Trade‐Adjusted Concentration Ratios in the US Manufacturing Sector

Abstract

This paper incorporates international trade into the four‐firm concentration ratio to get a more realistic measure of market structure in the US manufacturing sector using 1997 and 2002 NAICS data. As expected, trade‐adjusted CR₄ is significantly lower than the published CR₄. Moreover, the effect of international trade is higher in 2002 than in 1997, offsetting the increase in domestic concentration and leaving the US manufacturing sector in 2002 as competitive as it was in 1997. Furthermore, different tests are used to check the validity of the results. All of them confirm that trade‐adjusted CR₄s are significantly lower than their published counterparts.

Key Words:

• Concentration Ratio
• International Trade
• Manufacturing Sector
• NAICS

JEL classifications:

• L11
• L60

Notes

1. Note that this generalization about imports assumes that the large firms do not control any imports and label them as their own. In addition, the effect of imports on specific industry can be positive or negative depending on many factors. For example, for heterogeneous products, if imports are substitutes for the products produced by small domestic firms, then an increase in imports leads to an increase in the CRs. While if imports are substitutes for the products produced by large domestic firms, then an increase in imports leads to a decrease in the CRs. Pryor (1994) finds that the first effect dominates for the manufacturing sector in 1958 and 1982. Another factor is technology. In technology intensive industries, CRs may fall if older/large firms cannot keep up with the rapid changes in technology. On the other hand, CRs may increase because of the cost advantages older/large firms have over small firms. Specifically, since fixed costs in research and developments are high in technology intensive industries and large firms have more advantages here; this will lead to a higher concentration. Pryor (1994) finds that the latter effect dominates in the same period mentioned above.

2. HHI of 1,800 approximates the existence of five competitors in the market. (If a market has five competitors with equal market shares, the HHI will be 2,000, close to the DoJ guidelines.) As the number of competitors increases, HHI decreases below the DoJ benchmark (If the market has six competitors with equal market shares, HHI is 1,666.67). While as the number of competitors decreases, HHI increases well above the HHI benchmark. (At four competitors with equal market shares, HHI is 2,500 and at three competitors with equal market shares, HHI is 3,333.33.)

3. The most commonly used concentration measures are the concentration ratio and the Herfindahl–Hirschman Index (HHI of 1,800 approximates). The question of which index to use has been extensively debated in the literature. The common belief is that the choice of any concentration measure to describe an industry or to relate its structure to performance does not matter much as most concentration indices are highly correlated. Scherer and Ross (1990) mentioned that the correlation between the CR₄ and the HHI in 1982 was 0.954. The correlation increases even more to 0.992 when the logarithms of the HHI and the CR₄ are used. Other scholars (Bailey and Boyle, 1971, cited in Kowka, 1981, p. 446; Rosenbluth, 1955, cited in Kowka, 1981) have shown similar results of high correlation (greater than 0.90) among different concentration measures. Most empirical studies comparing different measures of concentration found that there is no one measure that is superior to others. Drucker (2009) mentioned that concentration ratios are the most widely used measure of concentration and that the HHI – because of its weight – is insensitive to small firms. Shepherd and Shepherd (2004) mentioned that in most cases, HHI is inferior to CR₄. Furthermore, the US Census Bureau publishes the CR for all sectors while publishes the HHI for the manufacturing sector only. Therefore, using the CR₄ enables comparison of concentration among all sectors, while using the HHI makes the comparison among sectors more difficult.

4. In terms of regional concentration, Drucker (2009) compared regional concentration in the US manufacturing sector over different years and found that regional concentration decreased from 1963 to 1997. It increased slightly from 1997 to 2002, with an average CR increasing from 0.3610 in 1997 to 0.3697 in 2002. In addition, Izraeli and Murphy (2003), cited in Drucker (2009), found that unemployment rates are lower and per capita disposable income is higher in the states with low industrial concentration in the United States.

5. Note that the usual formula used for trade correction assumes that the four largest firms either control a share of imports that is smaller than their share of domestic production or they don’t control any imports.

6. Note that the 2002 distribution is less dispersed than the 1997 one.

7. This assumption has been extensively used in the literature; for example, Lunn (1984), Cannon 1978 (cited in Utton, 1982), Green 1978 (cited in Utton, 1982), Utton (1982), Weiss and Pascoe (1986), and others. Utton (1982) finds that for the UK manufacturing industries, the exports share of the largest producers is greater, on average, than their sales’ share.

8. This causes the corrected CR₄ to overstate (understate) the actual CR₄ if the four largest firms’ share in exports is greater (smaller) than their share in domestic production.

9. The true measure of trade‐corrected CR₄ should have a slightly different numerator. The numerator should be (S ₄ – X ₄ + M ₄), where X ₄ and M ₄ are the share of exports and imports respectively, controlled by the four largest firms in the industry. Because of the unavailability of data for X ₄ and M ₄ and to be comparable with other studies in the literature, I used the formula CR_4T.

10. See the results in the comparison of CR₄ and CR_4T section below.

11. A detailed table of the published CR₄ and the trade‐corrected CR_4T for all industries in the US manufacturing sector is posted on the author’s homepage www.gbc.edu/~raouff.

12. Note that the CR₄ < 40%, 40% ≤ CR₄ ≤ 60%, and CR₄ > 60% are the traditional market structure demarcations.

13. These subsectors were also heavily affected by international trade in Pryor (1994) paper.

14. Other industries include 322122, 312140, 315224, 315993, 316211, 316214, 325192, 327111, 331111, 331112, 332115, 332913, 332998, 333991, 334112, 334416, 336415, and 336611.

15. Other industries include 316991, 321213, 321219, 322221, 325193, 327991, 333311, 333997, 334417, 334611, 335991, 336312, 336330, and 337920.

16. CR_4T will be less than CR₄ as long as M > 0. (Note that this is true if none of the four largest producers are foreign.) To see that

17. CR₄ and CR_4T in 2002 and 1997 are slightly positively skewed. In 2002, for CR₄, the mean is 45.34% while the median is 44.3%. For CR_4T, the mean is 33.51% while the median is 30.99%. The Pearson’s index (PI) of skewness is 0.1489 for CR₄, and 0.333 for CR_4T. In 1997, for CR₄, the mean is 42.38% while the median is 40.9%. For CR_4T, the mean is 33.27%, while the median is 31.38%. The Pearson’s index (PI) of skewness is 0.2132 for CR₄, and 0.2941 for CR_4T. Note that the PI of skewness shows that the distribution is not significantly skewed.

18. The 2002 calculated Z‐stat = (0.4534 − 0.3351)/{√[(0.209582)^↑2/375] + √[(0.227016)^↑2/375]} = 7.3938.

The 1997 calculated Z‐stat = (0.42375 − 0.3327)/{√[(0.20758)^↑2/380] + √[(0.1926)^↑2/380]}=6.2828.

19. This also implies that international trade has a greater impact on adjusting concentration above the mean than below the mean. This effect is greater in 2002 than in 1997 as the regression coefficient for 2002 is smaller than for 1997. This reinforces our earlier results that the effect of international trade is higher in 2002 than in 1997 and it increases with concentration.

20. The 2002 t‐stat The 1997 t‐stat=

21. Note that globalization has made the definition of imports blurry. Many companies have production plants in more than one country. For example, the automobile industry: some Toyota and Honda cars are assembled in the United States, while some GM cars are assembled in other countries. Another example is the brewing industry: a Japanese brand of beer, Kirin, is brewed in the United States and is, therefore, not considered an import, whereas, an Irish brand of beer, Guinness, is brewed in Canada and is considered an import. The same is true for other industries as well.

22. Utton (1982) used similar assumptions. Utton (1982) used two percentages only: 10% and 25%.

23. In 2002, the 10% imports share is significant at 10% level of significance only. In 1997, the 15% imports share is significant at 5% level of significance.

24.

Then

25. This contradicts Cowling’s comment. Cowling, cited in Utton (1982), mentioned that, including imports share by the four largest firms in the modification formula results in a corrected CR₄ that is similar to the published one. For this to hold, the imports share of the four largest firms has to be the same as the CR₄.

26. This is the same formula used by Pryor (1994) and many others.

27. Pryor (1994) uses higher level of aggregation. Pryor’s (1994) calculation of the weighted averages was done on a two‐digit SIC code, even though he used a four‐digit SIC code. This study uses a six‐digit NAICS code, which is equivalent to a four‐digit SIC code. Cannon (1978), cited in Utton (1982), mentions that when studying competition, the appropriate level of aggregation is the product or product group that is a seven‐ or six‐digit NAICS code. In addition, Pryor (1994) uses customs value imports (which excludes transportation cost, insurance, and tariffs) while this research uses C.I.F. (cost, insurance, and freight) imports value. Pryor (1994) acknowledges that ‘this introduces some statistical noise into the calculations’ but it does not make a big difference.

28. In addition, there is the problem of the market definition: Whether the market is narrowly or broadly defined. For more information on this shortcoming, see Abdel‐Raouf (2009).

29. Schwartzman and Bodoff (1971).

30. Local and regional industries are provided in the table posted on the author’s homepage http://www.gbc.edu/~raouff and are identified by L and R respectively.

31. See Table 3 above for comparison.

32. Most transactions with value of fifty million dollars or more (indexed for inflation) have to file premerger notifications (Shughart II, the Concise Encyclopedia of Economics).