has heightened importance to the computer sector is the role and importance of innovation and technical growth. Or, put more formally, whether the focus on short-term price changes to define relevant markets is appropriate for new economy industries such as the computer sector.[7] Some argue that this static approach to market definition is wrong and does considerable harm; others believe that the criticisms are exaggerated, and the adverse consequences small.

Underpinning these views are different “theories” of the new economy. One school believes that new emerging markets should be tightly regulated because developments in the formative stages of the industry will determine its future market structure. If these create monopolies and tight oligopolies, it will be impossible to impose structural remedies to re-create a more competitive industry at a later date. First mover advantages, network effects and tipping all point to a heightened risk of monopolisation which need to be prevented. According to this view, static market definition is adequate, and any harm it does is the lesser of two evils – the more serious danger is monopolisation.

The other school contends that this is a misreading of the economics and history of innovation. Proponents argue that initially concentrated markets are common and critical for innovation to take place and, secondly, that these markets are subject to the “perennial gales of creative destruction”, so that competition should be assessed not “in the market” but “for the market”. That is, a process of serial monopolisation occurs which is short-lived. The standard approach, it is argued, results in excessively narrow product market definitions, and too much regulatory intervention which may chill investment and stall innovation.[8] Market definition should use a wide lens to take into account longer term non-price factors.

The latter view was given a boost in 1995 with the publication of Gilbert and Sunshine’s influential article arguing that traditional market definition procedure was flawed in innovative industries.[9] The authors argued that market definition ignores technological change, and hence the effects on future competition and market power. They proposed defining an innovative market as well as a product market. The former would include firms likely to innovate in the market whether they currently produce or not. The competition authority would then be required to determine the effect of a merger on concentration in that innovative market, and finally the effect of this on innovative activity.[10] While this has not been widely endorsed, there are signs that competition authorities are paying more attention to dynamic analysis and the impact of intervention on longterm competitive factors.


In the analysis of the computer industry, it is often necessary to decide whether there is one market which includes hardware and complementary services, or several separate markets – a primary market (mainframe computers) and a number of aftermarkets consisting of the complementary services/products supplied both by the computer manufacturer and third-party suppliers. This issue arises regularly because firms in the primary market are usually vertically integrated, supplying both the primary and complementary products/services. Those who supply, or who wish to supply, only the latter often regard the competition as unfair, it is frequently observed that the prices for complementary services are high, giving rise to concerns that there is monopoly pricing.

Some theory is needed to examine the aftermarket issue.[11] A firm selling, say, both mainframe computer equipment (primary product) and computer maintenance (the aftermarket) faces a trade-off when it sets the price for maintenance services. Those who have purchased a computer will be “locked-in” by the initial capital costs of buying and setting up the computer system, and any subsequent costs of moving to a new computer system. This provides the commercial temptation to the seller to exploit those who have already purchased his computers – there is the possibility of what is sometimes called “installed user opportunism”. A higher price will allow it to earn increased profits on maintenance sales to consumers who have already purchased its computer equipment.

The incentive of the computer manufacturer to raise maintenance prices will have a feedback effect. Higher maintenance prices will reduce computer sales, because some or all potential buyers will take into account the higher expected costs of continuing maintenance services – so-called “life cycle” or “whole life” costs. If a high price in the aftermarket deters a sufficient number of potential buyers, then anticompetitive pricing is not a rational strategy. Here the word sufficient is to be understood as loss of computer sales which make a price hike in the aftermarket unprofitable. Thus if there is effective competition in the computer market this acts as an effective competitive constraint in the aftermarket. [12]

If, on the other hand, the competitive constraint is weak, because buyers do not have adequate information, anticompetitive practices in the aftermarket may be profitable. Employing the EC Commission’s “five percent test”, the integrated computer manufacturer and maintenance provider could raise maintenance prices by 5% without inducing losses in computer sales sufficient to wipe out the profits from “overcharging on maintenance”. In the latter case, aftermarket services and hardware would delineate separate relevant markets.

In summary there are two views which ultimately depend on the specific facts in each case:

  • View I holds that buyers of computers are sophisticated and have or should have adequate information to judge “whole life cost”. They are, therefore, unlikely to make wrong decisions, especially if they are repeat purchasers. Moreover, given the interrelationship between hardware and maintenance, maintenance costs have an effect on new computer sales, which in turn limits excessive charges for maintenance services. Thus, hardware sales act as a continuing discipline on manufacturers not to hike maintenance prices. That is, competition in the primary market implies competition in the aftermarket, and hence both constitute the one relevant market for competition law purposes.

  • View II casts doubt on the ability of buyers to foresee “whole life” costs at the time of purchase, and high switching costs give the seller the ability to exploit captive customers. This “lock-in” or “hold-up” theory results in two (hardware and maintenance) or many separate (hardware and a different maintenance market for each type of hardware) relevant markets.

It should be appreciated that the above analysis does not do justice to the complexity of pricing in the computer sector. Pricing of computers and related services is the outcome of a number of conflicting considerations which alter over the life cycle of the product.

For example, computer systems are products where network effects may arise. This is where the value of the product increases with the number of users of compatible systems (direct network effects) or with the number of complementary products produced for the system (indirect network effects). This effectively means that the value of computer systems rises with the size of the installed user base, and this in turn encourages suppliers of computer hardware and operating systems to underprice these with the object of recouping profits down the road when the installed subscriber base is larger and the value of the system greater to them. As a consequence, suppliers engage in so-called “penetration pricing” which entails lower hardware costs and higher prices for secondary products. The best example is mobiles where handset subsidies are given in order to lower the costs of access, and enhance the value of mobiles to users who can connect to more users. Such penetration pricing is noted (although not termed as such) in the OFT’s Guidelines at para. 5.11) which state that in such circumstance “it may be more appropriate to treat the primary and secondary products as separate markets .”.

Second, customer “hold-up” (View II above) may not be the only explanation for high aftermarket prices. High aftermarket prices may be a method of metering and hence a form of price discrimination used to identify consumers with intense demand. The implication is that high aftermarket prices are counterbalanced by low(er) hardware prices. This, argue some economists, is not anticompetitive in the same sense that a restaurant “overcharges” for wine or a pub “undercharges” for food in order to maximise profits.[13] What is at issue is whether the total price of the hardware and maintenance package is set at near competitive levels, not the individual prices. Moreover, it is not the role of competition authorities to regulate component prices but to ensure that the total price is set at a competitive level. Others argue that the practice is inefficient in the broader sense by (a) creating excessive hardware purchases and therefore encouraging buyers to economise on “overpriced” aftermarket services, and (b) leading to excess potential entry of independent service providers, and obviously a spate of antitrust actions.

In contrast it has been argued that aftermarket prices may be set too low. This was the allegation in the US Microsoft case where it was alleged that Microsoft’s bundling of its Web browser free with its Window operating system was predatory. The Department of Justice argued that operating systems (OS) and Web browsers were two separate markets, and that the bundling of the two products by Microsoft was an attempt to leverage its market power in the OS market onto the Web browser market. Microsoft countered that the two products were in the same market, and that the free provision of Internet Explorer was not anticompetitive because the marginal cost of adding IE to Windows was negligible and most likely negative because it boosted the sales of Windows. The case pointed both to the complexity of the competitive issues, and the critical role that market definition can play in the enforcement of competition law.

Checklist Approach

The discussion can be turned into a checklist of factors as has been done in some recent guidelines by competition authorities. It is likely that, say, hardware and computer maintenance would form one market if all or most of the following conditions existed:

· effective competition in the hardware market

· a high proportion of current to historical sales of hardware

· initial purchasers know and can predict lifetime service costs

· transparent service prices

· many repeat purchasers

· low switching costs

· open service markets

· high degree of technical change leading to short life of equipment

· business rationale supporting metering justification

· evidence that hardware discounting policy supports metering rationale.

Many of these factors have obvious implications, others less so. For example, View I places considerable emphasis on effective competition in the hardware market to discipline the supplier in the aftermarket. This force may be diminished if the proportion of new sales to installed subscriber base of hardware is low, so that the anticipated loss of sales due to high maintenance costs is less. Lost hardware sales are also unlikely to make an impact where the manufacturer is making no sales, losing market share and/or in a declining market. A supplier in a declining or obsolete market is likely to “harvest” its captive customers by charging supra-competitive prices for secondary products/services.[14]

High switching costs or infrequent purchase may dampen the link between vertical products/markets. The OFT’s Guidelines state (para. 5.10) that:

“If replacement is very infrequent or switching costs are high there may be a significant number of secondary product customers who are captive. Depending on the relative size of the primary market, even if new customers [appreciate] whole-life cost, the supplier may find it profitable to exploit these captive customers, implying that the secondary products will be a separate market.”

This allows for the possibility that, while some customers may be able to switch between products, others cannot, and that this may mean that the latter constitute a separate relevant product market.

Recent Cases

The case law has at different times adopted the two different views on aftermarkets. The European Court has held that there are two separate product markets, and that the hardware purchaser subsequently becomes captive irrespective of whether there was effective competition in the market for hardware. Thus, spare parts (Hugin v Commission[15]), consumables (Hilti v Commission[16]) and computer maintenance (Digital Undertaking[17]) have been defined as separate relevant markets. The US Supreme Court in[18] was more moderate in its analysis but came to the same conclusion. It held that, if there was effective competition in the hardware market, there could only be an abuse in the aftermarket if switching costs were high and if the information the purchaser received when the initial purchase was made was inadequate.

Some more recent decisions have taken the opposite view. For example, the OFT in ICL/Synstar[19] held that mainframe computers and hardware maintenance were one market. This case concerned a complaint by Synstar that ICL was abusing its dominant position under the Competition Act 1998 by not providing access to its full diagnostic service, and thus foreclosing the market. The OFT concluded that the relevant market was mainframe computers, that ICL was not dominant in that market, and that hardware maintenance was not a separate relevant market because buyers take into account the whole life costs when purchasing a mainframe computer. This was based on the fact that buyers are sophisticated, the costs of maintenance relative to that of the primary product was small, and price information transparent. The OFT dismissed Synstar’s view that there was a significant base of legacy systems with “captive customers”.

Concluding Remarks

Market definition is a critical first step in any competition law inquiry. It can determine the outcome, and it is an area where there is often considerable dispute and analysis in the practical enforcement of competition law. It has also been the case that competition authorities have had a more or less free hand in defining the market, often (as the critics would point out) in a contrived and narrow way that has little empirical support. Whatever the merit of this claim, this is now changing. The definition of markets and identification of market power is becoming more rigorous and quantitative, and the increased scrutiny of competition regulators by the courts and specialist appeal tribunals will ensure that regulators do their homework as the OFT found to its cost recently when its decision was annulled by the Competition Commission Appeal Tribunal for failure to properly define the market.[20] Having said this, the discussion above shows that market definition in the computer industry is complex and there will remain genuine areas of controversy over the theory, economics and probative value of practices and outcomes.

Cento G. Veljanovski, Managing Partner, Case Associates, London & Associate Research Fellow, Institute of Advanced Legal Studies, University of London. Case Associates are competition and regulatory economists with a particular expertise in the communications sector. Contact: © C. Veljanovski, October 2002

[1] EC Commission Notice on the definition of the relevant market for the purposes of Community competition law, Official Journal C 97/C 372/03, 1997, 5-13.

[2] Market Definition, OFT 403, March 1999. Under the proposed EC Framework Directive which will govern ex ante regulation of the communications sector, market definition will be critical. Draft Guidelines on market analysis and the calculation of significant market power under Article 14 of the proposed Framework Directive on common regulatory framework for the electronic communications networks and services 21.02.2002.

[3] U S v. E.I. du Pont de Nemours & Company 351 U.S. 377 (1956).

B. C. Harris & J. J. Simons, “Focusing Market Definition: How Much Substitution is Necessary?” Research in Law & Econ. (1989); B. C. Harris & C. G. Veljanovski “Critical Loss Analysis – Its growing use in competition law” (memo 2002).

Critical Loss is equal to Y/(Y + CM) x (100%), where Y = the hypothesized price increase of 5 or 10 % expressed as a proportion (.05 or .10), and CM = the contribution margin defined as the difference between the original price and average variable cost stated as a proportion of the original price.

172 F. Supp. 2d 172, 182, 186­92 & n.21 (D.D.C. 2001).

[7] C. G. Veljanovski, “EC Antitrust & the New Economy – Is the EC Commission’s view of the network economy right?” 10 Eu. Competition LR, 115 (2001).

[8] Bittlingmayer’s study of 21 major industries in the US covering 1947-1991 investigated the statistical association between antitrust case filings and investment. Each antitrust filing was associated with a significant decline in investment in the respective industry. G. Bittlingmayer “Regulatory Uncertainty and Investment: Evidence from Antitrust Enforcement” (2001) 3 Cato Journal

[9] R. Gilbert & S. Sunshine “Incorporating Dynamic Efficiency Concerns in Merger Analysis – The Use of Innovation Markets” 63 Antitrust Law Journal 569 (1995). Cf R. Rapp, “The Misapplication of the Innovative market Approach to Merger Analysis” 64 Antitrust Law Journal 19 (1995).

[10] OFT/Oftel, Innovation and Competition Policy – Economic Discussion Paper 3 (2002); W. J. Baumol, The Free Market Innovation Machine (Princeton UP 2002).

[11] There is now a literature on the economics of aftermarkets in the wake of the US Kodak and Microsoft decisions. See C. Shapiro & D. J. Teece, “Systems Competition and Aftermarkets”, 39 Antitrust Bulletin, 135 (1994); C. Shapiro “Aftermarkets and Consumer Welfare: Making Sense of Kodak”, 63 Antitrust Law Journal, 483 (1995); S. Borenstein, et al., “Antitrust Policy in Aftermarkets”, 63 Antitrust Law J. 455 (1995).

[12] Indeed recent theoretical works suggests that even if the primary market is effectively competitive, in the absence of perfectly contingent long-term contracts governing secondary/complementary services manufacturers are still able to price above the competitive price (greater than marginal costs) for aftermarket products and services. See S. Borenstein, J.K. MacKie-Mason & J.S. Netz, “Exercising Market Power in Proprietary Aftermarkets”, 9 J. Economics & Management Strategy, 157 (2000).

[13] B. Klein, “Market Power in Aftermarkets” in F. S. McChesney ed. Economic Inputs’ Legal Outputs – The Role of Economists in Modern Antitrust (1998).

[14] Katz, M. L. & C. Shapiro, “Antitrust in Software Markets” in A. Eisenach & T. M. Lenard, eds, Competition, Innovation and the Microsoft Monopoly: Antitrust in the Digital Marketplace, Kluwer Academic Publishers (1999).

[15] Hugin v Commission Case 22/78 [1979] ECR 1869.

[16] Hilti AG v Commission [1991] ECR II-1439.

[17] Commission Press Release IP/97/868.

[18] 122 S.C.T (1992).

[19] ICL/Synstar Case No. CA98/6/201 (20 July 2001). The OFT stated that it preferred to follow the EC Decision in Pelican v Kyocera, Competition Newsletter, Vol. 1

[20] Aberdeen Journals Ltd v Director General of Fair Trading & Aberdeen and District Independent Ltd (intervenor) Case No. 1005/1/01 (19 March 2002); [2002] CompAR 1.