Professional Documents
Culture Documents
Industry Concentration Paper 2010
Industry Concentration Paper 2010
net/publication/46528769
CITATIONS
READS
24
3,223
2 authors, including:
David Nawrocki
Villanova University
64 PUBLICATIONS 688 CITATIONS
Some of the authors of this publication are also working on these related projects:
All working papers co-authored with Fred Viole constitute my current working projects. View project
All content following this page was uploaded by David Nawrocki on 29 May 2014.
Applied Economics
Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t713684000
To cite this Article Nawrocki, David and Carter, William(2010) 'Industry competitiveness using Herfindahl and
entropy concentration indices with firm market capitalization data', Applied Economics, 42: 22, 2855 — 2863, First
published on: 03 August 2010 (iFirst)
To link to this Article: DOI: 10.1080/00036840801964666
URL: http://dx.doi.org/10.1080/00036840801964666
This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.
Applied Economics, 2010, 42, 2855–2863
I. Introduction
have no effect on the use of market capitalization
This article received its start when the new Standard values as proxies for market concentration and (2) it
and Poors eight-digit GIC industry classification only includes firms that are publicly traded.
codes were introduced in January 2001. As part of Even with these limitations, the improvement in the
a project to generate 125 new survival bias-free quality of data by using market capitalization data
industry indices, an extensive database of market over the problematic use of market share and sales
capitalization values grouped by industry classifica- data should provide interesting results. The remain-
tions going back to 1971 was developed. As no der of this article is organized as follows: Section II
previous study (that we are aware of) of industry discusses the type of data used previously in industry
concentration has utilized market capitalization data, concentration studies. Section III presents an over-
an opportunity for an interesting research project view of the industry concentration measures used
presented itself. within the academic literature. Empirical results are
We argue that if the financial markets are reported in Section IV. Finally, conclusions are
successful at valuing a firm’s prospects, then market provided in Section V.
capitalization of firms within an industry is a good
proxy for computing market concentration measures
within an industry. This approach has two inherent
weaknesses: (1) it requires that firms’ debt to equity
II. The Data
ratios1 have no effect on the market concentration of
an industry; in other words, all firms within an
The best test for the existence of monopoly power
industry have the same leverage ratio, and therefore,
within an industry is to acquire and analyse cost and
Applied Economics ISSN 0003–6846 print/ISSN 1466–4283 online © 2010 Taylor & Francis 2855
http://www.informaworld.com
Applied Economics, 2010, 42, 2855–2863
DOI: 10.1080/00036840801964666
2856 D. Nawrocki and W.
Carter
demand data for firms within an industry. Although used the Sharpe, Treynor and Jensen performance
these data have been studied (Dickson, 1994), Curry measures in their study.
and George (1983) note that these data are very Industry indices and market capitalization values
difficult to analyse effectively. Even when cost and were calculated for 13 industries (using the eight-
demand data are obtainable, Dickson (1994) notes digit Global Industry Classification (GIC) sub-
that it has to be adjusted carefully for firm size. As industry classifications introduced in 2001) using
a result of this poor data availability, researchers daily and monthly data from the CRSP dataset from
have attempted to use sales and market share data January 1971 to December 2001. These data provide
that has also proven difficult to obtain (Kelly, 1981; another advantage over sales and market share data
Weinstock, 1982). Another problem is that Curry and by being available on a daily and monthly basis.
George (1983) argue that sales data will favour
distribution firms over manufacturing firms. They
suggest that an alternative is to use only the size of
a firm’s assets, but that approach can be misleading III. Concentration and
due to dependence on accounting conventions. Competition Measures
Market capitalization data avoids the problem of
dependence on accounting conventions and is fairly The use of different measures of concentration or
easy to obtain from the University of Chicago Center competition within an industry is also subject to
for Research in Security Prices (CRSP) dataset. 2 The debate. Kowka (1985) states that no single concen-
only problem is whether market capitalization data tration measure can capture everything happening
Downloaded By: [Nawrocki, David] At: 03:12 4 August 2010
are a fair proxy for firm size because they ignore debt within an industry. A thorough economic analysis of
information. However, because any additional firm an industry is sure to uncover much that is missed
value or loss of firm value due to monopolistic by a single index number. Curry and George (1983)
behaviour (or through economies of large-scale opine that no single concentration measure is best
operation) will accrue to the shareholders rather because of the complexity of business life. They
than bondholders, this assumption should be a recommend that a number of measures need to be
reasonable one. tried because the measures will be dependent on the
This brings us to an important issue raised by particular industry. Accordingly, this study utilizes a
Gale and Branch (1982) who asked whether the number of measures and provides another approach
profitability of an industry results from economies to the problem of analysing industry concentration.
of scale due to larger market shares or whether it This approach is exemplified by Nissan and Caveny
is due to oligopolistic behaviour within the industry. (1993) who used six different measures to find that
Utilizing business unit data (that is no longer there was an increase in industry concentration in the
available), they concluded that market share was US in the 1980s.
more important than industry concentration in The traditional measure that has been published by
determining profitability. They conclude that anti- census departments is the n-firm concentration ratio.
trust policies may reduce market efficiency. However, numerous authors have demonstrated the
Woodrow (1990) adds support to this view by finding superiority of the Hirschman–Herfindahl index
that increases in labour productivity (scale efficiency) (commonly called the Herfindahl index) over the n-
can increase industry concentration. We cannot firm concentration ratios (Weinstock, 1982; Curry
add to this debate because we cannot separate out and George, 1983; Michelini and Pickford, 1985;
the market share and industry concentration effect Sleuwaegen and Dehandschutter, 1986; Cortes,
without the business unit data. However, we can 1998), with one dissenting vote (Kowka, 1985). The
point out that the four-firm concentration ratios Herfindahl Index (HI) is computed as the sum of
used in the Gale and Branch study are inferior to the squared firm proportions within an industry (s)
other concentration ratios (see discussion given and will vary between 0 and 1.
further).
HI ¼ Σs2 ð1Þ
This article provides an additional measure of
market power concentration and reminds the profes- where Σs ¼1.0. The HI is an index of concentration
sion that equity risk and return measures have not within an industry. A high number is indicative of a
been used in these studies since Melicher et al. (1976) high degree of concentration or monopoly power and
2
CRSP has had data accuracy problems with delisting
codes and values used to compute survivor bias free
industry indexes (Shumway, 1997). Recently, CRSP has
spent a great deal of time correcting these problems. Even
so, CRSP data have to be watched carefully for problems,
especially with recent data.
Industry competitiveness using Herfindahl and entropy concentration 2857
indices
a low number indicative of a high degree of equivalent number of firms’ calculation is available
competition. from information theory.
Another measure described by Horowitz and
Horowitz (1968) is the entropy measure of competi- E ¼ eH ð5Þ
tion. In information theory and the physical sciences, where E is the number of equal size firms that would
entropy is a measure of the degree of disorder or
result in the same amount of entropy H. The
uncertainty of a system. Let s be the probability of a
equivalent number of equal size firms (E) can be
discrete event, the entropy H is given by:
used to compare the degree of competition in one
industry with that in another industry or to compare
H ¼ —Σs ln s ð2Þ
the degree of competition at one period of time with
that at another.
Maximum entropy is reached when each prob-
Finally, the SD and the semi-deviation (downside
ability of each discrete event is equal, or s¼ 1/n where
risk measure) are utilized to track changes in the
n is the number of events. The entropy of a
equity risk within an industry.3
communication source provides a model for measur-
ing the ‘competitiveness’ of an industry. If entropy is
used as a measure of the degree of competitiveness,
then s is the firm’s proportional market value within
IV. The Empirical Results
the industry and the higher the entropy value the
higher the degree of competitiveness.
Armed with these measures of competitiveness, the
Downloaded By: [Nawrocki, David] At: 03:12 4 August 2010
No. of companies
% Concentration
(a)
Telecommunications diversified (AT&T) (b) Telecommunications (AT&T) equivalent and total
number of companies
Concentration index
Downloaded By: [Nawrocki, David] At: 03:12 4 August 2010
No. of companies
Time
Time (r = 0.9907)
Fig. 2. Telecommunications
Still the R2s are very low, thus supporting the view of
were estimated for each company:
Attaran and Saghafi (1988) who found little
evidence of any relationship between the mean CRt ¼ α þ þEt þ "t ð6Þ
return and SD of assets and the entropy asset
CRt ¼ α þ þRt þ "t ð7Þ
concentration index.
One interesting idea to test is whether the largest CSt ¼ α þ þRt þ "t ð8Þ
firm within an industry benefits from a reduction
in competitiveness. Therefore, the largest firm was CSVt ¼ α þ þRt þ "t ð9Þ
selected within each industry and conditional return,
where CRt is the conditional return for the largest
conditional SD and conditional semivariance
company in the industry, CSt is the conditional
measures were computed by subtracting the industry
monthly SD for the company, CSVt is the
return from the company return for each data
conditional monthly semi-variance for the
observation. Monthly returns were utilized from the
company, Et is the equivalent number of companies
monthly CRSP dataset. The daily CRSP dataset was
in the industry (Equation 5), and Rt is the relative
utilized to compute monthly SDs and semi-variances
entropy concen- tration measure for the industry
for both the industry indices and the individual
(Equation 3).
companies.
A positive regression coefficient with the entropy
In Table 1, the conditional statistics for the largest
regression Equations 7, 8, 9 indicates that as com-
company in the industry are regressed using the
petition increases (entropy increases), the conditional
entropy or equivalent number of firms as the
return or variability of the firm increases. A negative
independent variable.4 The following regressions
regression coefficient indicates that as competition
increases, the conditional return or variability of the
4
See Whitelaw (1994) for a description of the methodology for computing conditional returns and the monthly SD and semi-
deviation statistics.
Industry competitiveness using Herfindahl and entropy concentration 2859
indices
(a) (b) Equivalent number of companies
Computers and peripherals and the number of companies
Concentration index
No. of companies
Time (r = 0.9735)
Time
(a)
M E
o q
Concentration index
n u
e i
y v
Downloaded By: [Nawrocki, David] At: 03:12 4 August 2010
a
C l
e e
n n
t t
e
r n
u
B m
a b
n e
k r
s
a
n
d
n
u
m
b
e
r
T
i o
m f
e
c
(
o
r
m
p
= a
n
0 i
. e
2 s
4
9
1
No. of companies
)
Time
No. of companies
Time (r = 0.9266)
No. of companies
Time (r = 0.9070)
Concentration index
No. of companies
(a) Oil and gas (b) Integrated oil (Exxon-Mobil) equivalent number and number of companies
Concentration index
No. of companies
firm decreases. With the equivalent number of firms regressions indicate that as the industry com-
regression Equation 6, a positive (negative) coeffi- petitiveness declines, the industry return and risk
cient indicates that as the number of firms decrease, are increasing. The company regressions for
the return of the firm decreases (increases). If the Microsoft indicate that as competitiveness (entropy)
relationship is insignificant, the competitiveness of declined, Microsoft’s return and variance increased.
the industry has no effect on the company. The most However, if only downside risk (semi-variance) is
likely explanation is that the company is not large studied, Microsoft’s downside risk declined as the
enough to derive any marginal changes from changes competitiveness of the industry declined. This latter
in the competitiveness of the industry. Another result is very interesting as it indicates Microsoft
explanation is that the industry competition is reduced its risk as it grew to dominate the industry.
stable enough that no competition effect is noted. In addition, the R2 for the semi-deviation regression
First, we will take an in-depth look at Microsoft is almost four times higher than the R2 for the
and the software industry and then we will SD regression.
summarize the other industry results. The results in Table 2 provides more detailed results for
Fig. 1, Table 1 and Table 2 for Microsoft are Microsoft. First, autocorrelation (missing variables)
interesting. In Fig. 1, the entropy measure indicates is not a problem as evidenced by the Durbin–Watson
that the software industry was pretty competitive statistics.5 The equivalent number of companies is
before Microsoft started trading publicly in March highly correlated with the entropy and Herfindahl
1986. The HI is also steadily improving during this measures but the overall number of companies is not.
period indicating increasing competitiveness. After Therefore, the overall number of companies variable
Microsoft went public, both indices started a long is used in the following regressions,
decline. The number of firms in the industry increase
throughout the period but the equivalent number of CRt ¼ α þ þ1Rt þ þ2Nt þ "t ð10Þ
firms peaks in early 1986 and starts to decline
thereafter. Although the R2s are low, the industry CRt ¼ α þ þ1HICt þ þ2Nt þ "t ð11Þ
5
Generally, the Durbin–Watson statistics indicates no autocorrelation in all of the regressions reported in this study.
Industry competitiveness using Herfindahl and entropy concentration 2861
indices
Table 1. Simple linear regressions for company conditional statistics against entropy and equivalent number of companies
Number of Durbin–
Dependent Entropy Herfindahl companies Adjusted R2 Watson Equation
Conditional —1.5887 (—14.38) 0.4202 (6.10) 0.5477 2.1243 10
return
Conditional —2.0684 (—14.80) 0.3490 (5.21) 0.5617 1.9244 11
return
MSFT R/SV —27.7178 (—11.60) 7.6688 (5.14) 0.4415 2.0087 12
ratio
MSFT R/V —14.3658 (—13.73) 3.5283 (5.40) 0.5211 2.0027 13
ratio
Correlation matrix
Equivalent
Number of number of
Entropy Herfindahl companies companies
Entropy 1.0000
Herfindahl 0.9358 1.0000
Number of 0.2015 0.1296 1.0000
companies
Equivalent 0.9343 0.8326 0.5108 1.0000
Downloaded By: [Nawrocki, David] At: 03:12 4 August 2010
number of
companies
Notes: MSFT R/SV ratio – Monthly reward to semi-variability ratio for Microsoft. MSFT R/V ratio – Monthly reward to
variability ratio for Microsoft. t-statistics are in parentheses.
In Fig. 4, the Money Center Banks show a long- leads to increases in risk. Citigroup shows a positive
term decline in competitiveness and number of relationship between SD and competition but a
companies. Alcoholic beverages (Fig. 5) demonstrate negative relationship between semi-variance and
that a long-term decline in competitiveness can be com- petition, i.e. downside risk increased as
reversed as the competition ratios and the number of competition increases. Applied materials
firms increase starting in 1997. General retail (Fig. 6) (Semiconductor equip- ment), Wal-Mart (Retail) and
also presents a picture of declining competitiveness/ Enron (Multi-utilities) have very weak negative
increasing concentration since the 1980s. relationships between risk and competitiveness
Multi-line utilities (Fig. 7) is very interesting while AMBAC Financial (Financials) has a weak
because of the Enron effect. The competition indices positive relationship.
and the equivalent number of companies start to In terms of conditional returns, only Microsoft has
decrease in 1998, which is probably a result of the a significant relationship between conditional return
technology bubble inflating Enron’s market capitali- and competitiveness. The rest of the companies’
zation. After the Enron bankruptcy in 2001, all of the conditional returns are not affected by the competi-
indices increase. Enron’s fraudulent market capitali- tiveness of the industry. The equivalent number of
zation significantly biased these indices during this companies has a significantly negative effect on
period. conditional returns of AT&T, IBM, AMR, Enron,
Finally, Oil and Gas (Fig. 8) also exhibits a long- Exxon-Mobil and Pfizer. Microsoft, Nortel, Applied
term decline in competition even with an increase in Materials, Citigroup, AMBAC Financial, Anheuser
the number of firms in the late 1990s. Busch and Wal-Mart do not see their conditional
In the regressions in Table 1, Nortel (Communica- returns affected by the number of firms.
tions Equipment), AMR (Airlines), Anheuser Busch
(Alcoholic beverages), Exxon-Mobil (Oil and Gas)
and Pfizer (Drugs) are found to have strong negative V. Summary and Conclusions
relationships between their risk measures and the
entropy concentration measure, i.e. increases in Using market capitalization data obtained from a
competition lead to decreases in the risk of the firm. project on industry indices provides an interesting
AT&T and IBM show positive relationships between look into the competitiveness of 13 different indus-
risk and competitiveness, i.e. increase in competition tries. Competitiveness or concentration measures
were computed using firm market value rather than
Industry competitiveness using Herfindahl and entropy concentration 2863
indices Dickson, V. (1994) Aggregate industry cost functions and
market shares, sales or cost and demand functions.
The Herfindahl, entropy, relative entropy and the Herfindahl index, Southern Economic Journal, 61,
445–52.
equivalent number of firm measures were computed Gale, B. T. and Branch, B. S. (1982) Concentration
for the period 1971 to 2001. Regressions between the versus market share: which determines performance
largest firm in the industry and the competitive- and why does it matter?, Antitrust Bulletin, 27, 83–
ness measures indicate that whenever competition 105.
declines, it has a positive effect on the firm’s Horowitz, A. and Horowitz, I. (1968) Entropy, Markov
return conditioned on the industry returns (nine of processes and competition in the brewing industry,
Journal of Industrial Economics, 16, 196–211.
13 industries). However, most firms see their risk Jacquemin, A. P. and Berry, C. H. (1979) Entropy measure
relative to the industry increase as competitiveness diversification and corporate growth, Journal of
declines (nine of 13 industries). It should be noted Industrial Economics, 27, 359–70.
that results differ from industry to industry indicating Kelly Jr, W. A. (1981) A generalized interpretation of the
the wisdom of the Curry and George approach Herfindahl index, Southern Economic Journal, 48, 50–
7.
(1983). Microsoft is the only company to benefit
Kowka Jr, J. E. (1985) The Herfindahl index in theory and
significantly from a reduction in competitiveness in practice, Antitrust Bulletin, 30, 915–47.
its industry. AT&T and IBM saw significant Melicher, R. W., Rush, D. F. and Winn, D. N. (1976)
reductions in their risk/return performance after their Degree of industry concentration and market risk-
industries became more competitive in the 1980s. return performance, Journal of Financial and
Future research would include using the HI in the Quantitative Analysis, 11, 627–35.
Michelini, C. and Pickford, M. (1985) Estimating the
regressions and running more regressions with Herfindahl index from concentration ratio data,
Downloaded By: [Nawrocki, David] At: 03:12 4 August 2010
the equivalent number of firms. In addition, both Journal of the American Statistical Association, 80,
the entropy and relative entropy measures would be 301–5.
utilized. Furthermore, conditional risk and return Nawrocki, D. N. (1999) A brief history of downside risk
variables using macrovariable regressions could be measures, Journal of Investing, 8, 9–25.
attempted. Nissan, E. and Caveny, R. (1993) Concentration of sales
and assets of the top 25 fortune 500 firms: 1967–90,
Another interesting study would be cross-sectional Applied Economics, 25, 191–7.
regressions over time looking at relationship between Shumway, T. (1997) The delisting bias in CRSP data,
industry risk and return and the industry competi- Journal of Finance, 52, 327–40.
tiveness indices. Sleuwaegen, L. and Dehandschutter, W. (1986) The critical
choice between the concentration ratio and the H-
index in assessing industry performance, Journal of
Industrial Economics, 35, 193–208.
References Weinstock, D. S. (1982) Using the Herfindahl index
Attaran, M. and Saghafi, M. M. (1988) Concen- to measure concentration, Antitrust Bulletin, 27, 285–
tration trends and profitability in the US manufactur- 301.
ing sector: 1970–84, Applied Economics, 20, 1497– White, A. P. (1982) A note on market structure measures
510. and the characteristics of markets that they ‘measure’,
Cortes, B. S. (1998) Trends in industrial concentration in Southern Economic Journal, 49, 542–9.
Japan, 1983–92, International Review of Applied Whitelaw, R. F. (1994) Time variations and covariations
Economics, 12, 271–81. in the expectation and volatility of stock market
Curry, B. and George, K. D. (1983) Industrial concentra- returns, Journal of Finance, 49, 515–41.
tion: a survey, Journal of Industrial Economics, 31, Woodrow Jr, E. E. (1990) Concentration changes and large-
203–56. firm/small firm efficiency differences: evidence from
Davies, S. (1980) Measuring industrial concentration: an US manufacturing industries, Applied Economics, 22,
alternative approach, Review of Economics and 131–42.
Statistics, 62, 306–9.