Market Structure and Market Power - PDF

1. 5. 46 Foundations Appendix The Calculus of Competition 3 function or is mip, The comprise (i tum derined im's moblem may beas revenue sol...

1. 5. 46 Foundations Appendix The Calculus of Competition 3 function or is mip, The comprise (i tum derined im's moblem may beas revenue solved Ra) by we les the ailing cosfirm's (a) Price profitala, as, Market Structure and Market Power behavior implies R(q) = Pq. Hence: 7 (q) = R(q) - C (q) = Pq - C(q) (2.A1) Standard maximization then yields: dr- = P - C'(g) = 0 da (2.42) P = marginal cost C' (q) under perfect competition. For the monopoly firm, its output is the same as industry output @, that gives P(0). via the inverse demand curve. Hence, the monopolist's profit maximization problem is to Early industrial economics, working in the Structure-Conduct-Performance framework, choose output @ so as to maximize: recognize that decisi resen chard wil in face mean den minis maret. +(Q) = R(e) - C(l) = P(R)e - C(l) structure. Yet despite this change in perspective, it is clear that contemporary industrial (2.A3) economics must still address the issue of market structure, or the way the industry's Again, standard maximization techniques yield: producers are organized. In turn, this requires that market structure? and its close cousin di market power?be well understood in a way that allows distinction between different = P(Q) + QP (Q) - C' (Q) = 0 structures or different degrees of market power. de (2.A4) We know, for example, that markets work well when the market consists of numerous firms, each with a minimal market share. Yet such markets are relatively rare in the real P(Q) + QP(Q), is the firm's marginal revenue. The monopolist will maximize profit by world. Some markets have just two or three firms. Some have ten or twelve of unequal size. producing where marginal cost equals marginal revenue. For a linear demand curve of the In what ways is this difference important? If there are twenty firms, does it matter if one form of P(Q) = A - BQ we have P (Q) = -B. In this case, the firm's marginal revenue firm has 60 percent of the market and the other nineteen have just a bit more than 2 percent is A - BQ - BQ, or A - 2BQ. The monopolist's marginal revenue curve has the same each? Alternatively, can we measure market structure in such a way that enables us to make intercept as its demand curve twice the slope. Note that the profit-maximizing condition above can also be written as P(Q) - C' (Q) = -eP (Q) (2.A5) Dividing both sides by P(Q), we then have 3.1 DESCRIBING MARKET STRUCTURE P(Q) - C' (e) 11 One way to think about an industry's structure is to undertake the following simple P(l) =. P(e) (2.A6) where n is what economists call the elasticity of demand?a measure of how responsive the quantity demanded is to price movements. It is formally defined as: -n=.P ( Q ) / l P'(Q) (2.A7)  Market Structure and Market Power 49 48 Foundations 100 *ratios just given for the A, B, and C industries. Industry A appears more concentrated than does industry B using the CRs measure, but less concentrated when evaluated with the 80 CRy index. An alternative to CR, that attempts to reflect more fully the information in the concen- 60 tration curve is the Herfindahl-Hirschman Index or, more simply, the Hi. For an industry with N firms, this is defined as follows: 40 20 HI = Es} (3.1) 0 2 4 6 8 10 12 14 16 18 20 22 Rank size where s; is the market share of the ith firm. In other words, the HI is the sum of the squares of the market shares of all of the firms in the industry. Table 3.1 illustrates the calculation Figure 3.1 Some possible concentration curves of the HI for industry C in our example. If we measure market share in decimal terms so that a firm with 25 percent of the market has a share s; = 0.25, the HI for industry C is 0.20. Compare this to a maximum value of HI = 1.0, which would be the HI if the industry were a pure monopoly with one firm accounting for all the output. However, the practice is Figure 3.1 displays concentration curves for each of three representative industries: A, often to measure the shares in percentage terms in which case the HI for industry C is 2000 B, and C. The firms' ranked sizes are measured along the horizontal axis, again with the first firm being the largest. The cumulative market share is measured on the vertical axis. meisted in tes way. For industre mon, sly yalue of lions yield when shares are For example, Industry A has ten firms, each with a 10 percent market share. Industry B HI = 3,126.25, respectively. has twenty-one firms, the largest of which has a 55 percent market share. The remaining twenty firms each have a 2.25 percent share. Finally, in Industry C, there are three firms very to advane one telest has e orate over i dies have one. each with a market share of 25 percent and five firms each with a market share of 5 percent. combined influence of both unequal firm sizes and the concentration of activity in a few For Industry B, the vertical coordinates corresponding to the horizontal values 1 and 2 large firms. That is, rather than just reflect a single point on the concentration curve, the on the concentration curve are 55 and 57.25, respectively. This reflects the fact that the HI provides, in a single number, a more complete sense of the shape of that curve. It is largest firm has 55 percent of the market and the largest two firms have 57.25 percent this ability to reflect both average firm size and inequality of size between firms that leads conomy so preten he still simple concentratie one ve such reales disparity in pie. Table 3.1 Calculation of the HI for industry C Market Share Squared Market (%) Share Firm Rank s} is 80, 70.75, and 100, respectively. 1 25 625 625 Thus, for any n, we can define CR, as the n-firm concentration ratio. A little thought will 3 625 then i re tate road to a sa most industry ois the ation 25 5 25 information in the curve. Compare for example the four-firm and eight-firm concentration 6 25 25 " Those familiar with the GINI Coefficient typically used 10 measure income inequality will recognize the 8 5 25 Sum: 100 2000 (HI) concentration is derived. For curve furtheras the industrial details, structure see Damgaard analog of the Lorenz Curve from which the GINI Coefficient (2007). 50 Foundations Market Structure and Market Power 51 Consider nvo industries, each comprising ten firms, In industry A, the largest firm has a Table 3.2 Concentration measures for selected industries market share of 49 percent. The next three firms have market shares of 7 percent each, and HI CRA the remaining six firms have equal shares of 5 percent each. In industry B, the top four firms Industry share the bulk of the market with 19 percent apiece. The next largest firm accounts for 14 80.4 2425.5 Breakfast Cereals 40.9 602.8 percent, and the smallest five firms equally split the remaining 10 percent of the industry. Footwear Manufacturing 67.6 1448.8 Automobiles 19.6 160.2 a. Compute the four-firm concentration ratio and HI for each industry. Compare these Textile Mills 24.0 227.8 measures across the two industries. Which industry do you think truly exhibits a more Paper Manufacturing 47.5 806.5 competitive structure? Which measure do you think gives a better indication of this? Petroleum Refineries 58.2 1109.1 Pesticides 29.5 359.1 Explain. Pharmaceuticals 2025.3 67.1 b. Now let the three second largest firms in industry A merge their operations while holding Soap & Detergent Manufacturing 70.5 1995.3 onto their combined 21 percent market share. Recalculate the HI for industry A. Aluminum Sheet/Plate/Foil 63.4 2030.7 Computers & Peripherals Electric Light Bulbs & Parts 75.4 2258.3 33.6 388.8 Dolls, Toys, & Games 81.3 2652.2 Aircraft 58.1 1079.1 3.1.1 Measurement Matters: What Is a Market? Battery Manufacturing 60.5 2078.5 Telephone Apparatus 59.0 1828.5 Whether one uses CR, or HI as an overall measure of a market's structure, it should be clear Farm Machinery 72.8 1539.6 that the ability to make such measurements at all is predicated upon our ability to identify Tire Manufacturing 58.1 1094.5 Soft Drink Manufacturing 167.6 a well-defined market in the first place. In truth, this is not often easy to do. Consider, for 19.2 Medical Equipment & Supplies example, the soft drink Pepsi. Does Pepsi compete only against other carbonated beverages, Source: "Concentration Ratios in Manufacturing," Bureau of the Census, 2007 Census or should we also include such products as fruit juices, iced teas, and flavored milk in thinking about this market? Alternatively, think about multi-use products such as the Xbox of Manufacturing, January 2011. that, among other capabilities, can function as a DVR and a platform for streaming services such as Netflix and HBO GO. Is the relevant market for Xbox simply that of video games, or These data are published regularly by the Census Bureau. Table 3.2 shows both CRy and does it also include products such as TiVo? Unless we have a clear procedure for answering HI for a sample of well-known NAICS industries.? The two measures of industrial concentration, CR, and HI, are highly correlated, implying such questions, any summary measure of market structure such as HI will become an arbitrary statistic capable of being manipulated either upward or downward at the whim of that each gives roughly the same description of an industry's structure. Yet while the CRA and HI measures often tell the same story, the crucial question is whether or not it is the the researcher. An analyst can then make CR, or HI arbitrarily small or large by defining the market either broadly or narrowly. right story. As a little consideration quickly confirms, there is in fact reason to doubt that One commonly used set of market definitions is the set maintained by the US Census thisThe is the case.of the potential dissatisfaction with NAICS-based measures of industry source Bureau (in collaboration with other agencies) known as the North American Industry structure is that these measures are based more on commonality of production technologies Classification System (NAICS). The Bureau begins by first categorizing the output of and less on whether the output of the different establishments considered really do compete business units in the United States into broad sectors of the economy, such as manufacturing, with each other. Generally speaking, we would like to include production establishments primary metals, agriculture, and forestry products, each of which receives a numeric code. in the same market if the goods that they produce are close substitutes in consumption. Yet These sectors are then subdivided further, and each is given a two-digit code. The as noted, the NAICS procedure is based more on a similarity of production techniques. For manufacturing sector, for example, is covered by codes 31-33, while retail is covered by codes 44-45, and finance by 52. In each case, the two-digit codes are further disaggregated example, while wood flooring, resilient floor covering (e.g., linoleum tile), and ceramic tile are all flooring products, the NAICS categorization places them in different industries at the into the three-digit, four-digit, five-digit, and six-digit levels. Each additional digit represents three-digit and all higher levels (321 for wood flooring, 326 for resilient flooring, and 327 a further subdivision of the initial classification. The Bureau surveys companies and assigns each plant to a specific six-digit industry, defined primarily on the basis of similarity of for ceramic tile). In addition, the industries identified by the NAICS data calculate market production processes. For plants that produce more than one product, the Bureau makes shares on the basis of the national market when in fact practical considerations for a number an assignment based on the plant's primary product, as measured by sales. Once all the 2 Further details are available online at http://www.census.gov/epcd/www/naics.html. establishments are so assigned at the most disaggregated level, it is easy to add up to the A quite readable discussion of the advantages and disadvantages of each ratio is available in Sleuwaegen higher levels. It is also straightforward to work out the market shares and concentration and Dehandschutter (1986) and Sleuwaegen, Dehandschutter, and DeBondt (1989). indices according to who owns which plants for any industry at any level of aggregation. 52 Foundations Market Structure and Market Power5 3 of goods, such as newspapers and ready-mix concrete, imply that inter-firm competition is only local or regional while for others such as passenger aircraft or automobile manufacture competition in international. Reality Checkpoint If true product markets are defined by the substitutability between the various goods sold, then one would ideally like to define such substitutes on the basis of a direct measure of Concentrating on Concentration such substitutability. A formal measure that captures this relationship is what economists call the cross-price elasticity of demand ny. This is defined as the percentage change in Just as we can measure the fraction of an This of course does not mean that firms are demand for good i that occurs when there is a I percent change in the price of another good industry's output accounted for by its largest not getting bigger. If each firm grows at the j. The mathematical definition of this elasticity is firms, so we can measure the fraction of the same rate as the economy, each of us will economy's entire output, GDP, accounted for find ourselves employed in larger and larger (3.2) by its largest corporations. However, while it Agi Pj organizations over time even though concen- may make some sense to speak of a concen- tration is stable. White shows that this too Ap; 9i tration index based on just the top four to eight has been happening and that the size of the In brief, if ni is large and positive then goods i and j would be considered to be reasonably firms when speaking of a single industry, such average firm has, correspondingly, grown. close substitutes. * Precisely what values of ni, ought to be considered large and positive, a small number of firms would account for More recently, Elaine Tan (2008) has esti- much too little of GDP to seriously consider. mated both trends in concentration and trends though, is not immediately clear. A practical approach to market definition that implicitly embodies the cross-elasticity So, in the case of aggregate economic activity, in firm size inequality over a slightly longer measure is that employed by the Department of Justice and the Federal Trade Commission we consider concentration ratios such as CR so time period than White (2002), namely, from in evaluating mergers. This is typically referred to as the SSNIP standard. This is an or CRa00. Such measures can be constructed 1931 to 2000, using total assets to measure acronym for a "small but significant and non-transitory increase in price." Essentially, using data from the Census Bureau's Census firm size and assuming that the distribution of of Manufactures. Economist Lawrence White the authorities start with the narrowest possible definition of the market. For example, firm sizes is generated by a Champernowne (2002) made such calculations for the US for they might consider grocery stores selling natural, organic products such as Whole Foods. distribution. Like its close cousin, the logistic In that case, the SSNIP test would be whether, if all such natural foods grocery stores various years up to the end of the 20 century. distribution, the Champernowne distribution Some of his results are shown below. were monopolized, the monopolist could profitably raise the price of any of the goods is like the normal distribution except that it has by some small amount?usually 5 percent. If the answer is yes, then these grocery stores Aggregate Concentration for Manufacturing "fatter tails." Rather than look at the four-firm by themselves constitute a relevant product market. The intuition is that the hypothetical (Value Added Basis); Selected Years, concentration ratio, Tan (2008) looks at the monopolist could only profitably raise prices if other stores left out of this market did not 1947-97 asset share of the largest 500 US firms both for offer close substitutes and therefore the market analyzed is well-defined. If the answer to the Year CRso all corporations and limited to non-financial 1947 17% CR100 C R200 corporations. In both cases. Tan finds that SSNIP test is no, however, then the potential market is widened perhaps to all grocery stores 23% 30% natural or not, and the test of a profitable price increase by a hypothetical monopolist is 1958 the share of total assets owned by the top 500 23 30 38 repeated. If the answer is still no, then this process continues through ever wider definitions 1967 25 33 firms increased from 1931 to the middle of the 42 1977 24 33 44 second world war but then fluctuated with the of the retail food market until the "yes" moment is realized. 1987 25 33 end result that the share was about the same in While the SSNIP standard is clear however, its implementation is not. For example, 43 1992 24 32 2000 (55 percent) as it was in 1942, although in the early 2000s, Oracle, People-Soft, and SAP were the three leading firms engaged 1997 42 24 32 40 that share rose steadily during the 1990s. She in developing, manufacturing, marketing, and servicing the complicated software used by also finds that the inequality among firm sizes businesses to manage their large organizations. This software is often further subdivided These data suggest that, at least since the has grown. Both of these results are consistent into separate modules for a corporation's many parts. Thus, there is a software module for 1950s, aggregate concentration in manufac- with White's earlier findings. turing has shown no increasing or decreasing human resources management (HRM), for financial management (FM), for supply chain trend. It was approximately the same in 1997 management (SCM), and customer relations management (CRM). Besides the three firms as it was in 1958 whether one looks at the top Sources: L. White. 2002. "Trends in Aggregate already mentioned, there were also many smaller firms specializing in just one of these Concentration in the United States," Journal of 50, 100, or 200 largest firms. White shows Economic Perspectives (Fall): 137-60: and E. S. modules. The DOJ filed suit to block the proposed merger of Oracle and People-Soft. In its that somewhat similar results can be obtained Tan. 2008. "Champernowne Model Estimates of suit, the Justice Department argued that the relevant market based on an SSNIP test was the if one looks at all nonfinancial corporations Aggregate Concentration in the US, 1931-2000," or focuses on shares of employment or profit. Working Paper, University of London, Royal Hal- loway. ? However, a high monopoly price may inflate the cross-elasticity measure, a point originally emphasized by Stocking and Mueller (1955). Because the marginal consumer may look for substitutes when faced with a high monopoly price that would not be sought if the good were priced competitively, the cross price elasticity measured at the current industry price may be unduly large. Since 1997, the European Commission has also explicitly recognized this standard as a tool for market definition. 54 Foundations Market Structure and Market Power 55 I minus the fraction of production within the region that is shipped to consumers outside the region. In other words, LIFO is a rough measure of I minus the fraction imported an area prit such compler firms to bu janet poll be took ac in as while LOFI is roughly I minus the fraction exported. Based on a review of historical none on ich aly the tre ages vendovare plan. The many scale each im regar evidence, Elzinga and Hogarty (1978) suggest that for regions in which either the LIFO noel do even toimpete for these it a uniquel contracts because configured soft thethey an faru used sand on let tillins led selling and ims serving didsome an a and LOFI measures both exceed 75 percent, or where their average exceeds 90 percent, the presumption that the region is a well-defined geographic market is strong. cal not co predict that, while appropriate for muge halt an as, could never senate Yet while the Elzinga and Hogarty (1978) standard is easy to understand and measure, Metrol te are complex enterprises The Don a range that an i test come e the threshold value needed to establish a definite market is not. As noted, Elzinga and mark detention. However. Oracle disagreed and armand thate a proper application or Hogarty (1978) suggest a 75/90 threshold as a practical rule of thumb. Yet this critical SIP andand would indicate that a much broader market definition including al mos value is based on economic intuition and a rough sense of what is workable?not on any is preparing busines ofvare was appropriate." Thus, while the standard may be dies probability distribution that would allow us to construct confidence intervals or test for the proper way to impose it is more ambiguous. statistical significance. It is not entirely clear, for example, why thresholds of 65 and 80 would not be better at capturing actual market boundaries. Note too that the definition of the product market has implications for the definition of the geographic market and vice 3.1.2 More Measurement Matters: Geography and Vertical Relations versa. In the Oracle case, for example, the LIFO and LOFI measures vary depending on whether the product market consists of just the individualized business software sold by the big three firms or all business software sold by any firm. cqual need i0 define the marker's geographic space. Consider for example the newspape A further issue with both product and market definitions relates to the relationships that serve the same city. Assuming that printed news media is determined to be a relevan between firms operating at different stages of the production process. The delivery of a product market, we still need to determine which printed media constitute the citywide final good or service to the customer often represents the last of many steps. These include newspaper market. Is it just newspapers printed and sold within the city? Or do nation the acquisition of the raw materials, their transformation into a semi-finished good, the newspapers that are printed far beyond the city such as the New York Times and USA Toda refinement of the semi-finished good into a final consumer product, and thereafter, the also constitute some of the market? retailing. In economics jargon, the initial raw materials phase is typically described as the upstream phase after which the product flows "downstream" through the various stages Alternatively, consider a region with a number of software producers who sell i toward its final sale to the consumer. The relationship between upstream and downstream global market. If most of these firms' customers live outside the region then the geographic phases is therefore a vertical one, and there are several forms that this relationship can take. boundaries must be expanded. Similarly, it could be the case that local producers sell to An upstream producer may own all the subsequent phases in which case we say the firm is a broad regional or even global market. As a result, although most local consumers bo, from local producers, the presence of many consumers outside the local area means that the vertically integrated. Alternatively, an upstream producer may offer franchising agreements or long-term contracts to downstreamsellers. The existence and variability relevant geographic market extends beyond the local region. relationships can cause difficulty in measuring the structure of the market at any one stage In principle. an SSNIP test could be used to define the locations that constitute a geographic of production. For instance, there are many bottling companies so that conventional measures market just as it is used to define the goods that determine a specific product market. Ti of market concentration in the bottled can and soft drink industry are rather low. In turn, this is, once we have defined the relevant product market, we can define a region such as the suggests a fairly competitive market. However, the reality is that most bottling companies do local city and ask whether a firm that monopolized production within this area-say of not compete with each other but, instead, are tied through strict contractual agreements to use newspapers or software-could impose small but significant and non-transitory increase one of the national upstream suppliers, such as Coca-Cola or Pepsi.' As a result, there may in price. If the answer is yes, then the city does define a relevant geographic market. If the be much less competition among bottlers than the concentration measure would suggest. answer is no, then a wider geographic region needs to be considered. In sum, the interpretation of structural market measures such as CR, and HI is greatly In fact, the merger guidelines put out by the DOJ and the FTC point to exactly that kind complicated by a variety of factors beginning with the fact that any such structure of test. In practice however, alternative approaches are often used to define the relevam is endogenous. Much of our data collection organizes markets on the basis of similar geographic market. A common test in this regard is the Elzinga-Hogarty (1978) test. This production techniques rather than on a measure of substitution. Implementing more test considers two features of a geographic region to determine whether it represents a appropriate market measures such as SSNIP requires careful application and interpretation legitimate regional market. The first is LIFO (Little in from Outside) defined as I minus the of the evidence, and this can be problematical. Geographic definitions such as the Elzinga fraction of all the products consumed within a region that are produced outside the region. and Hogarty (1978) standard are easy to understand, but it is not obvious which values Returning to our newspaper example, if 20 percent of all newspaper subscriptions witho of that standard should serve as a threshold of market determination. The a local area are national newspapers, then LIFO would be 80 percent. The other measur while much of the theoretical work covered in this text will take the market at hand to be of the Elzinga-Hogarty (1978) test is LOFI (Little out from Inside). This is calculated as well understood, the real world measurement of actual markets is fraught with difficulties. ? 31theF.merger Sup 2d 1098: was 2004-2 allowed Trade Cas. (CH) 4.542. In the main, the court accepted Oracle's view d to proceed. Some authors, for example, Gort (1962) and, more recently, Davies and Morris (1995), have tried to obtain a precise, quantitative measure of the extent of vertical integration. 56 Foundations Market Structure and Market Power 57 Al the same time, one should not overstate the case. For example, categorizing industries on the basis of closeness of shared production techniques does not necessarily indicate production in the monopoly outcome. To drive the point home, recall that the perfectly competitive firm faces an infinitely elastic or horizontal demand curve. When such a large the substitutability necessary for competition. However, it has the advantage that it likely value is substituted for the elasticity term, equation (3.4) implies a Lerner Index of 0. Again, does group firms with similar production costs. This is important because any analysis that the perfectly competitive firm sells at a price equal to marginal cost. Note too that the Lerner links an industry's configuration to the underlying structure of its technology and costs Index can never exceed 1 and that it can only hit this maximum value if marginal cost is 0. only makes sense if the production technologies are sufficiently similar that we can make For an industry of more than one but not a large number of firms, measuring the Lerner general, industry-wide statements about a typical firm's cost structure. Index is more complicated and requires obtaining some average index of marginal cost. A particularly straightforward case in this regard is that in which the commodity in question is homogenous so that all firms must sell at exactly the same price. If this is so, then we can 3.2 MEASURING MARKET POWER measure a market-wide Lerner Index as: Throughout this chapter, we have been thinking about market structure in the quite literal sense of how the industry's production of output is allocated across different firms. We have P- s; MC; (3.5) seen how summary statistics such as the CR, or HI attempt to describe this configuration L I = of firms in an industry much as a census taker might use similar statistics to describe the number and size of families in a geographic region. A large part of the motivation for these Here, as before, s; is the market share of the ith firm and N is the total number of firms. measures is the desire to summarize succinctly just where an industry might lie relative The Lerner Index is a very useful conceptual tool and we will make reference to it to the ideal of perfect competition. There is nothing wrong with this structural approach throughout the remainder of this book. Like the CRy or the HI, the Lerner Index is a so long as one clear caveat is kept in mind: that a particular structure does not necessarily summary measure. The difference is that the Lerner Index is not so much a measure of how an industry's production is structured as it is a measure of the market outcome. The greater When we say that an industry is highly concentrated we are saying that the industry the Lerner Index, the farther the market outcome lies from the competitive case-and the output is dominated by a few firms, in contrast to the configuration that we associate with more market power is being exploited. In this sense, the Lerner Index is a direct and useful the competitive model. Does that necessarily mean then that prices charged in this industry gauge of the extent of market competition. are above what would prevail in a perfectly competitive market? The answer is not so For example, Ellison (1994) tries to get evidence on game-theoretic models of cartel straightforward. As we shall see in subsequent chapters, markets with even just two or three behavior. For this purpose, he studies railroad prices over time in the late nineteenth century. He estimates that, apart from price war periods, the Lerner Index was about 85 percent of firms may sometimes come quite close to duplicating the competitive or efficient outcome. The Lerner Index is one way to measure how well a market performs from an efficiency what it would be under pure monopoly pricing. In other words, the collusive behavior of point of view. The Lerner Index, LI, measures how far the outcome is from the competitive railroads at this time was capable of coming within 15 percent of the pure monopoly price ideal in the following way: distortion. However, much like the structural indices, the Lerner Index also has its problems. To p.- ? ? LI = begin with, calculating the Lerner Index for an industry runs into the problem of market (3.3) definition. Further, even when the market definition is reasonably clear, the Lerner Index is still difficult to measure. It is one thing to count the number and estimate the sizes of Because the Lerner Index directly reflects the discrepancy between price and marginal cost it captures much of what we are interested when it comes to the exercise of market power. the various firms in an industry. Measuring the elasticity of demand is trickier. Measuring For a competitive firm, the Lerner Index is zero because such a firm prices at marginal cost. marginal cost is even more difficult. Unfortunately, even small changes in the assumptions For a pure monopolist, on the other hand, the Lerner Index can be shown to be the inverse one makes about the data can lead to sizable differences in estimated price-cost margins. of the elasticity of demand-the less elastic the demand the greater the price-marginal cost For example, Ellison's (1994) study relied on data studied earlier by Porter (1983). Yet distortion. (See the Appendix of Chapter 2 for a formal derivation.) To see this, recall that Porter's (1983) estimate of the price distortion during collusive periods is only half as large for a monopolist the marginal revenue of selling an additional unit of output can be written as Ellison's (1994) estimate. as MR = P + Moreover, even when the Lerner Index is accurately measured its interpretation can ? ? Q. For profit maximization we set marginal revenue equal to marginal remain ambiguous. Suppose for example that each firm in an industry has to incur a ?? cost, or P + one-time sunk cost F associated with setting up its establishment. Assume further that each A l @ = MC. Rearranging and dividing by price P we obtain firm's marginal cost is constant. Because each firm needs to earn enough operating profit to P _ MC ?? ? cover its sunk cost, the equilibrium price level will need to rise above marginal cost. That = - is, the Lerner Index will need to be positive. However, the more positive that difference ?O P (3.4) ?the greater the number of firms that can cover is?the greater is the price-cost margin- where I/n is the inverse of the elasticity of demand. The less elastic is demand, or the the one time sunk cost. As a result, we might observe a high Lerner Index in a setting in smaller is n. the greater is the difference between market price and marginal cost of which there are numerous firms, none of which is very large. In such a case, the high Lerner 58 Foundations Market Structure and Market Power 59 Index might erroneously indicate little competition even though no firm has any significan. Remember that the elasticity of demand n is the proportionate increase in output in market power.® response to a given proportionate decrease in price. If the price were to fall from its current Conversely, the Lerner Index might underestimate market power in settings in which P level to the competitive level of P = MC, then output would rise to the competitive level cost-reducing innovations are important. Suppose for example that an industry has an old and not very efficient incumbent firm with high marginal cost. As long as demand is of QC. That is: somewhat elastic, such a firm may have no choice but to price relatively close to marginal (l° - l)/l (@° -@) = (P n- - MC) (3.8) " = (P - MC)/ P Because we also know that the industry Lerner Index is (P - MC)/P, we can rewrite incumbent's marginal cost when the relevant but unavailable comparison is the price with equation (3.7) as: (3.9) the potential rival's lower marginal cost.' WL WL' = Now recall from equation (3.4) earlier in the chapter that, for a pure monopolist, the 3.3EMPIRICAL APPLICATION: MONOPOLY POWER-HOW Lerner Index is given by: LI = (P - MC)/P = 1/n. Then, in this case, the deadweight BAD IS IT? loss relative to industry sales will be: (3.10) A recurrent question in antitrust policy is just how costly imperfect competition is for the economy overall. If the losses from monopoly power are not large, then devoting WL' = PO any significant resources to antitrust enforcement to prevent such losses is probably not That is, for the perfect monopoly case, the deadweight loss as a fraction of current worthwhile. Such scarce resources would be better used in, say, increasing homeland industry sales is simply one-half the Lerner Index or one over twice the elasticity of security or providing relief to hurricane victims. If the economic costs of market power are demand. The intuition is that as the demand elasticity increases, the welfare loss shrinks large, however, then allocating resources to combat the abuse of that power is likely to be because other goods are increasingly viewed as substitutes to the monopolized commodity. warranted. Hence, it would be useful if economists had some sense of just how serious the Note further the sensitivity of the welfare loss to the elasticity estimate. An estimate that losses from monopoly power actually are. n = 1.5 produces a welfare loss equal to 33 percent of revenue. An estimate of n = 2 In principle, economists have a clear measure of the economic loss caused by monopoly reduces this amount to 25 percent of revenue. That is, a 0.5 change in the elasticity estimate power. It is the deadweight loss or triangle that results from prices above marginal cost. yields an eight percent change in the welfare loss. In practice, however, measuring this loss is not so easy. This is because it requires getting The first person to make calculations along the foregoing lines on a large scale was estimates of cost and/or demand but, as with any estimate, these values are subject to some Arnold Harberger (1954). Using a sample of 73 manufacturing industries, Harberger (1954) error. Unfortunately, rather small changes in the estimates can lead to rather large changes took the difference in the five-year average industry rate of return and the five-year average in the estimated welfare cost. for manufacturing overall as an approximation of LI. Because he worked with industry To understand the issues involved, let us start with the basic measurement of the welfare data, and because none of the industries was a pure monopoly, Harberger (1954) could not or deadweight loss that results from pricing above marginal cost. As shown in Chapter 2, assume that his LI estimate is the inverse of elasticity of demand, as we did in equation this is the area whose height is given by the difference between price P and marginal cost (3.10). Instead, he combined his L/ estimates with an assumed demand elasticity of n = 1 in MC, and whose base is given by the difference between the competitive output @C that equation (3.9). The dollar value of these estimated distortions is then given by multiplying would the sell ifloss welfare P = WL MC is: and the actual market output @ that sells at the actual price P. Hence, WL' by industry sales PQ. When Harberger (1954) added these dollar values up and extrapolated the results across the entire economy he found a surprisingly small welfare cost of monopoly-on the order of one-tenth of one percent of Gross Domestic Product. W L = The low value of Harberger's (1954) estimate thus raised a serious question about the (3.6) cost-effectiveness of antitrust policy and litigation. Harberger's (1954) approach, however, did not go uncriticized. Bergson (1973) noted that to yield It is convenient to express this welfare loss as a proportion of total sales revenue PQ Harberger's (1954) procedure essentially used a partial equilibrium framework to obtain a general equilibrium measure. He demonstrated that, in principle, this could mean that WL Harberger's (1954) estimate considerably understated the actual loss. Cowling and Mueller 1P - MO (OC - e) WL' = (1978) used firm-level data for 734 companies in the United States and 103 companies in the Pe 2 United Kingdom. The use of firm-level data means that Cowling and Mueller (1978) could (3.7) apply equation (3.10) directly. Their estimated monopoly welfare costs range from 4 to 13 See, for example, Elzinga (1989). percent of GDP in the United States and from 4 to 7 percent in the United Kingdom-far Hovenkamp (1994), among others, has made this argument. larger than Harberger's (1954) estimates. 60 Foundations Market Structure and Market Power 61 Table 3.3 Hall's (1988) estimated Lerner Index for selected industries Substituting this result into equation (3.9), we obtain: Lerner Index (3.12) Industry W L HI WL' = Food & Kindred Products 0.811 Tobacco 0.638 is greater than one because MC is the marginal cost Textile Mill Products -0.214 Note that the term (P - MC) Apparel 0.444 that would prevail under competition. Aiginger and Pfaffermayr (1997) measure this Lumber and Wood 0.494 competitive MC as the marginal cost of the most efficient firm in the industry under the Furniture and Fixtures 0.731 assumption that this is the cost efficiency that would be required for competitive firms to Paper and Allied Products 0.930 survive. Effectively, their approach permits them to decompose the welfare cost of market Printing 0.950 power into two parts. One is the traditional welfare loss measure due to prices not equal to Rubber & Plastic 0.337 industry average marginal cost, P - MC. The other is due to the fact that market power Leather Products 0.524 allows the survival of firms with higher than minimum costs, MC - MC. Using data from Stone, Clay, and Glass 0.606 10,000 cement and paper firms in the European Union, Aiginger and Pfaffermayr (1997) Primary Metals 0.540 Fabricated Metals find that the total welfare loss of market power in these industries is on the order of 9 to 0.394 Machinery 11 percent of industry sales. Perhaps not surprisingly, they find that these welfare losses 0.300 Electric Equipment are largely due to the cost inefficiencies that imperfect competition permits. Thus, their 0.676 Instruments estimate of the traditional welfare loss measure is on the order of 2 to 3 percent, while the 0.284 Miscellaneous Mfg cost inefficiency loss is on the order of 7 to 7.5 percent. Extrapolating these estimates to the 0.777 Communication 0.972 entire economy would yield results that are considerably closer to the Cowling and Mueller Electric, Gas & Sanitary Services 0.921 estimates (1978) than those obtained by Harberger (1954). Motor Vehicles 0.433 In evaluating all of these estimates, it is useful to bear in mind at least two caveats. First, an Average 0.57 implicit assumption in all these calculations is that it is feasible to have perfect competition in all industries. As we shall see in the next chapter, however, costs and technology make this an unlikely outcome. In this sense, the estimates of welfare losses due to monopoly Robert Hall (1988) used a production theory approach to derive estimates of the Lerner price distortions are too high as there is no way in which all industries could be freed of Index for twenty broad manufacturing sectors in the United States. These are shown in such market power. Second, the measures are taken from data in which active antitrust Table 3.3. Domowitz, Hubbard, and Petersen (1988) obtained similar but generally lower enforcement has been the norm. In this sense, the measures are an understatement of the estimates of the Index using Hall's (1988) approach corrected for changes in raw material potential for monopoly-induced welfare losses. Had there been no antitrust enforcement, usage. Whereas Hall (1988) found an average price-cost margin of 0.577, Domowitz, there would have presumably been more market power abuses and the associated welfare Hubbard, and Petersen (1988) estimate the average to be only 0.37. Even this lower value, losses would have been greater. however, indicates a substantial degree of welfare loss in the manufacturing sector due to non-competitive pricing. An important source of variation in Cowling and Mueller's (1978) analysis is how Summary only reflects the top firm shares but also the advertising costs are treated in measuring LI. This calls attention to the critical importance This chapter has focused on the measurement differences in relative firm sizes. of the marginal cost measure in determining welfare losses. This issue has been addressed of market structure and market power. We are An explicit measure of market power is the more recently by Aiginger and Pfaffermayr (1997). They start by recognizing that without very often interested in summarizing the extent to Lerner Index. Because it is based on a compari- the pressure of perfect competition, firms can operate in an industry with different cost which an industry departs from the competitive son of price and marginal cost, this index directly efficiencies. Thus, the average industry marginal cost MC is very likely not the minimum ideal in a single number or index. The issue then addresses the extent to which the market outcome average cost that would be enforced if perfect competition were the rule. Aiginger and becomes whether and how we can construct such deviates from the competitive ideal. However, the Pfaffermayr (1997) then make use of a result (one that we shall derive in Chapter 9) from a a summary measure. need to measure marginal cost accurately, along standard oligopoly model. The result is that the industry price-cost margin measure using Concentration indices, such as the CR, or HI, with other measurement issues, makes the Lerner are explicit measures of a market's structure. MC is equal to the industry Herfindahl Index HI (scaled from 0 10 i), divided by the Index as difficult to employ as the structural elasticity of industry demand. That is: Both look at firm shares as a fraction of the indices. Estimates of the Lerner Index also serve industry's total output. Both encounter impor- as a useful starting point to estimate the actual tant problems, such as the difficulty of accurately P - MC _ HI defining the relevant market. The HI, however, efficiency costs of monopoly power. Many efforts (3.1 1) have been made to do this for the entire economy is generally preferred by economists since it not  ;. - 6 3 Market Structure and Market Power 62 Foundations As long as the foregoing problems are rec- in an attempt to get a general view as to how seri- ous the problem of market power really is. These ognized, the CR,, Hil, and Lerner Index mea- References Case of Coal," Antitrust Bulletin 23 (Spring): sures are useful starting points to characterize empirical studies have yielded a wide range of Aiginger and Pfaffermayr. 1997. "Looking at the 1-18. estimates of the aggregate deadweight loss as a an industry's competitive position. However, an Cost Side of Monopoly," Journal of Industrial Gort, M. 1962. Diversification and Integration in percentage of GDP. The lower bound estimate industry's degree of concentration and price-cost Economics 44 (September): 245-67. American Industry. Princeton: Princeton Uni- is that monopoly power imposes only a small margin do not materialize out of thin air. Instead, Bergson, A. 1973." O n M o n o p o l y Welfare versity Press. inefficiency cost of a few tenths of one percent ofthese indices all derive from the interaction of Losses," American Economic Review 63 Hall, R. 1988. "The Relation Between Price and GDP. However, upper bound estimates range as a number of factors. One of those factors is the (December): 853-70. Marginal Cost in U. S. Industry," Journal of high as 14 percent. A crucial parameter in such nature of production costs. The role that tech- Cowling, K., and D. C. Mueller. 1978. "The Political Economy 96 (October): 921-47. studies is the elasticity of demand assumed to be nology and cost play in shaping the industrial Social Cost of Monopoly Power," Economic Harberger, A. 1954. "Monopoly and Resource Journal 88 (December): 727-48. Allocation," American Economic Review 45 typical. outcome is examined in the next chapter. Damgaard, C. 2007. "Lorenz Curve." From MathWorld-A Wolfram Web Resource, cre- Hovenkamp, H. J. 1994. Federal Antirust Law Problems ated by Eric W. Weisstein. http://mathworld Policy: The Law of Competition and Its Prac- ?wolfram.com/LorenzCurve.html tice. St. Paul: West Publishing 1. The following table gives US market share data in percentages for three paper product markets Davies, S. W., and C. Morris. 1995. "A New Porter, R. 1983. "A Study of Cartel Stability: The Index of Vertical Integration: Some Estimates Joint Executive Committee, 1880-1886," Bell for UK Manufacturing," International Journal Journal of Economics 14 (Autumn): 301-14. Facial Tissue Toilet Paper Paper Towels of Industrial Organization 13:151-78. Company % Share Company % Share Company % Share Kimberly-Clark 48 Proctor & Gamble 30 Proctor & Gamble 37 Proctor & Gamble 30 Scott 20 Scott 1 8 Scott James River 16 James River 12 Georgia Pacific Georgia Pacific 12 Georgia Pacific 11 Other 9 Kimberly-Clark 5 Scott Other 16 Other 18 a. Calculate the four-firm concentration ratio for each industry. b. Calculate each industry's HI Index. c. Which industry do you think exhibits the most concentration? 2. Consider a market comprised of three firms. Firm I produces and sells 23 units per period. Firm The ingense make unis per period, while firm 3's periodic production and sales are 15 unis 3. not large enough to sustain two efficient-sized air carriers. Evaluate this argument. Why might Monopoly Air flights be so under-booked? Does this prove that there is no "market room" for a new competitor? We defined the Lerner Index as LI = 1/u where i is the absolute value of the elasticity of demand. We also showed that LI can be alternatively expressed as (P - MC)/ P. Use these relationships to show that LI can never exceed 1. What does this imply is the minimum demand elasticity we should ever observe for a monopolist?