Market Concentration: Measurement and Trends Firm Fi rman anssyah, S. S.E. E.,, M. M.Si Si., ., Ph Ph.D .D FEB UNDIP
Introduction •
•
•
•
Analisi Analisiss kompetis ompetisii perusa perusahaa haan-p n-peru erusah sahaan aan,, mencak mencakup up key-elements of industry structure structure The most important characteristics characteristics of industry structure include the number and size distribution of , exit, and the degree of product differentiation Seller concentr concentration ation refer referss to the the first first of these these elements: elements: the number number and size distribut distribution ion of firms. In empirical research in industrial organization, seller concentration concentration is probably the most widely used indicator of industry structure
Markets and industries •
•
The definition definition of a market market is straight-f straight-forwa orward rd in theory theory, but often more problematic in practice. Serviceable Serviceable theoretica theoreticall definitions definitions can be found found in the the works of the earliest, nineteenth-century economists. –
–
•
The entire territory of which parts are so united by the re ations o unrestricte commerce t at prices t ere ta e the same level throughout, with ease and rapidity (Cournot, 1838 pp. 51–2Fn). prices of the same goods tend to equality with due allowance for transportation costs (Marshall, 1920, p. 270)
For practic practical al purposes, purposes, the definition definition of any any mark market contains both a product dimension and a geographic dimension
•
•
•
•
The product market definition should include all products that are close substitutes for one another, both in consumption and in production. Goods 1 and 2 are substitutes in consumption if an increase in the price of Good 2 causes consumers to switch from Good 2 to Good 1. The degree of consumer substitution between Goods 1 and 2 can be measured using the cross-price elasticity of demand Good 1’s elasticity of demand with respect to a change in the price of Good 2 is:
A large and positive cross-price elasticity of demand indicates that the two goods in question are close substitutes in consumption (for example, butter and margarine).
•
•
•
•
•
In contrast, a large and negative cross-price elasticity of demand indicates that the two goods are close complements (for example, camera and film). However, this could also imply that they should be considered part of the same industry. CD players and amplifiers might be grouped together as part of the hi-fi equipment industry. But what about cars and petrol? These goods are also complementary, but would it be sensible to include motor manufacturers and oil companies in the same industry group? Good 1 produced by firm A, and Good 2 produced with similar technology by firm B are substitutes in production if an increase in the price of Good 1 causes firm B to switch production from Good 2 to Good 1. In this case, firms A and B are close competitors, even if from a consumer’s perspective Goods 1 and 2 are not close substitutes. For example, Good 1 might be cars and Good 2 might be military tanks. increase in the price of cars. But on receiving the same price decide to switch to car production. The degree of producer substitution between Goods 1 and 2 can be measured using the cross-price elasticity of supply. Good 1’s elasticity of supply with respect to a change in the price of Good 2 is: signal, a tank producer might decide to switch to car production. The degree of producer substitution between Goods 1 and 2 can be measured using the cross-price elasticity of supply. Good 1’s elasticity of supply with respect to a change in the price of Good 2 is:
•
The geographic market definition involves determining whether an increase in the price of a product in one geographic location significantly affects either the demand or supply, and therefore the price, in another geographic location. If so, then both locations should be considered part of the same geographic market. In principle, a similar analysis involving spatial cross-price elasticities could be used
Official schemes for industry classification •
The International Standard Industrial Classification of All Economic Activities is a United Nations system for classifying economic data. The United Nations Statistics Division describes it in the following terms: –
internationally, in classifying data according to kind of economic activity in the fields of production, employment, gross domestic product and other statistical areas. ISIC is a basic tool for studying economic phenomena, fostering international comparability of data, providing guidance for the development of national classifications and for promoting the development of sound national statistical systems.
ISIC Indonesia •
kbli_2009.pdf
Measures of concentration •
Seller concentration, an indicator of the number and size distribution of firms, can be measured at two levels: , located within some specific geographical boundary; –
2) for all firms classified as members of some industry or market, again located within some specific geographical boundary.
•
•
• •
The first type of seller concentration, known as aggregate concentration, reflects the importance of the largest firms in the economy as a whole Typically, aggregate concentration is measured as the share of the n largest firms in the total sales, assets or employment (or other appropriate size measure) for the economy as a whole The number of firms included might be n = 50, 100, 200 or 500 Aggregate concentration might be important for several reasons: –
–
–
if aggregate concentration is high, this might have implications for aggregate concentration data might reveal information about the economic importance of large diversified firms, which is not adequately reflected in indicators of seller concentration for particular industries; if aggregate concentration is high, this might indicate that the economy’s largest firms have opportunities to exert a disproportionate degree of influence over politicians or regulators, which might render the political system vulnerable to abuse
•
The second type of seller concentration, known as industry concentration or (alternatively) market concentration, reflects particular industry or market
•
economists have employed a number of alternative concentration measures at industry level. To assist users in making an informed choice between the alternatives that are available, Hannah and Kay (1977) suggest a number of general criteria that any specific concentration measure should satisfy if it is to adequately reflect the most important characteristics of the firm size distribution: –
–
–
–
Suppose industries A and B have equal numbers of firms. Industry A should be rated as more highly concentrated than industry B if the firms’ cumulative market share (when the firms are ranked in descending order of size) is greater for industry A than for industry B at all points in the size distribution. rans er o mar e s are rom a sma er o a arger rm s ou always increase concentration. There should be a market share threshold such that if a new firm enters the industry with a market share below the threshold, concentration is reduced. Similarly, if an incumbent firm with a market share below the threshold exits from the industry, concentration is increased. Any merger between two incumbent firms should always increase concentration.
•
As will be shown below, not all of the seller concentration measures that are in use satisfy all of the Hannah and Kay criteria. This section examines the construction and interpretation of the most common measures of seller . concentration ratio, the Herfindahl–Hirschman index, the Hannah-Kay index, the entropy coefficient, the variance of the logarithms of firm sizes, and the Gini coefficient.
Concentration Ratio
Herfindahl–Hirschman (HH) index
Hannah and Kay index Hannah and Kay (1977)suggest the following generalization of the HH index:
where α is a parameter to be selected. α should be greater than zero, but not equal to one, because HK(1) = 1 for any firm size distribution
The last two points are illustrated in Table 6.8. For I1 the contribution to the HK(1.5) index of firm 1 (the largest firm) is 45.8 per cent (0.2003 out of 0.4376). But firm 1’s contribution to the HK(2.5) index is 63.8 per cent (0.0686 out of 0.1076). Our earlier comments about the favourable properties of the HH index apply in equal measure to the HK(α) index. Furthermore, the larger the value of α, the smaller the degree of inaccuracy if the HK(α) index is calculated using accurate individual data for the largest firms, but estimated data for the smaller firms.
Entropy coefficient
E is an inverse concentration measure: E is small for a highly concentrated industry, and E is large for an industry with low concentration In Tables 6.5 and 6.6, E is 1.3721 (for I4), 1.7855 (I1) and 1.8236 (I5). Therefore E correctly identifies I4 as more concentrated than I1, and I1 as more concentrated than I5 The minimum possible value is E = 0, for an industry comprising a single monopoly producer. The maximum possible value is E = loge(N), for an industry comprising N equal-sized firms
Standardized entropy coefficient, whose maximum value does not depend on the number of firms (RE)
The minimum possible value is RE = 0 for a monopoly, and the maximum possible value is RE = 1 for an industry comprising N equal-sized firms.
Variance of the logarithms of firm sizes •
•
•
In statistics, a variance provides a standard measure of dispersion or inequality within any data set. In the case of data on the sizes of firms in an industry, the statistical property of dispersion or inequality is closely related to (but not identical to) the economic property of seller concentration. In Table 6.5, dispersion in I2 is zero (because all firm sizes are the same), but dispersion in I4 is much higher (due to the inequality in . , than in I2. Accordingly, the variance of the logarithms of firm sizes, VL, can be included among the list of concentration measures (Aitchison and Brown, 1966). VL is defined as follows:
For the purposes of calculating VL, the firm size data are expressed in logarithmic form for the following reasons: Most industries have a highly skewed firm size distribution, with large numbers of small firms, fewer medium-sized firms and very few large firms. The variance of the (untransformed) firm size data would therefore tend to be unduly influenced by the data for the largest firms. The log-transformation reduces or eliminates the skewness in the original distribution, enabling VL to provide a more reasonable measure of inequality across the entire firm size distribution. •
The variance of the (untransformed) firm size data would be influenced by the scaling or units of measurement of the data. VL, in contrast, is unaffected by scaling. For example, if inflation caused the reported sales data of all firms to increase by 10 per cent, the variance of the (untransformed) sales data would increase, but VL would be unaffected. In this case, there is no change in concentration or dispersion because the sales of all firms are increased in the same proportions. VL reflects this situation accurately. •
Although VL has occasionally been used as a measure of seller concentration, it is more accurate to interpret VL as a measure of dispersion or inequality in the firm size distribution. The distinction can be illustrated using the following examples taken from Tables 6.5 and 6.6: Both I2 and I3 have VL = 0 because in both cases all firms are equal-sized, so there is no inequality. From an industrial organization perspective, however, it seems clear that I3 is more highly concentrated than I2. I3 has fewer firms than I2, making it more likely that a cooperative or collusive outcome will be achieved. •
An economist would regard I6 as more concentrated than I1. However, the merger between the three smallest firms in I1 to form I6 implies I6 has a lower degree of inequality in its firm size distribution than I1. Accordingly VL is smaller for I6 than for I1. When we switch from I1 to I6, VL moves in the opposite direction to HH and HK(α), and in the wrong direction from the economist’s perspective. •
Lorenz curve and the Gini coefficient •
A Lorenz curve (named after Lorenz, 1905) shows the variation in the cumulative size of the n largest firms in an industry, as n varies from 1 to N (where N is the total number of firms)
•
The firms are represented in a horizontal array, from the largest to the smallest (reading from left to right) along the horizontal axis. The vertical axis shows the cumulative size (the sum of the sizes of all firms from firm 1 to firm n, as a function of n) –
If all of the firms are equal-sized, the Lorenz curve is the 45-degree line OCA. At point C, for example, exactly half of the industry’s member firms account for exactly half of the total industry size, represented by the distance OD.
–
the concave curve OBA. At point B, exactly half of the industry’s member firms account for three-quarters of the total industry size, represented by OD.
The Lorenz curve can be used to define a concentration measure due to Gini (1912), known as the Gini coefficient. With reference to Figure 6.1, the Gini coefficient is defined as follows:
The maximum possible value of G = 1 corresponds to the case of one dominant firm with a market share approaching one, and N − 1 very small firms each with a negligible market share. In this case the Lorenz curve approaches the line ODA, so the numerator and denominator in the formula for G are the same. . this case the Lorenz curve is the 45-degree line OCA, so the numerator in the formula for G is zero. •
•
The formula definition for the Gini coefficient is as follows: where xi is the size of firm i (as before, measured using sales, assets, employment or some other appropriate size indicator) when the firms ranked in descending order of size
Like the variance of logarithmic firm sizes measure, the Gini coefficient is most accurately interpreted as a measure of inequality in the firm size distribution. In fact, elsewhere in economics one of the best known applications of the Gini coefficient is for the measurement of inequality in household incomes. In our case, Tables 6.5 and 6.6 show that both I2 and I3 = , . As before, however, an industrial economist might regard I3 as more highly concentrated than I2. Furthermore, I6 has G = 0.3129, smaller than G = 0.4482 for I1. But an industrial economist would regard I6 as more concentrated than I1.
