Lee SCP NEIO and Beyond
Lee SCP NEIO and Beyond
Lee SCP NEIO and Beyond
2.1 Introduction
The relationship between firm behavior and market structure has been a central focus of
study in the field of industrial organization (IO). This emphasis is reflected in the manner
in which some economists have defined IO, namely as the study of firm behavior in
imperfectly competitive markets.1 Historically, the discipline's emphasis on firm
behavior and market structure is, to a large extent, influenced by the work of a group of
economists at Harvard in the 1930s. Edward Mason and his PhD student Joe S. Bain
formulated a framework for empirical analysis called the Structure-Conduct-Performance
(SCP) that attempted to describe how key aspects of market structure relate to each other.
Stephen Martin summarizes this framework approach succinctly:
The central hypothesis (of the SCP framework) is that observable structural
characteristics of a market determine the behavior of firms within that market, and
that the behavior of firms within a market, give structural characteristics,
determines measurable market performance.2
The SCP paradigm became the dominant framework for empirical work in IO between
the early 1950s until the early 1980s. Its influence only began to wane in the 1980s with
the emergence of game theoretical analysis of oligopolistic markets - an approach labeled
as the `New Industrial Organization' (NIO). The body of empirical associated with this
approach is known today as the New Empirical Industrial Economics (NEIO).
The chapter provides a survey of the SCP, NEIO and developments beyond. There have
been some excellent surveys of empirical work in IO, for example, Martin (2002) and
Schmalensee (1989). We draw from this body of literature (rather than re-inventing the
wheel) but at the same time attempt to extend the body of this literature to include new
research that has emerged since these survey papers were written.
1
For example:
Industrial organization or industrial economics is the study of the operation and performance of
imperfectly competitive markets and the behavior of firms in these markets. (Church and Ware
(2000), p.7)
Industrial organization is concerned primarily with the intermediate case of oligopoly, that is,
competition between
a few firms. (Cabral (2000), p.3)
2
Martin (2002), p.119.
The outline of the chapter is as follows. Section 2.2 provides a review of the structureconduct-performance paradigm. This is followed by a review of the empirical literature
on SCP in Section 2.3. The new empirical industrial organization is discussed Section
2.4 while Section 2.5 reviews developments beyond the SCP and NEIO. Section 2.6
concludes.
2.2 Structure-Conduct-Performance
(a) Origins: Debate on Theory and Empirics
The origin of the SCP paradigm can be traced to the work of the Harvard economist
Edward Mason in the 1930s. The theoretical work of Mason's colleague Edward
Chamberlin provided inspiration for both Mason and his student Joe Bain to study
empirically how pricing and production policies of firms (especially large ones) are
determined. Mason (1939)s (p.63) starting point was that market share is important in
determining production and pricing policy of a firm. In the 1930s, there were generally
two approach in understanding pricing policies of firms, namely:
1. Theoretical approach - involving the use of oligopoly and monopolistic models to
derive production and pricing policy of a firm.
2. Empirical approach - involving the correlation between observed prices and other
economic variables representing differences in market structure
Mason argued that empirical analysis is essential to ensure that the theories of firm are
useful. This is because theories are based on mathematical constructs such as demand
and cost functions which are not ascertainable (in Mason's words, p.64). Thus, it is not
that theories are not important, rather their relevance cannot be determined without
empirical observations. This leads to the question of the set of empirical observations that
are useful.
Interestingly, Mason argued that the price and production decisions of a firm is
influenced by both the internal organization of the firm as well as market structure.
Internal organization here refers to group relationships within the firm which exerts
influence on the firm's policy.3 According to Mason, market structure is a
multidimensional concept that is specified and measured by variables such as product
characteristics, cost and production characteristics, and the number and market shares of
buyers and sellers in the market. There are also other factors that influences firm
behavior such as industry life-cycle and the characteristic of the distribution channels.
The relevance of these factors are discussed by use of anecdotal evidence from different
types of industries such as automotive, steel, rubber tyre, distributive trade and
construction materials.
3
The emphasis on group relationship within firms is similar to the behavioral theories of the firm that was
developed in the 1950s and 1960s. See Simon (1955) and Cyert and March (1963).
In the concluding part of his paper, Mason (1939) drew up a framework for the analysis
of production and pricing policy of a firm:
The argument ... runs from differences in market structure to
differences in price response, and from differences in price response to the
consequences of these differences for the functioning of the economy."
(p.73-74)
Thus, elements of the SCP paradigm were already present in the Mason's work in the
1930s. During this period, Mason also lamented the lack of empirical work in this area:
Although a good deal has been written both on the effect of restrictive
policies on the distribution of resources and the effect of price policies on
fluctuations in employment and output, very little has been done to
formulate tests of undesirable price behavior applicable to public action.
Specifically, what sort of tests are indicative of the existence of a price
sufficiently high to restrict output and investment below desirable levels?"
(p.74, italics added.)
Mason challenge for future empirical-policy work was subsequently taken-up by his
Ph.D. student, Joe S. Bain. Despite being inspired by the work of Mason, the research
methodologies of the master and his student were a bit different. Bain used industry-level
data - an approach which Mason was a bit skeptical of. In contrast, Mason was more in
favour of case studies involving specific firms or industries. It was Bain's work which
proved to be more influential in charting the course of empirical IO after the 1930s.
The SCP paradigm posits specific causal relationships between market structure, conduct
and performance. In particular, market structure determines conduct and conduct in turn
determines performance:
Structure
Conduct
Performance
Structure
Conduct
Performance
Some contend that the relationship between conduct and performance is also weak. For
example, one can further argue that performance can affect conduct:
Structure
5
6
Conduct
Performance
For example, firms with substantial accumulated profits can incur losses in the short-term
to drive out rival firms. If this is true, the SCP has low predictive power.
Li = P(Q)-MC(qi) = si (1 + i )
P(Q)
(1)
where i = dqj
dqi
Theoretically, the conjectural variation variable i measures the output response of the
firm's rivals. Scherer and Ross (1990) further suggests that it is also a measure of the
degree of coordination (or collusion) between firms in the industry. The conjectural
variation variable is determined by other factors:7
i = f1(Cj , Bj , Xij)
(2)
where Cj is a measure of seller concentration, Bj a set of entry barrier measures and Xij
other industry or firm characteristics affecting the conjectural variation. The above
equation provides the link between market structure and conduct. Substituting the second
equation into the first, we obtain a link between structure and performance (the Lerner
Index) for firm i:
Li = f2(si , , Cj , Bj , Xij)
(3)
In reality, the Lerner Index may not be observable. If there is a correlation between the
Lerner index and measures of profitability (i), the above equation can be reformulated
as:
i = f2(si, , Cj , Bj , Xij)
(4)
j = f3( Cj , Bj , Xj)
(5)
It should be clear from the above specifications that the empirical test of the SCP entails
testing for the relationship between structure and performance, taking conduct as either a
black box or theoretically proven. The hypothesis underlying the above specifications is
that concentration determines profitability.8
Measuring Performance
A key issue in the empirical literature in SCP is the measurement of performance. A
number of measures of performance have been used. Theory suggests that the Lerner
index is a good measure of the extent of a firm's market power:
Lerner Index = Price Marginal Cost
Price
When the Lerner index > 0, firms are said to have market power. However, it is not
always possible to derive the Lerner index empirically. It may be difficult to obtain
marginal cost data. Furthermore, firms may have numerous products, each priced
differently.
A measure or performance that is conceptually closest to the Lerner index is the price
cost margin (PCM):
PCM = value added payroll
value of shipments
(6)
where value added is calculated by substracting input cost from total sales.
One weakness of using the PCM as a measure of performance is that it requires
controlling for the normal rate of return on capital across different industries.
Another measure of performance that has been used is Tobin's q ratio which measures the
ratio of firms stock market value to replacement cost of capital:
q = Mc + Mp + Md
Ar
(7)
where Mc is the market value of ordinary shares, Mp market value of preference shares,
Md outstanding loan capital and Ar total assets at replacement cost. When q > 1, firms
have intangible resources or advantages that are not captured in asset valuation such as
market power. The advantages of using the Tobin q ratio is that it captures all available
information on a firms future profitability (adjusted for risk). However, it suffers from
some severe limitations such as limited coverage (only listed firms which biases the
results towards larger firms), subjective and volatile valuation of firms and the difficulties
in estimating replacement cost.
Finally, accounting measures of performance are also used. There are various versions:
1 =
profit
revenue
(8)
2 = profit
capital
(9)
3 = profit
equity
(10)
4 =
(11)
profit
net worth
equity
revenue
(12)
The major source of data for this approach is published annual reports or financial
statements. In this approach, profits can be defined as profits before or after tax. If the
accounting data short used covers only a short period, it can be affected by heavy
discretionary investment expenditures (R&D, marketing) in a given year. The use of
accounting data also engender further debate about the appropriate depreciation method
(straight line or accelerated) to be used. Firms may also differ significantly from one
anther in terms of their gearing ratio (debt-to-equity ratio). There is also debate on
whether inflation should be taken into account via the use of historical or replacement
cost.
There is no consensus on which is the best measure of performance. The choice of
measure obviously depends on data availability and the desired aggregation level of
analysis i.e. industry, firm or plant.
Measuring Concentration
The theoretical link between the Lerner Index (L) and market share (si) implies that we
can measure market power by measuring market concentration:
Li = P(Q) MC(qi) = si ( 1 + i )
P(Q)
(13)
where si is firm i's market share, ( ) is price elasticity of demand, (i) its conjectural
variation and i = dqj . Since si is directly related to the Lerner index, an obvious
dqi
measure of concentration is the total market shares (si ) of firms. One such measure is
the concentration ratio which measures the total market share of a given number of (m)
firms with the largest market shares:
CRm = si
(14)
One critique of the concentration ratio is that it does not take into account the distribution
of market share across all firms in an industry. A concentration index that does not share
this weakness is the Herfindahl-Hirschman Index (HHI):
HHI = si
(15)
The HHI is also directly related to the Lerner index. For an industry with n firms, the
industry's weighted average Lerner index is:
L = si Li
= si ( 1 + i )
(16)
(17)
The CR and HHI are the two most commonly used concentration indices used in
empirical SCP studies. There are other measures of market concentration that are used to
measure the degree of inequality in firm size distribution. These measures include the
Hannah-Kay index, Gini Coefficient and the Entropy Index.9 Debates on the choice of
which concentration measures to used have revolved around correlation between the
different measures and the sensitivity of these measures to changes in the number of
firms and market shares.
Other Independent Variables
Aside from industry concentration, the functional specification for SCP includes barriers
to entry as an explanatory variable for performance. Barriers of entry can be either
9
(18)
Bain used the eight-firm concentration ratio (CR8) to measure market concentration:
CR8 = s1 + s2 + + s8 = si
(19)
(20)
His analysis indicates that average industry profitability tend to be higher in concentrated
industries.
(21)
where CR4 the four-firm concentration ratio, Di for i=1,2 are dummy variables denoting
different levels of barriers to entry. Bain classified industries into three classes of barriers
of entry by subjectively evaluating factors such as scale economies, product
differentiation and absolute cost advantages. In the study, industries with high barriers of
entry tend to exhibit higher profitability. Bain's studies went on to inspire many
econometric analyses of the SCP in the 1960s and 1970s.
Subsequent Work
One of the most important work during this period is that of Comanor and Wilson (1967)
which became a classic reference for industry-level econometric analysis of the SCP.
The basic econometric specification adopted in the paper was as follows:
= 0 + 1ASR + 2 log(ACR) + 3 log(RGD) + 4 DlOC +
(22)
where:
Another important contribution to the empirical literature on SCP in the 1960s was that of
Collins and Preston (1969) who used an alternative definition of performance, namely,
price-cost margin (PCM) which was defined as:
(23)
10
Also note that Bain used the CR4 in his subsequent work.
10
(24)
GEO is a measure of geographic dispersion derived from the sum of the absolute
differences between the percentage of value of shipments accounted for by
establishments in each region and the percentage of population in that region
In general, the econometric specification for firm-level analysis is similar to industrylevel specification (ala Comanor and Wilson (1967)) except that additional variables are
use do capture the impact of firm characteristics. Such variable include market share (as
distinct from concentration ratio) and firm size (measured in terms of asset size). An
example of such specification can be found Shepherd (1972):
= 0 + 1MS + 2 GROUP + 3 log(SIZE) + 4 ASR + 5 GROWTH +
(25)
where:
Many of empirical studies on SCP conducted in the 1960s and 1970s including those
cited above provided some support for the SCP hypothesis that concentration is a
determinant of profitability.11 However, the significance of concentration is reduced
when barriers to entry variables are included as independent variables. This is because
concentration is correlated to barriers to entry variables such as MES and capital
11
For good surveys of the empirical findings in SCP see Schmalensee (1989), Scherer and Ross
(1990), Hay and Morris (1991), Weiss (1991) and Martin (2002).
11
12
12
(26)
where si is firm i's market share, ( ) is price elasticity of demand, (i) its conjectural
variation and i = dqj . Re-writing the above equation, we obtain the following:15
dqi
(27)
The conduct of the firm in terms of its conjectural variation (dqj / dqi ) can then be
inferred if we are able to obtain estimates of marginal cost (MC(qi)) and the slope of the
industry demand curve (dP / dqj ). Once the value of the conjectural variation can be
estimated, conduct of the firm can be inferred (see Table 1 below).
1 + (dqj / dqi )
0
1
2
(dqj / dqi )
-1
0
1
Lerner Index
0
1/(2)
1/
The NEIO approach has been applied to a number of industries. The include:
automobiles, rubber, textile, electrical machinery, tobacco, food processing, banks,
coffee, aluminium, retail gasoline, soft drinks and long-distance telephony. 16 Substantial
market power have been detected in some of the industries studied. Collusive strategies
have also been detected in some cases (e.g. trigger strategies in railroads).
13
Another line of research that examines industry disequilibrium is the literature on firm
turnover and mobility.18 The early theoretical motivation for this line of research inquiry
took the form of the Law of Proportionate Effect (LPE) or Gibrat's Law which states
that the expected value of the increment to a firm's size in each period is proportional to
the current size of the firm.19 The implication of LPE is that the limiting distribution of
firm size is lognormal. The empirical literature on LPE between the 1950s and 1970s
seemed to support Gibrat's Law. However, a few studies in the 1980s suggest that the
distribution of firm size is not lognormal in more complete data sets.20
In relation to the SCP paradigm, the LPE suggest a possible link between concentration
and mobility and turnover. Empirical studies have studied the causality between
concentration and mobility and turnover in both directions.21 The results appear to be
inconclusive and not definitive given the small number of studies. The LPE literature has
also generated studies that incorporates explicit treatment of the process of change - in
particular the interactions between random disturbances and structural factors that include
barriers to entry (e.g. MES and sunk cost).22 An alternative approach to the use of
structural factors is to embed the random processes within an industry life-cycle model
driven by technological change (innovation) and diffusion.23 Again, there have been
relatively few studies based on these approaches - possibly due to the intensive data
requirements of such studies.
There has been some attempts to review the IO literature for developing countries. For
example, Tybout (2000) observes that the highly skewed distribution of firm size in
developing countries is due to the presence of small geographically diffuse markets and
the predominance of consumer goods industry. Interestingly, Tybout's survey of the
empirical literature suggests that manufacturers in developing countries are not
necessarily inefficient. An interesting line of empirical studies on developing countries'
experience is the impact of policy changes (e.g. trade liberalization) on firm performance
in terns of price-costs mark-up and productivity.
2.6 Conclusion
The empirical literature in IO is well established covering a period of at least 50 years.
Within this period, two distinct methodological frameworks for empirical IO can be
discerned, namely the SCP paradigm and the NEIO. The SCP paradigm dominated
empirical IO between the 1950s and early 1980s and had significant influence on
policymaking especially in the area of antitrust. However, since the 1980s mainstream
18
For a recent survey of the literature, see Caves (1998). Caves used the term firm turnover to cover
three processes: births and deaths of firms, variations in sizes and market shares of survivor firms
(mobility) and changes in control of firms.
19
The origins of the Law of Proportionate Effect can be traced back to Robert Gibrat's work in 1931.
See Sutton (1997) and Lipczynski et al (2005), pp.264-273
20
See Cabral and Mata (2003).
21
Caves (1998), p.1964
22
For example, see Sutton (1997).
23
For example, Klepper (1996).
14
IO economists have carried out less empirical studies to test SCP. Even though the NEIO
has dominated the empirical IO literature in the past two decades, the number of NEIO
studies is far less that the number of SCP-based studies that have been carried out thus
far. It should also be noted that the SCP paradigm has evolved or led to new areas of
research. These include investigations on stochastic dynamic models (LPE, Lifecycle
models). Much of the literature on developing countries's experiences continues to be
based on the SCP paradigm. In comparison, NEIO and new variants of LPE have not had
much impact in this area. This is likely to be rectified in the future as the quality of data
in developing countries improves.
References
Bain, J. (1951). Relation of Profit Rate to Concentration: American Manufacturing,19361940. Quarterly Journal of Economics, 65(3), 293324.
Bain, J. (1954). Economies of Scale, Concentration, and the Condition of Entry in
Twenty Manufacturing Industries. American Economic Review, 44(1), 1539.
Bain, J. (1956). Barriers to New Competition. Harvard University Press, Cambridge, MA.
Bresnahan, T. (1989). Empirical Studies of Industries with Market Power. In R.
Schmalensee and R. Willig, editors, Handbook of Industrial Organization,
Amsterdam. North-Holland.
Bresnahan, T. and Schmalensee, R. (1987). The Empirical Renaissance in Industrial
Economics: An Overview. Journal of Industrial Economics, 35(4), 371378.
Brozen, Y. (1971). Concentration and Structural and Marker Disequilibria. Antitrust
Bulletin, 16, 244248.
Cabral, L. (2000). Introduction to Industrial Organization. MIT Press, Cambridge, MA.
Cabral, L. and Mata, J. (2003). On the Evolution of the Firm Size Distribution: Facts and
Theory. American Economic Review, 93(4), 10751090.
Caves, R. (1998). Industrial Organization and New Findings on the Turnover and
Mobility of Firms. Journal of Economic Literature, 36(4), 19471982.
Church, J. and Ware, R. (2000). Industrial Organization : A Strategic Approach. Irwin
McGraw-Hill, Boston.
Collins, N. and Preston, L. (1969). Price-Cost Margins and Industry Structure. Review of
Economics and Statistics, 51(3), 271286.
15
16
Supply
Raw Materials
Technology
Unionization
Product Durability
Value/Weight Ratio
Business Attitudes
Legal Framework
Demand
Price Elasticities
Substitutes
Rate of Growth
Cyclical and
Seasonal character
Purchase Method
Marketing Type
Market Structure
Number of sellers
Product Differentation
Barriers to Entry
Cost Structures
Vertical Integration
Diversification
Conduct
Pricing Behavior
Product Strategy and
Advertising
Research and Innovation
Plant Investment
Public Policy
Taxes and Subsidies
International Trade Rules
Regulation
Price Controls
Antitrusts
Information Provision
Performance
Production and Allocative Efficiency
Progress
Full Employment
Equity
17