The Transition of Market Structure in Russia: Economic Lessons and Implications for Competition
January 1999 Annette N. Brown✫! Western Michigan University SITE and CEPR J. David Brown" Stockholm Institute of Transition Economics CEPR
✫Brown
and Brown are not related. Department of Economics, Western Michigan University, 1201 Oliver St., Kalamazoo, MI 49008 USA; +1 616 387-5557 ph; +1 616 387-3999 fax;
[email protected]. " Stockholm Institute of Transition Economics, Stockholm School of Economics, Box 6501, S113 83 Stockholm SWEDEN; +46 8 736-9685 ph; +46 8 31 64 22 fax;
[email protected]. !
We would like to thank the participants of the Russian European Dialogue on Social Reform conference in Moscow, September 1998, and participants at the ACES session “Market Structure and Transition” at the ASSA conference in New York City, January 1999, for comments and suggestions. We also thank Russell Pittman for comments and suggestions. We are extremely grateful to John Earle, the MacArthur Foundation, and ACE-TACIS Grant No. T95-4115-R for help with the data. We thank the Ruben Rausings Foundation for Research on Entrepreneurship and Innovation and the Jan Wallander and Tom Hedelius Foundation for generous financial support.
Abstract: This paper has three main objectives. First, using a wide variety of indicators, we examine how the industrial structure of Russia is changing during transition. We analyze what the implications of these changes are for potential competition. We then investigate the economic processes that direct these changes and econometrically test several hypotheses, put forward by Sutton (1991) and others, concerning the determinants of market structure. We find that Russian industrial structure is indeed experiencing dramatic changes. The size distribution of firms is generally converging to that found in the United States. Manufacturing concentration is increasing on average, but these averages mask huge structural changes. Product concentration is decreasing. Considered together, the various structural changes suggest that the potential for competition is actually improving. Central planning factors still partly explain the observed levels of concentration. We find no systematic differences between Soviet Russian concentration and market equilibrium concentration. We find no evidence that Russian industry concentration is converging to U.S. industry concentration. The evidence strongly suggests that the economic processes directing increases in concentration are different from those directing decreases and that the processes determining market structure in exogenous sunk cost industries are different from those in endogenous sunk cost industries. The individual determinants of changes in concentration lend support to several other hypotheses theoretically developed by Sutton (1991).
I. Introduction The economic study of market structure has enjoyed a varied history. For many years, the Structure-Conduct-Performance (SCP) paradigm, which emphasizes the influence of market structure on economic outcomes, motivated much of the thinking and empirical research in industrial organization. With the advent of game theoretic modeling, the SCP paradigm was rejected in favor of specific industry models where many different factors may influence the conduct and performance firms. Some theories, such as the contestable markets theory of Baumol, Panzer, and Willig (1982), argued that market structure may have no effect at all on conduct and performance. Other new theories modeled market structure as an outcome of conduct and performance rather than the cause, a possibility well-recognized as the endogeneity problem in the earlier SCP empirical literature. These game theory models of the determinants of market structure, like the game theory models of the 1980’s generally, usually involved restrictive assumptions making them difficult if not impossible to test empirically using interindustry level data. In the late 1980’s, John Sutton, sometimes with co-authors, set out to develop a theory using game theoretic modeling that would yield empirical implications general enough to be tested across industries. His 1991 book, Sunk Costs and Market Structure, presents the theory in full in addition to extensive empirical analysis of the predictions. Sutton argues that the similarity between industry market structures across industrialized market economies suggests that market structure is a meaningful and revealing outcome of market processes. Therefore, the study of market structure and its determinants is important to economic theory. In this paper, we use the interesting example of Russia to further examine Sutton’s and others’ theories about market structure. The study of market structure in Russia, though, is not only useful for testing economic theories, it is also important for understanding transition and informing policy making, especially as market structure affects competition. Brown and Brown (1998) show that market structure does have a causal impact on competition, as measured by industry profitability, in Russia during transition. Specifically, industries that are oligopolistic and regionally dispersed enjoy higher profitability. It is especially important to understand the domestic conditions for competition in Russia because competition plays a larger role in determining allocative and productive efficiency than in other economies. First, most Russian firms need to be significantly restructured. Corporate governance, an important
1
determinant of firm conduct in developed market economies, is weak at best in Russia, leaving a larger role for competition to force change. Second, import penetration, especially in a country as large as Russia, is not pervasive enough to be the sole source for competition. This paper has three main objectives. First, using a wide variety of indicators, we examine how the industrial structure of Russia is changing during transition. We analyze what the implications of these changes are for potential competition. We then investigate the economic processes that direct these changes and econometrically test several hypotheses, put forward by Sutton (1991) and others, concerning the determinants of market structure. We find that Russian industrial structure is indeed experiencing dramatic changes. The size distribution of firms is generally converging to that found in the United States. Industries are becoming more heterogeneous in terms of firm size. Manufacturing concentration is increasing on average, but these averages mask huge structural changes. For example, while some industries are increasing in concentration, many others are decreasing. For another example, product concentration is decreasing. Considered together, the various structural changes suggest that the potential for competition is actually improving. These findings support several hypotheses presented in Brown, Ickes, and Ryterman (1994) and Joskow, Schmalensee, Tsukanova (1994). As of 1996, it is clear that market structure in Russia is still in the process of transition. Central planning factors still partly explain the observed levels of concentration. We find no systematic differences between Soviet Russian concentration and market equilibrium concentration. Although the Russian size distribution is converging to the U.S. size distribution, we find no evidence that Russian industry concentration is converging to U.S. industry concentration. The evidence strongly suggests that the economic processes directing increases in concentration are different from those directing decreases. The determinants of changes in concentration lend support to several hypotheses theoretically developed by Sutton (1991). Market structure in endogenous sunk cost industries, proxied by consumer goods industries, develops much differently than in exogenous sunk cost industries. In exogenous sunk cost industries where concentration is decreasing, we find that toughness of price competition, change in market size, product diversification within firms, and product differentiation within industries all significantly affect the degree, or speed, of concentration change.
2
The data used for this research come from the Russian Enterprise Registry Longitudinal Database, a panel database of all medium and large and many small Russian industrial enterprises from 1992 to 1996 with cross-sections from 1989 and 1991. We constructed this database from the Russian government statistical office’s (Goskomstat) yearly industrial census data. For a full description of the database construction, please see appendix A. The data include basic enterprise indicators on all enterprises, such as employment, output, capital, and location, and product data at the ten-digit level for most enterprises. The most common comparator in this paper is the United States. The U.S. is similar to Russia in that it covers a large territory, which can make a difference for market structure, and many comparative studies of market structure use the U.S. as a benchmark.
II. The Starting Point The conventional wisdom about the industrial structure of Soviet Russia was that enterprises were very large and industries were very concentrated, if not monopolized. Using the 1989 Russian industrial census, Brown, Ickes, and Ryterman (1994) (hereafter BIR) present extensive empirical evidence on the structure of Soviet Russian industry that disputes the conventional wisdom. They find that aggregate concentration and the incidence of very large firms is much lower in Soviet Russia than in industrialized market economies. The size distribution of firms in Soviet Russia is characterized by high percentages of manufacturing firms and employment in the medium and large categories (200-9,999 employees) and small percentages in the small and very large categories. This distribution contrasts with that in the United States, for example, where small firms and very large firms account for the majority of firms and employment in manufacturing. The correlation of concentration at the industry level between the U.S. and Soviet Russia is very low.1 On average, industries in Soviet Russia are more concentrated than U.S. industries, but previous studies of the Soviet economy failed to recognize that the concentrated industries actually account for a very small share of employment and output. So, for example, while 45.4 percent of industries in Russia in 1989 have four-firm
1
Brown, Ickes, and Ryterman (1994) use a dataset constructed by PlanEcon, Inc. using the 1989 Soviet Census of Industry. Using product data and Soviet industry codes, PlanEcon assigned each Soviet enterprise a United States Standard Industrial Classification (SIC) code. Thus the comparisons between U.S. industry concentration and Soviet Russian industry concentration are direct comparisons according to U.S. industry definitions.
3
concentration ratios (CR4s) of 80 or higher, compared to 5.1 percent of U.S. industries, these industries account for only 14.6 percent of output, compared to 4.9 percent of value added in the U.S. Only 0.23 percent of employment is in true monopolies in 1989. Joskow, Schmalensee, and Tsukanova (1994) ) (hereafter JST) also present evidence, using the 1991 Soviet census of industry, that U.S. and Soviet Russian industry structures are very different from each other and that Soviet Russian industry is not as concentrated as conventionally believed. They point out that many of the Soviet Russian industries that are highly concentrated are also very narrowly defined. Thus high measured concentration is often an artifact of Soviet industry definition.2
III. Changes in Russian Industrial Structure In this section and the next, we present and analyze the changes in industrial structure in three categories. First we examine the aggregate changes, including the changes in the size distribution of firms and in aggregate industrial concentration. Then we look at indicators of industry-level change, such as industry concentration and industry heterogeneity. Finally, we analyze changes at the product level. The various disaggregations of product categories provide us with different, additional definitions of markets so that we are not confined to defining markets only using the government’s industry codes.
The Size Distribution of Firms The size distribution of firms in Russia is generally converging to that found in the United States.3 Figures 1 and 2 depict the dramatic change in the size distribution of firms in Russia from 1989 to 1996 compared to the distribution in the U.S. in 1987. The entry of many 2
The data used by Joskow, Schmalensee, and Tsukanova (1994) are categorized into industries using the Soviet industry codes (324 industries). Whereas in the U.S. groups of firms must have a minimum market size in order to be classified as an industry, the Soviet classification pays no attention to market size. Thus there are examples of narrowly defined industries, such as polymer transport containers and tea leaf processing which have only one enterprise, where measured concentration is very high. 3 The analysis in this section of the paper compares Russian enterprises with U.S. companies. There is debate about the correct units for comparisons between the two countries. Russian enterprises are geographically organized more like U.S. establishments. The enterprise unit in Russia, though, is the unit of ownership and corporate decision making it analogous to the company unit in the U.S. Further, as seen later in the paper, Russian enterprises, although geographically organized like U.S. establishments, produce many more products in many more product groups on average than U.S. establishments making Russian enterprises more like U.S. companies in terms of the organization of production.
4
small firms during transition has significantly increased the share of manufacturing firms and employment in the left tail of the distribution. The one category where Russia is not converging is the right tail—the incidence of very large firms is not increasing yet. In the U.S. and other OECD economies, the very large firms are usually conglomerates and multi-national firms; they have employment and output in many locations including abroad. Conglomerates are not reflected at the enterprise level yet in Russia. Groups of firms do interact in Russia in the form of Financial-Industrial Groups (FIGs), although the degree of interdependence varies widely by FIG. Future research will examine the effects of these groups on market structure.
Share of Number of Manufacturing Firms
Figure 1. The Distribution of Firms by Firm Size in Manufacturing Sectors: Russia Compared to the United States 100 1989
80
1991 60
1992
40
1993 1994
20
1995
0
1996 1-99
100-249
250-999
1000-9999
>9999
US 1987
Employment Size Category Sources: RERLD, Goskomstat yearbooks and the U.S. Census 1987 Enterprise Statistics (company data).
5
Figure 2. The Distribution of Firms by Employment Size in Manufacturing Sectors: Russia Compared to the United States
Share of Manufacturing Employment
60 1989 1991 1992 1993 1994 1995 1996 US 1987
50 40 30 20 10 0 1-99*
100-249
250-999
1000-9999
>9999
Employment Size Category Sources: RERLD, Goskomstat yearbooks and the U.S. Census 1987 Enterprise Statistics (company data). *The 1-99 category is an estimate. Our estimate of the number and employment of enterprises we are missing for industry as a whole is based on the difference between Goskomstat’s published totals and the totals in the RERLD. To get the number for manufacturing, we take the percentage of output in the RERLD in manufacturing and apply this percentage to the missing numbers for industry as a whole. Since the mean employment of missing enterprises in later years is less than 20, we add all missing enterprises and their employment to the 1-99 category in 1992-1996 (we do not add missing enterprises in 1989 or 1991, since the mean employment of missing enterprises in those years is 1,493 and 496, respectively). This is a less good assumption in 1992, where the mean employment of missing enterprises is 52). The Goskomstat published totals are based on the manufacturing census plus an estimate of the small enterprise sector based on small enterprise surveys.
Expansion, Contraction, Entry, and Exit The fairly steady changes in the size distribution of firms reflect fairly persistent changes in individual firm sizes as measured by employment. Tables 1 and 2 show that firm expansion and contraction are persistent during transition. Table 1 reports the Pearson correlations from year to year for firms categorized as growing (or steady) and contracting. Firms that grow from 1992 to 1993 are roughly 30 percent more likely to grow from 1993 to 1994 than firms that contract from 1992 to 1993. Table 2 reports the year-to-year conditional probabilities of expansion and contraction. For example, of the firms that grow from 1992 to 1993, 38 percent grow again from 1993 to 1994. Of the firms that contract from 1992 to 1993, 80 percent contract again from 1993 to 1994.
6
Table 1. Persistence of Growth GROW 93 GROW 94 GROW 93 1 GROW 94 0.295 1 GROW 95 0.184 0.323 GROW 96 0.168 0.185
GROW 95
1 0.279
GROW 96
1
Source: RERLD. The sample includes all enterprises for which we have employment in three successive years. We classify enterprises into two groups in each year: those with the same or higher employment than in the previous year and those with lower employment. We calculate Pearson correlations between enterprise group status across years for each enterprise.
Table 2. Persistence of Growth and Contraction Prob. of Prob. of Prob. of Growing in Contracting Exiting in Following in Following Following Year Year Year GROW93 0.379 0.577 0.028 CONTRACT93 0.142 0.801 0.040 GROW94 0.479 0.458 0.047 CONTRACT94 0.185 0.746 0.052 GROW95 0.400 0.454 0.074 CONTRACT95 0.126 0.525 0.113 Source: RERLD. The sample is the 19,060 manufacturing enterprises with positive employment in 1992 and 1993. The rows do not add to one because of missing employment values for enterprises that report other data for that year.
The expansion, contraction, and exit of firms during transition varies greatly by firm size. Table 3 gives the transition probabilities of these changes in firms from the beginning of transition until 1996. Of domestic incumbents (those firms that we see in 1992), firms of all sizes have roughly equal probabilities of expansion, but larger firms have much higher probabilities of contraction and smaller firms have much higher probabilities of exit.
7
Table 3. Transition Probabilities of Russian Firms as of 1996* by Employment Size Categories Firm Size
1-20
21-50
51-99
100249
250- 1000- >9999 Missing** 999 9999
Total
Percent of Domestic Incumbents that: Expand
15.8
13.8
15.8
15.6
14.7
12.2
12.3
0.0
14.8
Contract
25.4
24.1
29.3
51.5
70.6
76.1
70.4
0.0
54.2
Exit
45.2
36.5
33.6
17.9
6.5
4.3
8.9
33.3
17.2
Missing**
13.6
25.6
21.3
15.0
8.2
7.4
8.4
66.7
13.8
Number of firms
469
1,793
4,724
7,614
7,948 3,334
203
18 26,103
Percent of Domestic Entrants that: Expand
51.7
30.3
16.1
18.3
17.8
14.4
17.2
0.0
24.4
Contract
35.1
35.3
22.5
36.9
34.8
24.6
10.3
0.0
31.2
Exit
6.8
28.1
47.4
33.5
35.7
54.1
55.2
11.2
32.6
Missing**
6.4
6.4
14.0
11.3
11.7
6.9
17.2
88.8
11.8
1,315
1,493
2,178
2,130
983
305
29
169
8,602
Number of firms
Percent of Foreign Firms that: Expand
32.7
30.2
25.1
26.1
26.7
16.7
0.0
0.0
29.4
Contract
14.3
26.0
32.2
43.1
51.5
66.7 100.0
0.0
20.6
Exit
42.2
33.9
33.9
23.9
16.0
13.3
0.0
26.3
37.1
Missing**
10.8
9.8
8.8
6.9
5.8
3.3
0.0
73.7
12.8
3,487
906
410
364
206
60
2
262
5,697
Number of firms
*The first year is 1992 or the first year we observe employment. Firm size categories are based on initial employment. **This category comprises firms not reporting employment. Unless these are in the exit row, they did not exit. Source: RERLD.
The pattern is quite different for domestic entrants.4 Among these firms, expansion is more likely among small firms, contraction is roughly equally likely across size categories, and exit is more likely among large firms. The first two results are not too surprising; the third result is. Theoretical modeling and empirical evidence on firm selection in market economies (see Jovanovic 1982) find that small firms grow more rapidly and have a higher probability of failing than large firms. In Russia, though, the majority of entrants in 1993 and 1994 were different
4
Here the transition probability is calculated from the year of entry until 1996.
8
from regular market entrants.5 Table 4 shows that in 1993 and 1994, the majority of firms entered in the larger size categories—84 percent of our sample of entrants in 1993 and 83 percent of our sample of entrants in 1994 have 51 or more employees, compared to only 35 percent in 1995. Many if not most of the entrants that we observe in 1993 and 1994 are spin-offs from incumbent firms rather than de novo entrants. These spin-offs suffer the same restructuring needs as incumbents, and now that they are smaller firms, their inability to restructure is more likely to lead to exit than contraction. By 1995 we see that the size distribution of entry reflects a decrease in spin-offs and an increase in de novo entry, an interpretation supported by the similarity in distributions between domestic entry in 1995 and foreign entry in all years.
Table 4. Domestic and Foreign Entry by Employment Size Class 1-20 21-50 51-99 100-249 250-999 1000>9999 Missing Total 9999 Number of Domestic Firms that Enter in: 1993 35 127 282 312 232 133 12 20 1,153 1994 122 535 1,510 1,336 541 117 16 44 4,221 1995 1,158 831 386 482 210 55 1 105 3,228 1996 46 54 52 79 85 22 1 9 348 Number of Foreign Firms and Joint Ventures that Enter in: 1993 1,270 290 134 99 31 3 0 66 1,893 1994 1,001 237 84 59 22 3 0 104 1,510 1995 309 57 15 15 9 1 0 36 442 1996 78 9 3 0 2 0 0 8 100 Source: RERLD. 1996 entry statistics are biased downwards because they have not been corrected yet by the 1997 data.
Among foreign firms, small firms are more likely to expand, large firms are more likely to contract, and small firms are more likely to exit. One troubling finding is that the entry of foreign firms is decreasing significantly from 1993 to 1996 in all size categories.
Aggregate Concentration Aggregate concentration has increased slightly overall in Russia during transition. Figure 3 shows aggregate concentration in Russia measured at four levels of aggregation from 1989 to 5
It is important to keep in mind here that for the domestic firms we observe only a sample of firms with fewer than 200 workers. We have no reason to believe that there is sample selection bias that would affect the year to year comparison here, however. For foreign firms we have a population in all size categories.
9
1996, with comparisons to the United States for two levels. At all four levels of aggregation, aggregate concentration decreased from 1989 to 1993 suggesting that the country’s largest firms suffered larger output declines early in transition. Aggregate concentration increases after 1993 to levels slightly higher than 1989. It might seem surprising that aggregate concentration is increasing over this period while the share of employment in large enterprises is decreasing as shown in Figure 1. The results in Figure 3 reflect the fact that the overall decreases in the very top firms have not been as great as the decreases in the middle. Aggregate concentration in 1996 is still much lower than in the U.S.
Percentage of Manufacturing Employment
Figure 3. Aggregate Manufacturing Concentration by Employment: Russia Compared to the United States 35.0 1989 1992 1993 1994 1995 1996 US 1982
30.0 25.0 20.0 15.0 10.0 5.0 0.0 Largest 10 Firms
Largest 20 Firms
Largest 100 Firms
Largest 200 Firms
Sources: RERLD and Scherer and Ross (1990, p. 59).
Industry Concentration In the next few sections, we examine changes at the industry level, where industries are defined according to Russian government five-digit industry codes (OKONH). See Appendix B for a description of this classification. On average, industry concentration has increased during transition. Figure 4 illustrates this result, which is robust for concentration measured by one-firm through fifty-firm concentration ratios. The 1992 concentration curve stochastically dominates the 1996 concentration curve meaning that weighted average concentration across industries for each
10
concentration ratio is greater in 1996 than in 1992. The results are robust to measuring concentration using employment rather than output.
100 90 80 70 60 50 40 30 20 10 0
OUT96 OUT92 EMP96
49
45
41
37
33
29
25
21
17
13
9
5
EMP92
1
Share of Industry Employment/Output
Figure 4. Russian Manufacturing Industry Concentration
Number of Firms Source: RERLD. Concentration is weighted by industry output.
These averages mask the variance in changes, though. Of 180 well-defined industries, 127 experienced an increase in concentration, 51 experienced a decrease in concentration, and 2 stayed the same according to the CR4’s. Figure 5 shows that the increasing concentration industries are on average initially low-concentration industries, while the decreasing concentration industries are on average initially high-concentration industries. Table 5 provides further evidence that increasing concentration is not a concern for high-concentration industries. The industries with CR4’s of 80 or higher in 1992 experience a decrease in concentration of 6.9 points on average, while the other industries experience an increase of 4.6 on average. This decrease among high-concentration industries is predicted by BIR. When we restrict the sample of industries to exclude those that are very small or are not well-defined markets, we find that the high-concentration industries are still decreasing on average.
11
CR4
Figure 5. Manufacturing Industry Concentration Change 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
AVE RISING FALLING
1992
1993
1994
1995
1996
Year Source: RERLD. Rising and falling refer to industries for which the CR4 rose and fell, respectively, between 1992 and 1996.
Table 5. Change in Concentration in High-Concentration Industries HighOther HighOther Concentration Industries Concentration Industries Industries Industries Industry Set Including 180-Industry Set Small Industries Average Employment 1992 7,546 69,899 16,552 81,503 Average Employment 1996 8,104 45,419 17,350 52,836 Share of Output 1992 7.5 92.5 6.2 93.8 Share of Output 1996 6.8 93.2 6.0 94.0 Average CR4 1992 95.0 45.5 92.9 43.3 Average CR4 1996 89.1 50.1 91.0 48.1 Number of Industries 68 188 22 158 Source: RERLD. High-concentration industries are those with a CR4 in 1992 of 80 percent or more. The industry set including small industries contains 256 industries, all manufacturing industries except miscellaneous ones.
The incidence of dominant firms has increased slightly during transition. Russian competition policy defines a dominant firm as one that has 35 percent more of market share. The Russian Anti-Monopoly Committee identifies dominant firms according to markets it defines on a case-by-case basis. For this analysis we define markets as we have been so far in the paper— five digit industries at the national level. Table 6 shows that the number of industries with at least one dominant firm increased from 83 to 87 from 1992 to 1996 and from 32 to 38 for the
12
well-defined industries. The number of industries with just one dominant firm decreased from 78 to 77 over all industries and increased from 31 to 34 in the well-defined industries. In the well-defined industries in 1996, 10 of the 38 industries had a second ranked firm with market share within 20 percentage points of the dominant firm.
Table 6. Industries with Dominant Firms
Number of dominant firms Number of industries Average output share of dominant firms Number of industries where second-ranked firm’s share 0.40 below leader Has no competitors
1992 1996 1992 1996 All Manufacturing 180-Industry Set Industries 88 97 33 42 83 87 32 38 0.592 0.554 0.481 0.498 18 24 8 10 36
35
19
18
18
25
5
10
11
3
0
0
Source: RERLD. Dominant firms have a national market share of at least 35 percent.
Market Share Mobility and Industry Heterogeneity Russian industries are undergoing significant structural changes during transition. One important measure of industry changes is change in market leadership of firms. Concentration statistics can hide changes within industries; for example, a four-firm concentration ratio could remain the same even if the identity of the top four firms changes. Changes in market leadership may indicate increased competitiveness even when concentration statistics suggest no change (Curry and George 1983). Table 7 shows that 164 out of 180 industries had at least one firm in the top four change from 1992 to 1996.
13
Table 7. Turnover in Market Leadership Number of firms in top 4 in 1992 still in top 4 in 1996 0 1 2 3 4 Number of 11 40 59 54 16 industries CR4 92 0.361 0.380 0.441 0.600 0.707 CR4 96 0.343 0.424 0.484 0.651 0.727 Share of 0.072 0.146 0.336 0.260 0.128 manufacturing Average 0.0067 0.0038 0.0059 0.0050 0.0083 industry share* Average number 320 136 119 62 26 of firms *This is the average of each industry’s share of total manufacturing output. Source: RERLD.
Table 8 provides further evidence that market shares are fluctuating significantly. The average rank correlation of market shares within industries in Russia between 1992 and 1996 is 0.431; the average Pearson correlation is 0.625. This latter correlation is much lower than that for the U.S. between 1947 and 1954.6 During the post-war period, the U.S. also exhibited a high correlation between market share stability and concentration levels. Russia during transition exhibits no such correlation, suggesting that highly concentrated industries are just as likely to experience changing market shares as less concentrated industries. The Gort Stability Coefficient is another measure of how concentration is changing—values greater than one mean that large firms are gaining share at the expense of small firms and less than one mean that small firms are gaining share at the expense of large firms. Table 8 shows that, although the Gort statistic is greater than one for Russia, it is within a standard deviation of one, and it is not significantly greater than the Gort statistic for the U.S. from 1947-1954. Gort argues that the increases in concentration in the U.S. between 1947 and 1954 were not a great concern for competition.
6
We are confined to using these years of data for the U.S. because the only available data come from Gort (1963).
14
Table 8. Correlation in Output Shares Average Std. Dev. Average Std. Dev. Average Std. Dev. Pearson Spearman Gort Correlation (Rank) Stability Correlation Coefficient Russia 1992-1996 0.625 0.271 0.431 0.289 1.188 0.378 U.S. 1947-1954 0.824 0.236 1.052 0.490 Sources: RERLD and Gort (1963). We calculated the U.S. numbers using the data in Appendix Table A in Gort (1963), which is for 205 4-digit U.S. SIC industries. The Gort stability coefficient is what Gort calls “Coefficient B”, where the coefficients for each industry are the geometric mean of the coefficient on the earlier year output share in a regression of later year output share on earlier year output share and the reciprocal of the coefficient of the later year output share in a regression of earlier year output share on later year output share. A Gort Stability Coefficient greater than one suggests that larger firms are gaining market share at the expense of smaller firms, and a coefficient less than one suggests the opposite. Following what Gort did with the U.S. data, we assign firms having data in one year but not in the other the minimum output share for the industry minus 0.0001. This assumes that the reason for the omission is that firms existed but did not report.
These fluctuations in market share are causing industries to become more heterogeneous internally and across the economy. Table 9 presents evidence on the changing variation in firm size by employment. In the first column we see that on average the coefficient of variation in firm size within industries is increasing. This observed change is biased downward since we have only a sample of the small firms that enter between 1992 and 1996, which would increase the coefficient of variation in the later years. Thus, industries are becoming more heterogeneous internally. In addition, industries are becoming more heterogeneous relative to each other as shown in the second column. The variation in industries’ average firm size increases from 1.15 in 1992 to 1.55 in 1996. For comparison, the coefficient of variation of average firm size across industries is 1.97 in the United Kingdom in 1989 (Fingleton, Fox, Neven, and Seabright 1995). This comparison is rough because industries are defined differently in the two countries.
Table 9. Change in Size Heterogeneity Within and Across Industries Average coefficient of Coefficient of variation of variation of employment average employment within industries across industries Russia 1992 1.20 1.15 Russia 1996 1.37 1.55 United Kingdom 1989 1.97 Sources: RERLD and Fingleton et. al (1995).
15
Russian Industry Concentration Compared to the United States and a Case Study Where it is possible to compare across countries, the structures of Russian industries continue to be very different from those found in the United States. Table 10 presents a partial comparison across countries, and appendix table C1 presents a comprehensive comparison and also shows the changes in concentration and number of firms for the selected Russian industries. For these tables we carefully matched Russian industries as defined by Russian industry codes to U.S. SIC industries using the detailed descriptions of both industry classifications. Appendix C describes fully the matching process. These matches are roughly similar to those found in the appendix to JST.
16
Table 10. Comparison between Industry Concentration in the United States and Russia United States USSIC U.S. Industry name
U.S. 1987 # of CR4 firms 1,948 34 101 87 469 37 48 53 196 73 380 16
2051 2082 2084 2085 2098 2371
Bread, cakes, related Malt Beverages Wines, brandy Distilled, blended spirits Macaroni and spaghetti Fur goods
2411 2421 2500 2700 2822 2844 2851 2861 2874
Logging 11,852 Sawmills, planing mills 5,252 Furniture and fixtures group 10,775 Printing and publishing 57,376 Synthetic rubber 58 Toilet preparations 648 Paints, allied products 1,121 Gum, wood chemicals 52 Phosphate fertilizers 55
18 15 10 7 50 32 27 59 48
3011 3221 3241 3312
Tires, inner tubes Glass containers Cement Blast furnaces, steel mills
69 78 28 44
114 35 123 271
3317 Steel pipe, tubes 3441 Fabricated structural metal
155 2,334
3511 Turbines, turbine generator sets 3531 Construction machinery
68
3534 Elevators and moving stairways 3541 Machine tools, metal cutting 3554 Paper industry machinery 3556 Food products machinery
Russia OKONH Russian Industry name 18113 18144 18143 18142 18115 17361 +17362 15100 15210 15271 19400 13310 18131 13150 15400 13112
13351 16513 16112 12130+ 12160 23 12140 11 14820
Bread baking Beer Wine Liquor, vodka Macaroni Fur and fur products and by customer order Logging Sawmill production Furniture industry Printing industry Synthetic rubber Perfume and cosmetics Varnish and paint Wood chemicals Phosphate fertilizers, inorganic chemicals Tires Glass packages Cement Ferrous metal products and coking Steel pipes Containers and prefabricated (metal) buildings and housing
80 14111 Turbine building
872
48 14511+ Road constuction, excavating, 14512+ construction machines and 14520 construction materials equipment
158
52 14154 Elevator production
Russia 1992 Russia 1996* # of CR4 # of CR4 firms firms 1746 3 1612 4 252 16 221 21 126 19 149 33 109 19 142 18 52 22 60 24 71 57 97 59 2278 362 631 868 15 35 85 15 18
2 16 9 23 68 48 59 61 52
1590 468 623 1493 15 45 78 15 18
5 25 12 23 85 93 32 64 74
12 40 51 74
63 40 23 47
12 40 49 92
67 40 22 53
22 10
69 77
19 13
74 69
19
64
25
75
133
28
129
26
9
96
9
91
381 256 483
31 14210 Metal cutting tools 88 26 91 30 14185 Pulp, paper equipment 10 89 10 28 14620 Food and animal feed 101 29 93 equipment 3561 Pumps, pumping equipment 333 19 14194 Pumps 23 52 29 3562 Ball and roller bearings 113 58 14350 Ball bearings 23 48 28 3585 Refrigeration, heating 746 31 14187 Refrigeration machine 9 99 12 equipment building *1996 data are less complete because they haven't been corrected by 1997; the number of firms is biased downwards.
Casual analysis of Table 10 suggests that there is no pattern of change in Russia that can be explained by U.S. structure. Although there are some industries, such as furniture and paints and
17
41 96 34 68 59 81
allied products, where concentration moves toward U.S. levels, there are many other industries, such as synthetic rubber and macaroni and spaghetti, where concentration levels move further away from U.S. levels. Table 11 shows that the Pearson correlation between the U.S. 1987 CR4’s and Russian CR4’s for all 71 matched industries is 0.142 in 1992 and 0.125 in 1996.
Table 11. Correlation Between U.S. and Russian Concentration During Transition USHHI USCR4 Pearson Spearman Pearson Spearman Correlation (Rank) Correlation (Rank) Correlation Correlation CON92* 0.241 0.172 0.142 0.118 (0.044) (0.154) (0.238) (0.327) CON93* 0.217 0.187 0.177 0.157 (0.072) (0.120) (0.140) (0.192) CON94* 0.172 0.170 0.119 0.094 (0.152) (0.157) (0.319) (0.435) CON95* 0.098 0.172 0.115 0.111 (0.415) (0.149) (0.332) (0.351) CON96* 0.152 0.168 0.125 0.120 (0.210) (0.164) (0.298) (0.320) N 71 71 71 71 Sources: RERLD and U.S. Census 1987 enterprise statistics. *CON is Russian HHI in the correlations with the USHHI and Russian CR4 in the correlations with the USCR4. The P-values are in parentheses.
A detailed case study can give an idea about how industries are changing. The four-firm concentration ratio for the Russian perfume and cosmetics industry (comparable to the U.S. toilet preparations industry) increased from 48 in 1992 to 94 in 1996. The U.S. concentration was 32 in 1987. The increase in concentration in Russia occurred despite an increase in the number of firms from 24 to 45. Real industry output fell by 75 percent, and the average import penetration ratio was 32 percent.7 The driving force of the change in concentration was one firm whose share increased from 19 percent in 1992 to 86 percent in 1996, while its real output increased by 269%. The number of 4-digit products it produced increased from 3 in 1993 to 4 in 1996; the
7
The measure of import penetration was calculated using data from the Russian Customs Committee on traded goods classified by United Nations product codes. These codes were assigned, as best possible, to Russian OKONH industries. Combining the RERLD data on output and the Customs data on traded goods, we calculated imports/(imports+output-exports). The measure is imprecise, not only because the industry assignment is imperfect, but also because the trade data include all sales, while the output data are biased by firms missing from the database.
18
number of 6-digit products increased from 5 in 1993 to 13 in 1996. At the same time, employment in this firm fell from 1762 to 1540. All other incumbent firms experienced a decline in market share from 1992 to 1996, although the year-to-year shares fluctuated quite a bit. Only two of the top four firms in 1992 remained in the top four in 1996. One firm entered the market from the soap and cleaning products industry in 1995, accounting for 10 percent of the perfume and cosmetics market in 1995, but only 3 percent in 1996. There were 21 entrants, although several of these firms report zero to negligible output in all reported years. By 1996, the largest market share for an entrant was 0.066 percent, a decline from its entering share in 1994 of 0.63 percent. Two firms exited the market and entered another—one the soap and cleaning products industry and one a miscellaneous food processing classification. Five firms exited completely, four incumbents and one entrant. The highest market share an exiter had in its last year was 0.2 percent. Three out of the five experienced a significant decline in employment before exit, with the average employment in the last year being 64 employees. The other two held employment steady before exit at roughly 100 and 25. The perfume and cosmetics industry represents just one example of how structures are changing. In this case, the industry is clearly becoming more concentrated domestically and facing its major competition from imports.
IV. Implications for Competition The most basic indicators, aggregate concentration and average industry concentration, suggest that market structure is becoming more concentrated in Russia, which would suggest that conditions for competition are eroding. Examined more carefully, however, the changes in market structure instead reflect an increasing potential for competition. There is dramatic entry of new, small firms, many more than we have in our data. Exit is a real threat. Firm sizes and firms’ market shares are changing noticeably, so that industries are restructuring much more than concentration indices reveal. There is quite high turnover in market leadership. The high concentration industries are experiencing a decrease in concentration on average and the decreasing concentration industries were more highly concentrated on average in the beginning. On the other hand, those industries experiencing increases in concentration tend to start at low
19
concentration levels. On average these industries start at a CR4 of 45 and increase to 53 by 1996. A CR4 less than 60 is not usually a concern to policy makers in market economies.
Product-Level Concentration An examination of changes in market structure at the product level provides further evidence that the environment for competition in Russia is improving. It is also instructive to examine changes at the product level because during the Soviet period most market structure analysis was conducted at the product level. Our database has product data at the 10-digit OKP level from 1993 to 1997 for a roughly random sample of about a third of the firms. See Appendix D for a full description of the product data. These 10-digit codes allow us to aggregate at the 6-digit, 4-digit, 3-digit, and 2-digit levels in order to analyze narrower (6-digit) and broader (4-digit) product groups and aggregations of product groups that could be considered industry (3digit) and sectoral (2-digit) groupings. Unlike the U.S. and other countries’ classifications, though, all these groupings are separate from the 5-digit industry code (OKONH) and no official mapping exists between OKP and OKONH. Table 12 reports the number of categories at each level in 1993 and 1997 compared to the number of categories at various levels in the U.S. SIC scheme. The Russian 4-digit OKP disaggregation is roughly comparable to the U.S. 5-digit SIC. The Russian 10-digit OKP disaggregation is more aggregated than the U.S. 7-digit SIC
Table 12. Comparison of Russian and U.S. Product Codes U.S. SIC No. produced in Russian OKP No. produced in 1982 1993 3-Digit SIC 144 2-Digit OKP 73 4-Digit SIC 452 3-Digit OKP 407 5-Digit SIC 1,600 4-Digit OKP 1,326 6-Digit OKP 3,794 7-Digit SIC >13,000 10-Digit OKP 3,932 6-Digit KOD_NOM 4,725
No. produced in 1997 74 429 1,422 3,860 4,458 4,859
Sources: RERLD and Streitwieser (1992).
The incidence of product monopolies and oligopolies at the 6-digit level in Russia is decreasing during transition. The left-most points in Figure 6 show that the percentage of products in monopolies decreased from 18.1 percent in 1993 (or 686 out of 3794 products) to 16.0 percent in 1997. These statistics are all biased upwards due to the fact that we have product 20
information for only a sample of the total firms, so initial concentration measured according to number of firms is lower than reported here. In addition, later year product concentration is biased upward even more because we are missing the data on the thousands of new small firms, some of which may be producing brand new products, but many of which are likely producing the same products as incumbents. In sum, product concentration is lower and decreasing more than figure 6 depicts. Figure 6 also shows that the outcome of decreasing product concentration is cumulatively robust from one-firm products to fifty-firm products. It is clear that Russian products are being produced by more firms.
100 90 80 70 60 50 40 30 20 10 0 49
45
41
37
33
29
25
21
17
13
9
5
1993 1997
1
Share of Products
Figure 6. Number of Firms Per 6-Digit OKP Product
Number of Firms Source: RERLD. These are cumulative distributions.
Product concentration is also decreasing according to concentration ratios calculated on physical output data. We do not have data in the RERLD on the value of output for each product for each firm, but we do have data on the physical volume of output for each product for each firm. This output variable is reported by KOD_NOM product categories, of which there are a few over 4,700. Again, please see Appendix D for further explanation. Figure 7 shows the concentration curves calculated by physical volume averaged over all products. So, for example, the average CR1 (the average over products of each product’s top firm’s share of total physical
21
output) decreased from 57.2 in 1993 to 56.2 in 1997. The result of decreasing concentration is robust for the CR1 through the CR15.
Share of Physical Volume of Production
Figure 7. Manufacturing Product Concentration 100 90 80 70 60 50 40 30 20 10 0
1993 1997
1
2
3
4
5
6
7
8
9 10 11 12 13 14 15
Number of Firms Source: RERLD. We use 6-digit KOD_NOM product codes here. Russian manufacturing enterprises produced 4,725 and 4,859 6-digit KOD_NOM products in 1993 and 1997, respectively.
The decrease in product concentration results in large part from an average increase in the number of products each firm is producing. Table 13 reports the changes in firms’ product behavior from 1993 to 1997 for the surviving incumbents in our product sample. These firms are increasing the number of final products they produce at each product level. The differences in the means are all statistically significant. The increase in products at the more aggregated levels, for example 3-digit OKP, suggests that firms are adding product lines, or diversifying. The increase in products at more disaggregated levels, for example 6-digit OKP, suggests that firms are differentiating their production. The number of final products per product line is also increasing, providing more evidence that firms are differentiating in addition to diversifying. The table also shows that firms changed their product mixes not just by adding products but also by dropping products. While 74 percent of firms added an average of 3.74 4-digit products, 58 percent of firms dropped an average of 2.94 products.
22
Table 13. Change in Number of Final Products Per Firm by Product Categories between 1993 and 1997 Product Avg. # Avg. # Percent Avg. # Percent Avg. # Percent Avg. # Categories in 1993 in 1997 Adding Added* Dropping Dropped* Keeping Kept at least 1 the the same Same* 2-Digit 2.26 2.51 48.4 1.46 30.3 1.52 97.4 1.85 OKP (18.20) 3-Digit 3.53 3.97 62.7 2.05 42.8 1.99 97.0 2.76 OKP (20.13) 4-Digit 5.98 7.07 74.4 3.74 57.7 2.94 96.7 4.43 OKP (26.29) 6-Digit 9.26 13.06 83.5 8.06 67.3 4.35 96.4 6.57 OKP (37.36) 3-Digit 1.69 1.71 OKP/2Digit OKP (2.79) 4-Digit 1.75 1.81 OKP/3Digit OKP (7.93) 6-Digit 1.55 1.85 OKP/4Digit OKP (30.18) *The mean is taken among enterprises conducting the action. Source: RERLD. The panel includes 9,010 manufacturing enterprises with at least one product and either employment or output in the RERLD in both 1993 and 1997. The t statistics for the test of the equality of means across years are in parentheses. Russian manufacturing enterprises produced 3,794 and 3,860 6-digit OKP products, 1,326 and 1,422 4-digit OKP products, 407 and 429 3-digit OKP products, and 73 and 74 2-digit OKP products in 1993 and 1997, respectively.
It is important to remember that diversification and differentiation are occurring in an environment where the majority of firms are cutting employment, so these changes in product mixes represent restructuring rather than just firm growth or acquisitions. In other words, diversification results from restructuring, not conglomeration. Figure 8 shows that at the same
23
time that firms increased the diversity of their production, they decreased the average number of employees per product.
Number of Employees
Figure 8. Average Number of Employees Per Product 160 140 120
1993 1994 1995 1996
100 80 60 40 20 0 Domestic Incumbents
Domestic Entrants
FDI
All Firms
Source: RERLD. The products are at the 10-digit OKP level. Russian manufacturing enterprises produced 3,932 and 4,458 10-digit OKP products in 1993 and 1997, respectively. FDI stands for foreign firms and joint ventures. Domestic incumbents are domestically owned enterprises with employment or output in 1992 or with products in 1993, and domestic entrants include all other domestically owned enterprises with products in 1994, 1995, or 1996.
In sum, when we define markets according to product categories and let firms compete in more than one market, we find more evidence that the competitive environment is improving in Russia during transition.
Russian Product-Related Behavior Compared to the United States There is evidence that Russian firms produce more products on average than U.S. establishments. Very little analysis has been done of U.S. market structure at the product level. We present here the little information that is available in order to provide some point of comparison for the Russian findings. The comparison is necessarily very rough. The U.S. product-level statistics that we do find are reported for the population of establishments, whereas in our data we are missing thousands and thousands of small Russian firms, which likely produce very few products. Thus our calculated incidence of single-product firms is very much biased downwards. In addition, the firm units differ across countries. For the U.S., Streitwieser (1992) 24
uses establishment-level data, where establishment is defined as each part of a firm that has a different physical location. All production at each location is considered one establishment. Our firm-level data for Russia are based on the Soviet distinction of enterprise. Enterprises are similar in definition to establishments in that Soviet enterprises usually located all their production in one place geographically. These single-location enterprises, though, were often quite large and comprised several production units, more like U.S. companies. Thus, we should expect to find more products produced at the enterprise level in Russia compared to the U.S. establishment level. Table 14 reports the incidence of single-product and multi-product firms and establishments in both the Russia and the U.S.. By 1997, the average employment size of the single-product firms in our sample is 152, compared to 573 for the multi-product firms. If we make the heroic assumption that all the new, small firms not represented in the data are singleproduct firms (the mean size of firms missing from the RERLD in 1996 is 19 (see table A1)) and then exclude all single-product firms and establishments from consideration, we can better compare the Russian and U.S. data.
Table 14. Single-Product and Multiple-Product Enterprises in the Russian Sample Compared to U.S. Establishments 1993
No. of enterprises/ Establishments % of enterprises/ establishments % of output/shipment value Mean No. of Products Mean No. of Employees
1997
U.S. 1982 Single Multiple Product Products 145,056 61,710
Single Product 1,587
Multiple Products 10,793
Single Product 1,939
Multiple Products 14,102
12.8
87.2
12.1
87.9
70.2
29.8
10.7
89.3
2.6
97.4
34.2
65.8
1.0
7.1
1.0
8.4
1.0
3.1
403
858
152
573
50
157
Sources: RERLD and Streitwieser (1992). Russian manufacturing enterprises produced 1,326 and 1,422 4-digit OKP products in 1993 and 1997, respectively, and there were 1,600 U.S. 5-digit products in 1982.
Table 15 classifies as “multi-product” firms and establishments that produce more than one 4digit OKP (1,400 total) or 5-digit SIC (1,600 total) product. The table shows how diversified
25
these firms are across more aggregated product and industry categories. For example, in Russia in 1997 25.5 percent of the multi-product firms produce products in only one 2-digit OKP sector. In contrast, in the U.S. 78.0 percent of multi-product firms produce products in only one 2-digit branch and 53.3 percent of multi-product firms produce products in only one 3-digit SIC sector. For another example, in 1997 6.8 percent of Russian multi-product firms produced products in more than 10 different 3-digit OKP “industries” while no U.S. multi-product establishments operated in more than 10 4-digit SIC industries.
Table 15. Distribution of Multi-Product (4-digit OKP) Russian Firms Across Broader Product Classes Compared to Multi-Product (5-digit SIC) U.S. Establishments Number of 1993 1997 U.S. 1982 industries 2-Digit 3-Digit 2-Digit 3-Digit 2-Digit 3-Digit 4-Digit in which OKP OKP OKP OKP SIC SIC SIC enterprises operate 1 36.1 12.1 25.5 6.5 78.0 53.3 33.6 2 31.4 21.6 28.6 21.3 19.3 35.9 45.3 3 14.4 18.7 19.6 16.3 2.2 7.9 14.0 4 7.6 14.4 11.4 12.0 0.3 2.1 4.6 5 4.3 10.3 6.0 9.9 0.1 0.5 1.5 6-10 5.4 18.7 7.4 27.3 0.1 0.3 1.0 11+ 0.9 4.1 1.5 6.8 F = 0.3439. *Significant at 10% in a two-tailed test. **Significant at 5% in a two-tailed test. ***Significant at 1% in a two-tailed test.
Appendix table E-3 reports the estimates with the U.S. concentration variable included for the subset of industries where we can match between the countries. Consistent with the findings for 1989 in table 17, U.S. concentration levels have no explanatory power.
Market Determinants of Changes in Market Structure The process of structural change during transition allows us to test for the determinants of market structure. Observed levels of concentration during transition are only partly determined by market forces, so in order to isolate the impacts of these forces, we use change in concentration from 1992 to 1996 as our dependent variable. Many of the hypotheses that we test have been formulated in terms of the effects of level variables on levels of concentration. As Caves and Porter (1980) explain, estimating the effects of levels on changes is theoretically justified when the industries begin out of equilibrium, which is clearly the case here. We test two hypotheses about the overall processes of change. In a market economy, where changes in concentration represent evolutionary movements between static equilibria, one might assume that increases in concentration and decreases in concentration follow the same
38
processes. Many studies include these industries in the same regressions. Caves and Porter (1980) find, though, that the pattern of changes for increasing concentration industries are different from those for decreasing concentration industries in the United States between 1954 and 1972. In a case like Russia, however, where increases and decreases originate from out-ofequilibrium levels, there is no reason to assume that the processes are the same. We test whether the processes are different by running the regressions separately on increasing and decreasing concentration industries and using the Chow test. We do not include in the regressions industries where measured concentration changed less than five percent in either direction between 1992 and 1996. Sutton’s theory argues that the determination of market structure is different for endogenous sunk cost industries than it is for exogenous sunk cost industries. We define endogenous sunk cost industries as those where the consumer goods share of total output is greater than 50 percent. All other industries we define to be exogenous sunk cost industries. Again, we run the regressions separately for these two groups and use a Chow test to test whether the coefficients are the same. We run the following specification based primarily on Sutton’s hypotheses.11 (5)
%∆CONC = β1 + β2CONC92 + β3CONC*GEODIS92 + β4SETCOST92 + β5OUTGROWTH + β6FIRMPRODS93 + β7INDPRODS93 + β8PROFIT92 + β9PRIVSHARE94 (+ β10CONSUMER93) + ε.
The dependent variable, the percentage change in concentration, can be interpreted as the speed of change or the degree of change. Without knowing the ending point (the point where concentration is completely market determined), we cannot distinguish between the two. The 1992 level of concentration is a control variable. The variable for 1992 profits tests the general theory that profitable industries, controlling for concentration, attract more entry. We avoid the endogeneity problem by only including the value in the initial year of transition. The variable for private share in 1994 controls for any relationship between the privatized share of industries and change in market structure. Because privatization and structural change are occurring
11
The model that Sutton presents in his book is essentially a static model. In several cases, though, he shows specifically how the results are robust to a sequential entry game and he frequently discusses the predictions in dynamic terms. We thus generalize his hypotheses to a dynamic setting, and our empirical results suggest that this generalization is indeed appropriate.
39
simultaneously, we cannot test for causal effects. Privatization shares change very little after 1994, so we use the 1994 value as the control. Sutton hypothesizes that in exogenous sunk cost industries (non-consumer goods), weaker price competition leads to lower concentration. Russia provides a unique opportunity to test this hypothesis because we have an exogenous measure of the degree of competition. At the beginning of transition, the structure of industry—firm number, size, and location—is exogenous to market forces. Brown and Brown (1998) find that the interaction between the inherited concentration of industries and their inherited geographic dispersion affects industry profitability in later years. Specifically, industries which inherited structure is both concentrated and dispersed enjoy higher profitability in later years controlling for structural changes, suggesting that they are characterized by weaker price competition. We include this interaction term in these regressions to test for the impact of toughness of price competition on market structure. Sutton hypothesizes that increases in market size cause concentration levels to converge to zero in exogenous sunk cost industries, whereas the effects of market size disappear in endogenous sunk cost industries when market size gets large. To study the effects of scale economies and market size together, Sutton employs the variable σ/S, which is setup cost divided by market size. He defines setup cost as µK, where µ is minimum efficient scale and K is the industry capital requirement. That is, the setup cost index σ/S is the minimum efficient scale times the industry capital-output ratio. Our SETCOST variable is calculated the same as Sutton’s setup cost index with one difference. As Sutton recognizes, Davies (1980) criticizes the standard definition of MES (median plant size) on the grounds that its relationship to concentration may be tautological. The standard definition of MES is based on an assumed technological relationship between firm size and efficiency within industries. We cannot make that assumption for Russian industries since Soviet planners maximized many more, if not completely different, objectives than technical efficiency when deciding on firm size. In order to try to capture technical efficiency, we therefore define minimum efficient scale in Russia as the size of the median firm when firms are ranked according to their cost-output ratios. As noted in Brown and Brown (1998), we find no relationship between firm sizes and cost-output ratios. Thus, our measure is not subject to the Davies critique.
40
The specification includes both the level of setup cost and the change in industry output, or change in market size. There is little change in measured MES over the observed years and very little investment. In addition, there is extreme noise in the capital variable due to changing valuations over those years, so we assume that the numerator of setup cost does not change and include only the change in the denominator in the specification. The final two variables in the specification address Sutton’s hypotheses concerning product differentiation. Sutton makes reference to two types of product differentiation: the variety of products across an industry, and the variety of products within firms. Product differentiation in an industry should facilitate entry by providing product niches. In endogenous sunk cost industries, this effect may be overshadowed by the positive relationship between product differentiation and advertising effectiveness. We measure industry differentiation, IND PRODS, with the natural log of the number of 6-digit OKP products produced in the industry. In exogenous sunk cost industries, product differentiation within firms should lower the lower bound on concentration although it also introduces multiple equilibria, which may yield higher levels of concentration. We measure firm product differentiation, FIRM PRODS, as the average number of products produced by firms in the industry as a share of the total products in the industry. In table 20, we report the results from running the traditional specifications (consumer and non-consumer goods included together) for increasing and decreasing concentration industries separately. We provide these results for comparison to previous work and in order to test the hypothesis that the processes of increasing concentration and decreasing concentration are different. The Chow test clearly rejects the null hypothesis that the estimates are the same across these two models. Also in table 20 we pool increasing, stable, and decreasing industries but run separate regressions for consumer (endogenous sunk cost) and non-consumer goods (exogenous sunk cost) industries. The Chow test again clearly rejects the null hypothesis that the processes are the same.
41
Table 20. Determinants of Increasing and Decreasing Concentration as Separate Processes and Consumer and Non-consumer as Separate Processes Variable RISE HHI FALL HHI CONSUMER NONCONSUMER Intercept 2.238 -0.835* 3.619 1.019 (1.13) (-1.98) (0.86) (1.50) HHI92 0.691 1.009*** -0.368 -0.114 (0.64) (2.66) (-0.07) (-0.15) HHI92*GEODIS92 1.427 -12.360*** -6.200 -11.808** (0.23) (-4.05) (-0.21) (-2.47) SETCOST92 1.331 0.106 3.539 -0.207 (1.11) (0.91) (0.36) (-0.41) OUT GROWTH 1.521 -1.447*** 4.595 -0.018 (0.76) (-3.25) (1.15) (-0.03) FIRM PRODS93 -1.098 -0.609*** -2.802 0.389 (-1.11) (-2.83) (-0.98) (0.86) IND PRODS93 -0.008 -0.117*** -0.126 -0.018 (-0.12) (-3.01) (-0.42) (-0.29) PROFIT92 0.794 0.231 1.399 0.272 (1.06) (0.61) (0.60) 0.64 PRIVSHARE94 -0.675 0.001 2.414 -0.724*** (-0.83) (0.01) (1.53) (-2.73) CONSUMER93 0.681*** -0.158* (2.62) (-1.90) Adj. R-square 0.137 0.531 0.034 0.073 N 117 43 50 130 The t statistics are reported in parentheses. The Chow test for the null hypothesis that the coefficients are the same for industries with increasing and decreasing concentration yields F (10, 140) = 3.56, Prob > F =0.0003. The Chow test for the null hypothesis that the coefficients are the same for consumer goods and non-consumer goods industries yields F (9, 162) = 4.01, Prob > F =0.0001. *Significant at 10% in a two-tailed test. **Significant at 5% in a twotailed test. ***Significant at 1% in a two-tailed test.
We correctly specify the estimating equation by dividing industries into four groups based on increasing/decreasing and consumer/non-consumer and running the regressions for each group separately. We do not have enough observations of consumer goods industries with falling concentration to run regressions on this group. Table 21 reports the estimates. The Chow tests comparing these various subgroups confirm the findings above that the processes are distinct.
42
Table 21. Determinants of Degree of Concentration Change with Consumer and Non-Consumer Goods Variable Consumer NonNonIncreasing Consumer Consumer HHI Increasing Decreasing HHI HHI Intercept 4.325 1.925* -0.983*** (0.61) (1.76) (-2.62) HHI92 0.905 0.281 1.124*** (0.28) (0.33) (3.13) HHI92*GEODIS92 -1.089 -3.824 -10.973*** (-0.06) (-1.02) (-3.70) SETCOST92 4.688 0.637 0.095 (0.86) (0.63) (0.88) OUT GROWTH 5.112 0.300 -1.522*** (0.68) (0.69) (-3.64) FIRM PRODS93 -3.537 0.022 -0.623*** (-1.27) (0.06) (-2.90) IND PRODS93 -0.051 0.022 -0.121*** (-0.19) (0.51) (-3.23) PROFIT92 1.996 0.204 -0.021 (0.81) (0.63) (-0.06) PRIVSHARE94 1.656 -1.579* 0.098 (1.25) (-1.87) (0.83) Adj. R-square 0.163 0.235 0.506 N 41 76 38 These are OLS regressions. Increasing HHI industries are those with a greater than 5 percent increase in concentration between 1992 and 1996, and decreasing HHI industries with a greater than 5 percent decrease in concentration between 1992 and 1996. The t statistics are reported in parentheses. The Chow test for the null hypothesis that the coefficients are the same for consumer good and non-consumer good industries with increasing concentration yields F (9, 99) = 2.61, Prob > F = 0.0095. The Chow test for the null hypothesis that the coefficients are the same for non-consumer goods industries with increasing vs. decreasing concentration yields F (9, 96) = 9.45, Prob > F =0.0000. *Significant at 10% in a two-tailed test. **Significant at 5% in a two-tailed test. ***Significant at 1% in a two-tailed test.
Consistent with Sutton’s hypotheses that the effects of several variables disappear in endogenous sunk cost industries, we find no statistically significant estimates for the consumer goods industries. According to these data, the theory does not explain the changes observed in exogenous sunk cost industries where concentration is rising, but it works quite well for exogenous sunk cost industries where concentration is falling. Remember that the left-hand-side variable is always negative, so a negative coefficient means that the variable causes concentration
43
to fall more—industries to be more fragmented. Three hypotheses are strongly supported. The weaker is price competition the more concentration falls. Increases in market size are associated with greater decreases in concentration. Greater industry product differentiation leads to greater decreases in concentration. We also find that firm product differentiation leads to greater decreases in concentration, consistent with the prediction that product differentiation lowers the lower bound on concentration. Consistent with earlier studies, we find that higher initial concentration leads to less decrease in concentration, although we only find an impact of initial concentration on changes in the one industry group. The lack of results for the group of non-consumer goods industries with increasing concentration could be due to a couple of factors. First, it may simply be the case that these industries are transforming in a random way or according to another process. It could also be the case that we incorrectly classified some consumer goods industries as non-consumer goods industries. We base our definition on the share of consumer goods in 1993. Given the increased demand for consumer goods during transition, it is possible that some industries have become more consumer-goods oriented and thus more influenced by endogenous sunk costs since 1993. Classifying the industries using a lower cut-off in 1993 does not change the results, however.
VI.
Conclusion This paper has three main objectives. First, using a wide variety of indicators, we
examine how the industrial structure of Russia is changing during transition. We analyze what the implications of these changes are for potential competition. We then investigate the economic processes that direct these changes and econometrically test several hypotheses, put forward by Sutton (1991) and others, concerning the determinants of market structure. We find that Russian industrial structure is indeed experiencing dramatic changes. The size distribution of firms is generally converging to that found in the United States. Industries are becoming more heterogeneous in terms of firm size. Manufacturing concentration is increasing on average, but these averages mask huge structural changes. For example, while some industries are increasing in concentration, many others are decreasing. For another example, product concentration is decreasing. Considered together, the various structural changes suggest that the potential for
44
competition is actually improving. These findings support several hypotheses presented in Brown, Ickes, and Ryterman (1994) and Joskow, Schmalensee, Tsukanova (1994). As of 1996, it is clear that market structure in Russia is still in the process of transition. Central planning factors still partly explain the observed levels of concentration. We find no systematic differences between Soviet Russian concentration and market equilibrium concentration. Although the Russian size distribution is converging to the U.S. size distribution, we find no evidence that Russian industry concentration is converging to U.S. industry concentration. The evidence strongly suggests that the economic processes directing increases in concentration are different from those directing decreases. The determinants of changes in concentration lend support to several hypotheses theoretically developed by Sutton (1991). Market structure in endogenous sunk cost industries, proxied by consumer goods industries, develops much differently than in exogenous sunk cost industries. In exogenous sunk cost industries where concentration is decreasing, we find that toughness of price competition, change in market size, product diversification within firms, and product differentiation within industries all significantly affect the degree, or speed, of concentration change.
45
Appendix A. Description of the RERLD The Russian Enterprise Registry Longitudinal Database (RERLD) is a panel database of all medium and large (the cutoff between small and medium varies from 50 to 200 employees, depending on the industry) and many small Russian industrial enterprises from 1992 to 1996 with cross-sections from 1989 and 1991. We have data for 1997 as well, but they have not yet been fully linked to the panel. The 1997 product data are operational, however (see appendix D). All data originated from Goskomstat’s annual censuses of industry (registries). The 1989 data were obtained from PlanEcon, Inc., which assigned each Soviet enterprise in the 1989 Soviet Census of Industry a United States Standard Industrial Classification (SIC) code, using product data and Soviet industry codes as guides. The 1992-6 panel comes from the 1993-6 annual censuses. The 1992 data come from the previous year values in the 1993 census, so the 1992 data do not contain enterprises that exited between 1992 and 1993 and have fewer variables. We constructed a panel by matching enterprise identification codes (OKPOs) across years. We found that each subsequent year’s registry typically contained 3,000-4,000 OKPOs not in previous registries and a similar number of OKPOs that dropped out of each subsequent registry. Some of this is due to enterprise entry and exit, some due to non-reporting enterprises, and some to enterprises that received different OKPOs. For all these OKPOs not having data in every registry, we searched in all the other registries for matching enterprises by using names, addresses, industries, and employment values. We found 917 enterprises whose 1993 OKPOs were different from their latest one, 522 whose 1994 OKPOs were different, and 414 whose 1995 OKPOs were different. Since the registries contain previous year as well as current year values, we were able to fill in entire years of data for several thousand enterprises that existed in a particular year, but did not report. We added some additional enterprises and filled in missing values for enterprises already in the database from a panel database constructed by Economics, Analysis, and Marketing, Inc. (EKAM) of Moscow using a second version of the Goskomstat annual industrial censuses. In the process of linking enterprises across years, we identified several hundred cases where both consolidated data and data for each subsidiary appeared. Subsidiaries’ names usually included the name of the parent enterprise and the word subsidiary, and employment of the subsidiaries usually added exactly to the employment of the consolidated record. We do not include the subsidiary data in the analysis in this paper. Subsidiaries that become independent enterprises in a later year are considered entrants in the later year. We also identified many spinoffs. The previous year employment of the spun off enterprise equaled the difference between the former parent enterprise’s current year report of previous year employment and its previous year report of its then current employment. Once again, the data for spin-offs prior to becoming independent are excluded, and the spin-offs are considered entrants in the first year they appear as independent enterprises. Appendix Table A-1 shows that the coverage of the RERLD ranges from 63 percent of employment in 1989 to 97 percent in 1993. The value for 1989 is especially low because the 1989 data do not contain enterprises in the military-industrial complex (MIC).
46
Appendix Table A-1. Russian Enterprise Registry Longitudinal Database Compared to Goskomstat Official Data on Industrial Enterprises RERLD Number of Industrial Industrial Employment Enterprises 1989 1991 1992 1993 1994 1995 1996
21,391 22,553 24,767 28,901 32,302 31,704 26,530
13,751,839 17,405,833 18,118,911 18,334,184 16,789,450 14,062,580 12,483,301
Mean
Median
643 772 732 634 520 444 471
211 218 209 175 142 121 133
Minimum
Maximum
1 1 1 1 1 1 1
100,605 125,297 134,378 96,038 98,873 99,242 97,588
Goskomstat Published Numbers Number of Total Mean Enterprises Employment Percent of Mean Emp. Industrial Industrial Industrial Missing Missing Employment of Missing Enterprises Employment Employment from from in RERLD Enterprises RERLD RERLD 1989 1991 1992 1993 1994 1995 1996
26,734 28,023 61,075 104,059 137,999 136,674 156,000
21,731,000 20,117,000 20,020,000 18,864,000 17,440,000 16,006,000 14,934,000
813 718 328 181 126 117 96
5,343 5,470 36,308 75,158 105,697 104,970 129,470
47
7,979,161 2,711,167 1,901,089 529,816 650,550 1,943,420 2,450,699
63.3 86.5 90.5 97.2 96.3 87.9 83.6
1,493 496 52 7 6 19 19
Appendix B. Description of the 5-Digit Russian Industry Codes (OKONH) The 5-digit Russian industry code classification contains 350 industries with at least one enterprise, 271 of which are in manufacturing. Unless otherwise noted, when using the industry as the level of analysis, we restrict the number to 180 of the 271 manufacturing industries. We eliminated all miscellaneous industries (15), all industries with less than 10 percent of the average industry output in 1992, and those that do not produce the largest amount of any six-digit (KOD_NOM) product in 1993 (see Appendix C for information about the KOD_NOM codes). Examples of industries that were eliminated due to insufficiently large size include the fiberglass industry, the tin industry, the elevator industry, the coloring equipment production industry, the casting equipment industry, the technical equipment for glass production industry, the production of wall blocks, the cotton cleaning industry, artificial fur production, the yeast industry, the tobacco fermentation industry, and the musical instruments industry. Industries eliminated because they do not produce the largest amount of any product include the gas byproduct processing industry, precious metals and alloys production, titanium and magnesium, difficultly fusible and heat resistant metals, and the defense industry (this is just one of the MIC industries). Appendix Table B-1 shows the representativeness of the 180-manufacturing industry set compared to manufacturing as a whole. The 180-industry set includes over 94 percent of manufacturing industry output in 1992. The U.S. industry set will be discussed in Appendix C. Appendix Table B-1. Representativeness of the Industry Sets Percentage of Total Manufacturing Output in Each Branch Branch All Manufacturing 180-industry set 71 U.S. 1992 1992 industries 1992 Energy and Fuel 9.4 9.8 22.2 Metallurgy 18.7 18.9 30.4 Chemicals 10.7 10.0 15.2 Machine Building 26.0 24.3 14.9 Forestry 4.4 4.7 6.3 Construction 3.4 3.5 1.4 Materials Light Industry 9.1 9.6 1.1 Food Processing 13.3 14.1 5.9 Other Industries 4.7 4.8 2.7 Share of Total 100 94.1 40.1 Manufacturing Output
48
Appendix C. Description of the match between Russian 5-digit codes and 4-digit U.S. SIC codes We matched Russian industries as defined by 5-digit Russian industry codes to 4-digit U.S. SIC codes using detailed descriptions of both industry classifications. In cases where more than one Russian industry matched with a single U.S. SIC code, we combined the Russian industries and recalculate the number of firms and concentration ratios. In cases where more than one U.S. SIC code match with a single Russian industry code, we could not combine the U.S. industries to recalculate the concentration ratios, but we show them next to each other in Appendix Table C-1. The 71-industry set used in the correlations with U.S. concentration, the 1992 and 1996 levels regressions and out-of-equilibrium regressions includes only those industries with one-to-one matches between U.S. and Russian codes and ones where Russian industries can be combined to form the equivalent of a single U.S. code. The representativeness of the 71-industry set is shown in Appendix Table B-1. It contains 40 percent of total manufacturing industry in 1992. The set of industries in Appendix Table C-1 is similar but not identical to those in JST’s Table A-1. Our set contains several industries not included in their set, such as canned fruits and vegetables (2033), prepared feeds (2048), logging (2411), and printing trades machinery (3555). Others that JST include are not included here, as they do not appear to be good matches, such as knit outerwear mills (2253), knit underwear mills (2254), soaps and detergents (2841), and sporting and athletic equipment (3949). JST say that the 5-digit Russian codes are more narrowly defined on average than 4-digit U.S. SIC codes. There are some cases where Russian industries are narrowly defined relative to U.S. industries (e.g., the Russian codes include a tractor production industry, farm machinery industry, and an animal feed equipment industry, which are all represented in a single U.S. farm machinery and equipment industry (SIC 3523)), but in the majority of cases where Russian industries do not appear to match one-to-one with U.S. industries, the Russian industry appears to be broader (e.g., the U.S. has separate four-digit codes for electronic computers (3571), computer storage devices (3572), computer terminals (3575), computer peripheral equipment (3577), and calculating and accounting equipment (3578), all represented in the Russian calculating equipment and spare parts industry). This may be at least partially due to the fact that Russian enterprises simply do not produce all the products produced in the U.S., and certainly not in the same quantities, as the computer industry example suggests.
49
Appendix Table C-1. Comparison between U.S. and Russian Industries U.S. 1987 Russia USSIC U.S. Industry name # of CR4 OKONH Russian Industry name firms 2011 Fresh and frozen meat 1328 32 18211 Meat industry 2013 Sausage and other prepared 1207 26 meats 2015 Poultry and egg processing 284 28 2033 Canned fruits and 462 29 18152 Canned fruits and vegetables vegetables 2046 Wet corn milling 31 74 18149 Starch and molasses industry 2048 Prepared feeds 1182 20 19220 Animal feed 2051 Bread, cakes, related 1,948 34 18113 Bread baking 2064 Candy, confectionery 623 45 18114 Confectionary, including chocolate 2066 Chocolate and cocoa 173 69 products 2082 Malt Beverages 101 87 18144 Beer 2084 Wines, brandy 469 37 18143 Wine 2085 Distilled, blended spirits 48 53 18142 Liquor, vodka 2086 Bottled and canned soft 846 30 18145 Non-alcoholic beverages drinks 2087 Flavoring extracts and 245 65 syrups 2098 Macaroni and spaghetti 196 73 18115 Macaroni 2141 Tobacco stemming and 62 66 18181 Tobacco fermentation redrying 2371 Fur goods 380 16 17361 Fur and fur products and +17362 by customer order
United States
50
Russia 1992 Russia 1994 Russia 1996* # of CR4 # of CR4 # of CR4 firms firms firms 746 10 763 11 688 16
265
18
268
23
209
31
63
37
56
47
43
51
261 1746 409
17 3 18
276 1759 409
10 259 10 1612 21 312
20 4 26
252 126 109 84
16 19 19 24
250 132 122 104
16 28 18 22
221 149 142 135
21 33 18 40
52 2
22 100
62 5
23 97
60 8
24 99
71
57
102
46
97
59
United States USSIC U.S. Industry name 2411 2421 2435 2436 2452
2500 2611 2621 2631 2700 2812 2821 2822 2823 2824 2844 2851 2861 2865
Logging Sawmills, planing mills Hardwood veneer, plywood Softwood veneer, plywood Prefabricated wood buildings
U.S. 1987 # of CR4 firms 11,852 18 5,252 15 274 22 131 604
Furniture and fixtures group Wood pulp
10,775
Paper mills Paperboard Printing and publishing group Alkaline, chlorine Plastic materials, resins
122 91 57,376
Synthetic rubber Cellulosic manmade fibers Organic fibers, noncellulosic Toilet preparations Paints, allied products Gum, wood chemicals Cyclic crudes, intermediates
26
Russia OKONH Russian Industry name 15100 Logging 15210 Sawmill production 15250 Veneer, plywood, all woods
38 11 15240 Prefabricated wood buildings and housing, wood containers 10 15271 Furniture industry 44 15310 Cellulose, wood pulp, paper, cartons 33 32 7 19400 Printing industry
27 288
72 13114 Chlorine 20 13130 Synthetic resins, plastic pulp 58 50 13310 Synthetic rubber 6 (D) 13120 Chemical fibers 46 76
648 1,121 52 131
32 27 59 34
18131 13150 15400 13170
Perfume and cosmetics Varnish and paint Wood chemicals Synthetic dyes
51
Russia 1992 Russia 1994 Russia 1996* # of CR4 # of CR4 # of CR4 firms firms firms 2278 2 1796 6 1590 5 362 16 446 20 468 25 30 44 28 46 35 38
10
86
9
90
11
92
631
9
715
10
623
12
119
25
115
35
114
32
868
23
799
24 1493
23
8 31
81 51
11 35
78 50
13 36
59 53
15 21
68 35
15 22
75 42
15 25
85 45
35 85 15 13
48 59 61 57
43 89 14 12
70 36 74 76
45 78 15 13
93 32 64 83
United States USSIC U.S. Industry name 2873 2874
Nitrogenous fertilizers Phosphate fertilizers
2911 2951 2952 3011 3021 3111
Petroleum refining Asphalt paving mixtures Asphalt felts and coatings Tires, inner tubes Rubber footwear Leather tanning and finishing Household slippers Men's footwear, except athletic Women's footwear, except athletic Athletic and other footwear Glass containers Cement Brick and structural clay tile Ceramic wall and floor tile Vitreous china Semi-vitreous china Lime Gysum products Asbestos products Mineral wool
3142 3143 3144 3149 3221 3241 3251 3253 3262 3263 3274 3275 3292 3296
U.S. 1987 Russia # of CR4 OKONH Russian Industry name firms 117 33 13111 Nitrogen industry 55 48 13112 Phosphate fertilizers, inorganic chemicals 200 32 11220 Oil refining 542 19 16272 Asphalt production 162 47 114 69 13351 Tires 54 39 13363 Rubber shoes 311 28 17310 Natural leather 34 110
77 17371+ Shoe production and 26 17372 by customer order
123
50
119 35 123 167
24 78 16513 Glass packages 28 16112 Cement 29 16152 Bricks and ceramic tiles
95 65 32 78 16551 China and pottery 43 (D) 56 43 16232 Lime, gypsum products 80 75 50 72 16250 Asbestos industry 173 71 16240 Insulation materials
52
Russia 1992 Russia 1994 Russia 1996* # of CR4 # of CR4 # of CR4 firms firms firms 20 45 21 49 22 48 18 52 23 68 18 74 54 23
32 57
61 32
40 37
91 33
37 41
12 22 70
63 75 39
12 26 79
66 57 50
12 28 71
67 48 38
249
21
304
23
317
37
40 51 686
40 23 6
44 52 683
42 22 6
40 49 550
40 22 7
24
38
24
40
27
43
54
35
55
31
45
47
7 49
97 34
6 49
98 29
5 42
99 38
United States USSIC U.S. Industry name 3312
Blast furnaces, steel mills
3313
Electrometallurgical products Steel wire and related products Steel pipe, tubes Primary aluminum
3315 3317 3334 3341 3412 3441
3443 3491 3511 3523
3531
Secondary nonferrous metals Metal barrels, drums, and pails Fabricated structural metal
Fabricated plate work (boiler shops) Industrial valves Turbines, turbine generator sets Farm machinery and equipment Construction machinery
U.S. 1987 Russia 1992 Russia 1994 Russia 1996* Russia # of CR4 OKONH Russian Industry name # of CR4 # of CR4 # of CR4 firms firms firms firms 271 44 12130+ Ferrous metal products and 74 47 80 52 92 53 12160 coking 25 55 12150 Electroferroalloy 6 94 6 90 6 91 274 155 34 365 118 2,334
1,584 310 68 1,576
872
21 12190 Intermediate metal products 23 12140 Steel pipes 74 12212 Aluminum, aluminous oxide, and sodium flouride 24 12610 Secondary nonferrous metals 30 14834 Metal transport containers
41
61
55
56
53
63
22 18
69 57
24 19
71 61
19 22
74 62
35
54
42
54
44
56
4
100
7
89
14
86
11 14820 Containers and prefabricated (metal) buildings and housing 13 14112 Boilers
10
77
11
78
13
69
28
70
32
74
35
76
20 14195 Industrial valves 80 14111 Turbine building
30 19
40 64
34 23
47 76
37 25
50 75
241
34
257
36
234
33
133
28
141
23
129
26
45 14410+ 14420+ 14430 48 14511+ 14512+ 14520
Tractors, agricultural machinery, animal feeding machinery Road constuction, excavating, construction machines and construction materials equipment 53
United States USSIC U.S. Industry name 3532 3533 3534 3535 3536
Mining machinery Oil, gas field equipment Elevators and moving stairways Conveyors and conveying equipment Hoists, cranes, monorails
U.S. 1987 # of CR4 firms 293 22 563 34 158 52
OKONH Russian Industry name 14140 Mining machinery 14183 Oil field equipment 14154 Elevator production
Russia 1992 Russia 1994 Russia 1996* # of CR4 # of CR4 # of CR4 firms firms firms 49 31 55 29 55 30 50 40 59 35 70 40 9 96 9 86 9 91
703
17 14153 Continuous transport
10
80
11
78
11
64
165
19 14151+ 14152 35 14155+ 14156
50
49
52
58
50
55
22
64
20
57
20
71
88
26
92
30
91
41
18 14230 Forge and pressing 31 equipment 37 14254 Mechanical welding equipment
46
34
63
35
53
20 14611+ Technical equipment for 14612 textiles, sewing, and knitting 32 14220 Wood processing equipment 30 14185 Pulp, paper equipment 44 14640 Printing equipment 28 14620 Food and animal feed equipment
83
28
77
29
63
42
37
39
40
37
36
44
10 12 101
89 82 29
10 13 114
93 77 35
10 12 93
96 87 34
3537
Industrial trucks and tractors
448
3541
381
3548
Machine tools, metal cutting Machine tools, metal forming Welding apparatus
3552
Textile machinery
475
3553
Woodworking machinery
280
3554 3555 3556
Paper industry machinery Printing trades machinery Food products machinery
256 408 483
3542
Russia
196 203
Cranes except for and for construction Auto and electric loading equipment and other loading and unloading equipment 31 14210 Metal cutting tools
54
United States USSIC U.S. Industry name 3561
U.S. 1987
# of CR4 OKONH Russian Industry name firms 333 19 14194 Pumps
3562 3563
Pumps, pumping equipment Ball and roller bearings Air, gas compressors
3579
Office machines, n.e.c.
191
3585
Refrigeration, heating equipment Records, tapes, CD's Radio receivers, televisions, phonographs, speakers, and related Telephone and telegraph apparatus Radio and television equipment Communications equipment Storage batteries
746
3652 3651
3661 3663 3669 3691 3692 3695
Primary batteries, dry and wet Magnetic recording media
Russia
113 223
462 360
58 14350 Ball bearings 36 14186+ Compressor machine 14191 building and vacuum pumps 47 14327 Instruments for mechanization and automization of engineering and administrative work 31 14187 Refrigeration machine building 63 13145 Records, tapes, CD's 39 14760 Communications equipment
403
63
572
37
364
34
125 59
64 14175 Accumulator and cell industry 88
181
31 14332 Information carriers
55
Russia 1992 Russia 1994 Russia 1996* # of firms 23
CR4
# of CR4 # of firms firms 52 27 63 29
23 15
48 54
27 21
57 44
28 18
59 53
20
58
19
72
15
88
9
99
10
86
12
81
8 2
100 100
12 83 2 100
13 10
86 100
32
55
36
31
45
47
1 100
CR4 68
U.S. 1987 Russia 1992 Russia 1994 Russia 1996* Russia USSIC U.S. Industry name # of CR4 OKONH Russian Industry name # of CR4 # of CR4 # of CR4 firms firms firms firms 3731 Ship building, repairing 547 49 14740+ Ship building and 265 31 279 36 127 48 14921 repairing 3732 Boat building, repairing 2,108 33 3743 Railroad equipment 150 52 14160 Railway machinery 53 52 57 42 56 39 3751 Motorcycles, bicycles, 242 66 14342 Motorcycle, bicycle 11 94 13 84 15 70 parts production 3812 Search, detection, 918 29 14325 Instruments for the 27 44 35 47 33 56 navigation, etc. instruments measurement of and equipment mechanical quantities 3822 Environmental controls 230 50 3829 Measuring and controlling 938 15 devices, n.e.c. 3823 Process control instruments 707 24 14321 Instruments for controlling 86 28 91 28 91 34 and regulating processes 3825 Instruments to measure 864 35 14322+ Products for measuring 57 34 70 23 67 26 electricity 14323 electricity and radio circuits 3827 Optical instruments and 236 35 14324 Optical and optical45 43 41 46 40 57 lenses mechanical instruments 3873 Watches and clocks 213 45 14326 Time-keeping instruments 19 55 21 48 23 62 3931 Musical instruments 402 31 19720 Musical instruments 41 35 40 29 29 52 3942 Dolls and stuffed toys 191 34 19770 Toy production 83 22 87 27 59 50 3944 Games, toys, children's 698 43 vehicles 3965 Fasteners, buttons, pins 247 33 17510 Buttons 7 90 5 97 3 100 3991 Brooms and brushes 293 19 17380 Bristle-brush industry 24 52 22 48 25 47
United States
*1996 data are less complete because they haven't been corrected by 1997; the number of firms is biased downwards. (D) signifies withheld to avoid disclosing data for individual U.S. companies.
56
Appendix D. Description of the product data The RERLD contains physical volumes of production of (final) products by enterprise for 1993 through 1997. Subsidiary enterprise product data in the original registries have been aggregated to the enterprise level. The number of products in each product code category is displayed in Table 12. The physical volumes are attached to the KOD_NOM codes, the most disaggregated codes in the database. Since we only have physical volumes, we cannot calculate volumes at higher levels of aggregation. Appendix Table D shows the representativeness of the product data sets we use. The all product set includes all enterprises producing manufacturing products in the given year. Appendix Table D-1. Representativeness of the Product Sets Percentage of Total Manufacturing Output in Each Branch Branch All Manufacturing Panel Product All Product All Manufacturing All Product 1993 Set 1993 Set 1993 1997 Set 1997 Energy and Fuel 9.6 12.0 11.1 8.0 7.4 Metallurgy 16.8 20.6 19.7 19.9 20.3 Chemicals 8.6 9.8 9.5 11.0 11.4 Machine Building 29.8 24.9 25.6 27.2 26.7 Forestry 3.9 3.3 3.6 4.0 4.0 Construction 4.3 3.9 4.1 5.0 5.0 Materials Light Industry 6.5 6.7 6.9 2.7 2.7 Food Processing 16.1 14.9 15.5 16.9 17.3 Other Industries 4.4 4.0 3.9 5.3 5.2 Share of Total 100 72.0 77.8 100 96.2 Manufacturing Output
57
Appendix E. Description of variables used in regressions Both change in concentration variables (by HHI and by CR4) use percent change in concentration, that is, (CONC96-CONC92)/CONC92. CONC*GEODIS92—the interaction of concentration, measured in same way as the LHS variable, and the geographic dispersion of industries. Geographic dispersion is one minus ((the sum of the absolute values of the differences between the output share of the regions (oblasts, of which we have 78 in the data) and the population share of the regions) divided by 2). CONSUMERXX is the industry’s output of consumer goods as a share of industry output in 19XX (we have values for consumer goods in 1989 and 1993). CR4XX is the four-firm industry concentration ratio (value of output of the four largest firms in the industry as a share of the total value of output in the industry) in 19XX. DIFMES92 is the difference in 1992 between observed Soviet minimum efficient scale (the output of the median firm according to output as a share of total industry output) and observed technological minimum efficient scale (the output of the median firm according to cost as a share of the total industry output. FIRM PRODS93 is the average number of 6-digit OKP products produced by firms in the industry as a share of the number of products in the industry. HHIXX is the Herfindahl-Hirschman Index (sum of squared output shares of each firm in the industry). IND PRODS93—the total number of 6-digit OKP products produced by firms in the industry expressed as a natural log. INDSHAREXX is the industry’s share of manufacturing output during 19XX. MICSHAREXX is the share of 19XX industry output accounted for by firms in military industrial complex ministries. MIC enterprises are contained in the 1992-1996 data. We have added MIC enterprises in the 1992 data to the 1989 data using the following method. We first obtained a conversion between Russian industry codes and US SIC codes. We linked approximately 13,000 enterprises from the EKAM database, which contains data for some enterprises each year since 1985, to the 1989 PlanEcon database (which contains US SIC codes but not Russian industry codes). The conversion uses the US SIC code in which a plurality of the enterprises in the EKAM database in a particular Russian industry fall. The 1992 output for the MIC enterprises was converted to 1989 output using branch average nominal growth rates between 1989 and 1992. We thus assumed that MIC enterprises grew (or contracted) at the same rate as other enterprises in their branch. The 1989 data with the MIC enterprises is only used for the final set of 1989 regressions for the level of concentration (the ones where MICSHARE89 is used). Elsewhere, the data without the MIC enterprises is used. In the 1989 regressions where
58
MICSHARE92 is used, we use the share of the 1992 industry output accounted for by MIC firms, and this variable is assigned to the 1989 industry codes in the same manner as 1992 MIC firms are assigned to the 1989 data. The 1992 MIC firms are not included in the set in the regressions where MICSHARE92 is used, however. MINISTRYXX is the Herfindahl-Hirschman Index for the concentration of industry production in 19XX when all firms in the industry that are in a given branch ministry are counted as one firm. OUT GROWTH—the percentage change in the industry’s real value of total output from 1992 to 1996. PRIVSHARE94 is the share of the total output in the industry from firms with non-state or mixed non-state and state ownership in 1994. PRODASS92 is the difference between the 1992 concentration measure (HHI in HHI regressions, CR4 in CR4 regressions) when all firms in the industry that are in a given production association are counted as one firm and the regular concentration measure. PRODASSSHXX is the share of 19XX industry output accounted for by enterprises in production associations. PROFIT92 is the sum over enterprises in the industry of the ruble value of profits (losses) divided by the sum over the same firms of the ruble value of output in 1992. RUSINDSH is Russia’s share of total output in the industry in the Soviet Union in 1989. SETCOST92 is the minimum efficient scale (defined here as the average plant size of the 50 percent least cost firms as a share of total industry output) times the capital-output ratio of the industry (the sum over enterprises in the industry of the average value of fixed assets used in industrial production (main activity of the enterprise) divided by the sum over the same enterprises of the value of output). SOVCON89 is the difference between the 1989 HHI for the industry in Russia and the HHI for the industry in the Soviet Union. USCR4 is the U.S. four-firm concentration ratio by sales in 1987 from the U.S. Census 1987 Enterprise Statistics establishment data. USHHI is the U.S. Herfindahl-Hirschman Index in 1987 from the U.S. Census 1987 Enterprise Statistics.
59
Appendix Table E-1. Means of Variables Mean Standard Deviation HHI89 0.273 HHI92 0.119 HHI96 0.141 USHHI (1987) 0.063 CR489 0.671 CR492 0.494 CR496 0.534 USCR4 (1987) 0.368 PRODASS89 (HHI) 0.007 PRODASS92 (HHI) 0.003 PRODASS89 (CR4) 0.013 PRODASS92 (CR4) 0.008 SOVIETCON89 0.120 CONSUMER89 0.420 CONSUMER93 0.315 MINISTRY89 0.640 MINSTRY92 0.584 INDSH89 0.003 INDSH92 0.005 MIC OUTSHARE89 0.037 MIC OUTSHARE92 0.105 OUT GROWTH -0.926 FIRM PRODS93 0.211 IND PRODS93 4.000 PROFIT92 0.258 SETCOST92 0.058 HHI92*GEODIS92 0.024 CR492*GEODIS92 0.119
60
0.282 0.116 0.143 0.060 0.289 0.243 0.239 0.199 0.032 0.010 0.033 0.020 0.110 0.400 0.366 0.263 0.254 0.005 0.011 0.129 0.249 0.063 0.147 0.871 0.097 0.108 0.014 0.037
Appendix Table E-2. Direction of Out of Equilibrium in 1992 Variable HHI UP HHI DOWN Intercept 0.539 0.447 (0.36) (0.29) USDIFHHI -0.026 -0.062 (-0.30) (-0.68) HHI92 6.033 13.671 (0.69) (1.52) INDSHARE92 182.718 163.222 (1.17) (1.00) DIFMES92 -4.693 -5.009 (-0.79) (-0.80) CONSUMER93 4.994 5.211 (1.01) (1.04) MINISTRY92 -0.622 -3.759 (-0.26) (-1.39) PRODASS92 -3.318 -12.400 (-0.68) (-1.06) MICSHARE92 2.941 0.491 (0.62) (0.10) SOVCON89 -0.017 0.508 (-0.01) (0.41) Adj. R-square 0.204 N 69 These are multinomial logit regressions, where the observations are divided into three groups, ones where change in concentration between 1992 and 1996 was less than or equal to 5 percent (STABLE), ones with a greater than 5 percent increase (UP), and ones with a greater than 5 percent decrease (DOWN). The z statistics are reported in parentheses. *Significant at 10% in a two-tailed test. **Significant at 5% in a two-tailed test. ***Significant at 1% in a two-tailed test.
61
Appendix Table E-3. Determinants of Russian Concentration Levels During Transition Variable HHI 92 HHI 96 CR4 92 CR4 96 Intercept 0.196** 0.114 0.495*** 0.460*** (2.20) (1.65) (3.75) (3.41) USHHI 0.184 0.059 (0.62) (0.22) USCR4 0.066 0.037 (0.42) (0.25) INDSHARE92 -2.463*** -2.104*** -4.645*** -4.173*** (-2.81) (-3.01) (-3.08) (-2.86) CONSUMER93 -0.067 0.009 -0.150* -0.121 (-1.41) (0.12) (-1.72) (-1.29) MINISTRY92 0.075 0.033 0.082 0.110 (1.19) (0.56) (0.70) (0.92) PRODASSSH92 0.207 0.205** 0.276** 0.303** (1.65) (2.29) (1.97) (2.44) MICSHARE92 -0.122 -0.093 -0.190 -0.107 (-1.37) (-1.26) (-1.42) (-0.94) RUSINDSH89 -0.201 -0.035 -0.052 -0.002 (-1.46) (-0.41) (-0.35) (-0.01) Adj. R-square 0.290 0.168 0.184 0.170 N 69 69 70 70 The t statistics are reported in parentheses. The F test for the coefficients in the two HHI regressions being equal yields F(7, 122) = 1.34, Prob > F = 0.2365. The F test for the coefficients in the two CR4 regressions being equal yields F(7, 124) = 0.33, Prob > F = 0.9395. *Significant at 10% in a two-tailed test. **Significant at 5% in a twotailed test. ***Significant at 1% in a two-tailed test.
62
References Brown, Annette N. and J. David Brown, “Does Market Structure Matter? New Evidence from Russia,” CEPR Discussion Paper No. 1946, 1998. Brown, Annette N., Barry W. Ickes, and Randi Ryterman, “The Myth of Monopoly: A New View of Industrial Structure in Russia,” World Bank Policy Research Paper No. 1331, 1994. Caves, R. E. and M. E. Porter, “The Dynamics of Changing Seller Concentration,” Journal of Industrial Economics, XXIX, (1980), 1-15. Curry, B. and K. George, “Industrial Concentration: A Survey,” Journal of Industrial Economics, XXXI, (1983), 203-255. Davies, Stephen W. and Bruce R. Lyons, “Seller Concentration: The Technological Explanation and Demand Uncertainty,” Economic Journal, XCII, (1982), 903-919. Freris, Andrew, The Soviet Industrial Enterprise, (St. Martin’s Press, 1984). Geroski, P., Masson, R. and J. Shaanan, “The Dynamics of Market Structure,” International Journal of Industrial Organization, V, (1987), 93-100. Ghemawat, Pankaj and Robert E. Kennedy, “Competitive Shocks and Industrial Structure: The Case of Polish Manufacturing,” William Davidson Institute Working Paper No. 53, 1997. Gort, Michael, “Analysis of Stability and Change in Market Shares,” Journal of Political Economy, LXXI, (1963), 51-63. Hannah, L. and J. Kay, Concentration in Modern Industry (London: Macmillan, 1977). Hay, Donald A. and Derek J. Morris, “The Development of Firms, and Market and Industrial Structure,” chapter 15 in Industrial Economics and Organization: Theory and Evidence (Oxford: Oxford University Press, 1991). Hewett, Ed A., Reforming the Soviet Economy, (Washington, DC: Brookings, 1988). International Monetary Fund, World Bank, Organization for Economic Cooperation and Development, and European Bank for Reconstruction and Development, A Study of the Soviet Economy, (Paris: OECD, 1991). Joskow, Paul L., Richard Schmalensee, and Natalia Tsukanova, “Competition Policy in Russia during and after Privatization,” Brookings Papers on Economic Activity: Microeconomics, (1994), 301-381. Levy, David, “Specifying the Dynamics of Industry Concentration,” Journal of Industrial Economics, XXXIV, (1985), 55-68. Martin, S., “Advertising, Concentration, and Profitability: The Simultaneity Problem,” Bell Journal of Economics, X, (1979), 639-647. Mueller, W. and Hamm, L. 1974. “Trend in Industrial Market Concentration, 1947 to 1970,” Review of Economics and Statistics, LVI, (1974), 511-520. Nelson, R. R. and S. G. Winter, “The Schumpeterian Trade-off Revisted,” American Economic Review, LXXII, (1982), 114-32. Newbery, D. and P. Kattuman, “Market Concentration and Competition in Eastern Europe,” World Economy, XV, (1992), 315-334. Pryor, Frederic L., “An International Comparison of Concentration Ratios,” Review of Economics and Statistics, LIV, (1972), 130-140. Roberts, Barbara M., “Transition and Changes in Industrial Concentration in Poland,” University of Leicester Discussion Paper No. 98/4, 1998.
63
Scherer, F. M. and David Ross, Industrial Market Structure and Economic Performance, (Boston: Houghton Mifflin Company, 1990). Sleuwagen, Leo and Wim Dehandschutter, “The Critical Choice between the Concentration Ratio and the H-index in Assessing Industry Performance,” Journal of Industrial Economics, XXXV, (1986), 193-208. Streitwieser, Mary L., “The Extent and Nature of Establishment-Level Diversification in Sixteen U.S. Manufacturing Industries,” Journal of Law and Economics, XXXIV, (1992), 504-34. Weiss, Leonard W. 1965. “An Evaluation of Mergers in Six Industries,” Review of Economics and Statistics, XLVII, (1965), 172-181. Wright, N., “Product Differentiation, Concentration, and Change in Concentration,” Review of Economics and Statistics, LX, (1978), 628-631.
64
