8. Market Structure, Industry Concentration, and Barriers to Entry

The market share for a particular firm is defined as the firm’s sales divided by industry sales, where sales are typically measured by output or by total revenue.

For a discussion of the history of this index, see Hirschman (1964).

To see this, note that the variance (σ ²) can be written as σ ² = [Σms _i ²/n − (Σms _i /n)²]; market shares sum to 1 (when measured in decimals), so that Σms _i  = 1; HHI = Σms _i ². Thus, σ ² = HHI/n–1/n ². Solving for HHI gives HHI = nσ² + 1/n. For a discussion of variance, see the Mathematics and Econometrics Appendix at the end of the book.

When market share is measured in decimals, note that HHI is 0.24 or 2,400/10,000. The numbers equivalent is 1/0.24 or 4.17. These are frequented rounded off to the nearest counting number, which would be 4 in this case.

The obvious drawback with HHI is that it requires sales data on every firm in the industry.

US Department of Justice and the Federal Trade Commission, Horizontal Merger Guidelines, April 2, 1992 and April 8, 1997. A comparison of the old with the new structural standard is difficult, because the new 2010 Merger Guidelines have more lenient standards and consider a broader set of factors. For further discussion of the 2010 Guidelines, see Chap.​ 20.

This system of classifying industries has been in effect since 1997. Prior to 1997, data were published according to the Standard Industrial Classification (SIC) system.

For a review of these studies, see Kwoka (1985), Scherer (1980, 58–59), and Scherer and Ross (1990, 72–73).

A detailed comparison for all industries listed in Table 8.6 is not possible because some industries are defined differently in the NAICS system and the older SIC system.

In particular, see Demsetz (1973), Peltzman (1975), and Sutton (1991). For a review of the literature, see Scherer (1980, Chap.​ 9) and Martin (2002, Chap.​ 6).

This excludes microbrewers or specialty brewers that make European style ales and lagers and began entering the market in the mid 1960s. Although the number of specialty brewers exceeds 1,600 today, their combined market share is less than 6% and they generally compete for a different type of customer.

In 1970, the two largest firms were the Anheuser-Busch and Miller brewing companies. In 2008, they were Anheuser-Busch and MillerCoors (the combined sales of the Miller and Coors brewing companies which formed a joint venture in 2008).

To compare it to CR₄, HHI is divided by 100 so that it ranges from 0 to 100.

For a more complete discussion of the geographic market in brewing, see V. Tremblay and C. Tremblay (2005, Chap.​ 3).

For an excellent review of the influence of Gibrat’s work, see Sutton (1997).

See the Mathematics and Econometrics Appendix at the end of the book for a review of a normal distribution and a standard deviation.

In a lognormal distribution, the logarithm of firm size is normally distributed.

These are the US, Canada, the UK, Sweden, France, and Germany.

For examples of different definitions of barriers to entry, see Bain (1956), Stigler (1968), von Weizsacker (1980), and McAfee et al. (2004).

Of course, firms could invest in research and development, which can change technology and lead to an increase or a decrease in scale economies.

There are certain market settings where limit pricing can be effective. For example, Milgrom and Roberts (1982) show that limit pricing can effectively block entry when there is incomplete information. In their model, M has either the same or lower costs than PE, but only M knows if it is a low or a high cost producer. They show that if the probability that M is a low cost producer is sufficiently low, then it may be optimal for a high cost M to behave like a low cost M by charging a low price. This action will deter entry of PE.

Sutton (2007, p. 2359) argues that if fixed costs are not sunk, then many of Sutton’s conclusions are invalid because it would then be more appropriate to assume that firms play a static rather than a dynamic game. When an investment such as this is made before any output is produced, it is a quasi-fixed cost (see Chap.​ 2).

Sutton assumed a Cournot model which produces an outcome that lies between cartel and perfect competition. We will discuss the Cournot model in Chap.​ 10.

Recall from Chap.​ 7 that differentiation can be vertical (e.g., quality differences) or horizontal (e.g., location differences). Because assuming vertical (quality) differentiation produces such dramatically different results, we focus on vertical differentiation here (found in Sutton 1991, Chap. 3). When differentiation is horizontal, Sutton shows that the relationship between concentration and market size is less precise than for the homogeneous goods case found in Fig. 8.9 (see Sutton 1991, pp. 37–42). With horizontal differentiation, acceptable concentration and market size values include the curve and all points to the north and east of the curve in Fig. 8.9. This is called a “bounds approach,” because the model provides bounds on the set of outcomes rather than pinning down a precise relationship.

For example, if quality is a normal good, an increase in consumer income could increase sales and the demand for quality, which would induce firms to increase the quality of their products.

In his work on research and development and sunk costs, Sutton (1999, 2007) also argues that concentration can vary, depending upon the type of technological trajectories that are characteristic of an industry. If, for example, goods are relatively homogeneous and firms compete in research and development that is designed to lower production cost (i.e., they follow a single technical trajectory), as in the aircraft industry, then concentration tends to rise over time and remain high. Alternatively, when many submarkets or niche markets exist, as in the flowmeter (i.e., devices that control the flow of gases and liquids through pipes) industry, firms may choose to compete in one or a few submarkets and pursue a proliferation of technical trajectories. This tends to keep concentration from increasing over time.

Stigler (1966, 227) puts it this way: “if one can conceal the profitability of his situation, entry will be slower.”

Sutton (1999) also finds support for his theory when research and development expenditures are the primary source of sunk costs.

Control variables include dummy variables for countries and industries; β ₂ is a vector of parameters conformable to x. For further discussion, see Sutton (1991, Chaps.​ 4 and 5).

Other studies include Robinson and Chiang (1996) for a sample of US consumer goods industries, Matraves (1999) for the global pharmaceutical industry, Lyons et al. (2001) for a sample of industries in the European Union, and V. Tremblay and C. Tremblay (2005) for the US brewing industry. See Sutton (2007) for a more extensive survey of the empirical evidence.

A curve is convex when it lies above any tangent line to the curve.