Measuring Competition Using Lower Bound Concentration Curve:

the Case of the Indonesian Manufacturing Industry

Achmad Shauki*

Centre for Regulation and Market Analysis

School of Commerce, Division of Business
University of South Australia

Phone: 61 8 8302 7043

Fax: 61 8 8302 7001
Email: [email protected]

* I am grateful to Professor David Round and DR. John Wilson for their excellent supervisions,
constructive comments, and helpful suggestions. I would like to extend my appreciation to the School
of Commerce and the Centre for Regulation and Market Analysis of the University of South Australia
for providing the resources. All errors are my sole responsibility.
 Preface

Title of Thesis : Competition, trade and industrial policy and rent seeking behaviour: the
case of Indonesian manufacturing industry.

Supervisors : Professor David Round and DR. John Wilson

The Indonesian trade liberalization and deregulation in nineteen eighties dubbed by as may as
a success story. One aspect of these policies that has not been widely studied is their
implications on domestic competition. The first part of my theses is to estimate and measure
domestic competition and identify how trade liberalization, competition policy, entry
regulation, and rent seeking affect domestic competition, using aggregate industry data. The
lower bound concentration curve based on stochastic frontier function is used for this
objective. In the second part, a model or rent seeking behaviour will be developed and an
empirical examination of this model will be carried out.
I. Introduction

Does trade liberalization promote domestic competition? The conventional wisdom

has been that a wider exposure to international trade will increase competitive pressure on
domestic oligopolies and thus have the effect of reducing their market power. However, the
effectiveness of trade liberalization in disciplining domestic competition depends on a number
of factors. It is determined by how the trade liberalization policy is designed, by the reaction
of imports to domestic collusion, and on the response of domestic oligopoly to the policy. The
Indonesian trade liberalization in the nineteen eighties is a good case to test these aspects of
trade liberalization given the uniqueness of Indonesia’s trade liberalization. The emphasis of
the Indonesian trade liberalization was more on promoting exports rather than promoting
efficiency in resource allocation. One aspect of Indonesia’s trade liberalization of particular
interest is the extent of rent seeking activities in influencing trade policies in Indonesia.

The first step in analysing the competition effect of a policy is to measure the
intensity of competition. The common method used in measuring the intensity of competition
is the price-cost margin or some proxy for this measure. A large gap between price and cost
indicates that firms in the market have significant power to increase price above marginal
costs without provoking entry. The implication is that variations in price cost margins across
industry can be used as an indicator of variations in market power and thus, on the variations
of the intensity of competition across industry. The variation is influenced by a number of
factors, among them public policy.

An important criticism of the SCP approach is that it ignores specific industry

characteristics. The type of firm interaction which prevails in each industry and the
characteristic of products in that industry determine the equilibrium for that industry. For this
reason critics of the traditional SCP approach maintain that the predictions of this approach
lack precision.

A more recent development in industry analysis due to Sutton (1991) has allowed a
new method of measuring competition. The model looks at the statistical regularities that
apply across industry and at the same time is robust across different assumptions about price
competition among firms. In particular it looks at the relationship between market
concentration and market size under the assumption of free entry. This relationship
establishes a lower bound concentration curve which is a benchmark from which competition
is measured. Deviation from the lower bound indicates that the entry condition facing the
industry deviates from the free entry conditions of the benchmark. Once the deviations are

1
identified and measured they can be used to explain variation in the degree of competition
across industries resulting from entry barriers.

The lower bound curve can be estimated using the stochastic frontier method
proposed by Battese and Coelli (1995). The advantage of this method is that it allows us to
estimate the lower bound curve and simultaneously estimate the factors that cause
concentration to deviate from the lower bound. A stochastic lower bound concentration curve
is estimated using observed structural variables such as the size of exogenous sunk costs and
minimum efficient scale, whereas the entry condition function is estimated using a proxy for
government intervention such as the role of state enterprises, a proxy for strategic behaviour
such as the extent of rent seeking activities, and the pressure from international trade.

The contribution of this paper is two-fold. It provides an explanation of the dynamics

of market structure in Indonesia during the trade liberalization era, and an understanding of
the competition aspect of the liberalization policy. The paper will be organized in the
following manner. A brief literature review will be provided in the next section, followed by a
discussion of theoretical background on the lower bound concentration curve. The empirical
model of a lower a bound curve will be provided in the section that follows. The next two
sections discuss the empirical model of the lower bound function and the estimation method
respectively. This is followed by a discussion of the empirical specification of the lower
concentration function. Discussion of data sources and the empirical result will be given in
the section that follows. A concluding section follows.

II. Brief literature review

Although competition lies at the heart of efficiency in the market economy, the
concept is difficult to define. For this reason, research has tended to focus on the
characteristics of a competitive market and to measure the intensity of competition in a
market by examining deviations from this ideal. One of the most important characteristics of a
competitive market is that firms in this market earn normal profit, which is defined as the
profit high enough for firms to stay in the market but not to encourage entry and exit.
Deviation from normal profit indicates deviation from a competitive market.

Economists are interested in explaining what causes the market to deviate from its
competitive level as it entails welfare loss to society. The methodology that is commonly used
to explain the intensity of competition is the structure-conduct-performance (Collins and
Preston, 1969; Strickland and Weiss, 1976; Martin, 1988; Kalirajan, 1993; Bhattacharya and
Bloch, 1997; Bhattacharya, 2002) approach which was first popularized by (Bain 1956). The

2
approach is based on the notion that a proxy for competition such as market structure will
determine the type of interactions between firm and the result of these interactions will be
shown in the industry performance. Price competition is implicitly imbedded in the market
structure or explicitly in observed behaviour variable such as advertising expenditure.

An important determinant of competition is exposure to international trade. One of

the early studies incorporating trade as a determinant of competition was Esposito and
Esposito (1971) who maintained that foreign competitors, due to their large number, can pose
a bigger threat to the ability of domestic firms to raise price above average costs compared to
domestic competitors. Empirical studies along this line of reasoning tend to support this
hypothesis (Domowitz, Hubbard et al. 1986; Domowitz, Hubbard et al. 1986; Katics and
Petersen 1994). However, incorporating a measure of trade exposure as a determinant of
profitability disregards the competitive structure of the domestic market. If the domestic
market is competitive to begin with, introducing imports will have little effect on the
profitability of industry. Imports may push inefficient domestic firms out of the market but
will have little effect on the profitability of efficient domestic firms. Therefore the
competitive impact of imports has to be discussed in the context of non-competitive domestic
producers (Geroski and Jacquemin 1981). The empirical implication is that the effect of
imports can only be measured if imports are introduced into the model as an interaction with
market concentration. Studies along this line of argument tend to support the import-as-
market-discipline proposition although there are mixed results as to the strength of the pro-
competitive force of imports. Using data from Chile, Melo and Urata (1986) were not able to
show convincingly that increase in import reduces profitability although they did show that
the effect of market concentration falls with the increase in import. More conclusive results
were obtained by Tybout (1996) using the Chilean data and by Grether (1996) using Mexican
data, although in both cases the results are sensitive to the empirical specification of the
model. Ghellinck et al. (1988) modify the SCP approach to allow sufficient flexibility to
absorb industry-specific features and found support of the import-as-market-discipline
hypothesis.

The SCP approach has been subject to much scrutiny. Researchers working in the
tradition of the Chicago school of industrial economics maintain that high profitability in a
concentrated industry has to be interpreted as a reflection of superior efficiency of larger firms
over smaller ones, not as a reflection of market power. One of the most influential researchers
in this tradition is Demsetz (1973) who argued that if a high concentration industry enjoys
high profitability then the profitability of large firms in that industry should not be different
from the profitability of small firms. He found that that the profitability of larger firms was in

3
fact higher than smaller firms which he regarded as proof that larger firms were more efficient
than smaller firms.

Another criticism of the SCP approach is that the cost data used in the empirical
analysis do not reflect economic costs and thus provide a biased estimate of market
performance. Researchers in the new industrial economics school overcome this problem by
avoiding the use of cost data and instead they estimate the price to marginal cost ratio. The
approach, which draws from the work of Hall (1988), is based on the notion that a firm’s
profit maximizing choice of input and output is determined by output price, which in turn is
determined by the market structure. Therefore, if data on a firm’s output and inputs is
available it is possible to estimate the price-cost ratio that the data implicitly represent.
Levinsohn (1993) used this method to estimate the price-cost ratio of the manufacturing
sector in Turkey and found that the price cost ratio declined after trade liberalization. Konings
et al. (2001) used it to test the impact of competition policy on performance in Belgium’s and
the Netherland’s manufacturing industries. The model was also used by Kee and Hoekman
(2007) to test the effect of competition law on domestic competition using cross country data
of 28 industries in 42 developed and developing countries.

The SCP approach is also criticized for its assumption of a one-way relationship
between market structure, conduct, and market performance. It is argued that if market
concentration contributes to a firm’s profitability then there will be an incentive for firms to
engage in strategic actions to reduce the number of existing firm and to prevent potential
competitors from entering the market. Furthermore, if import competition has the effect of
reducing profitability there is also an incentive for domestic firms, especially those in high
concentration industry, to engage in rent seeking activities to influence trade policy or to
request compensation. The endogeneity problem is usually perceived as an econometric
problem which can be resolved using instrumental variables or a simultaneous equation
approach. However, Schmalensee (1989) maintains that in a cross industry analysis
essentially all variables are endogenous such that no exogenous variables can be used for
instruments.

A more recent approach to industry analysis, due to Sutton (1991), provides an

alternative method of estimating competition. In this model, Sutton used game-theory to
explain firms’ profit maximizing behaviour when there are sunk costs associated with
production. He distinguished two types of sunk costs: the exogenous sunk costs needed to set
up a plant of minimum efficient scale and the endogenous sunk costs which cover the cost of
advertising and R&D. He showed that when entry and exit are assumed to be free, the
equilibrium number of firms, and thus market concentration, depends on the market size, the

4
magnitude of sunk costs, and the intensity of price competition. Given the intensity of price
competition, the model predicts that there is a lower bound to concentration which decline to
zero as the market size increase towards infinity. 1 The empirical implication is that we would
observe market concentration of similar industries to be lower in countries that has larger
market size. The requirement that the industries being compared are similar is to ensure that
the relationship between market size and market concentration being compared is estimated
based on industries that have similar exogenous sunk costs and similar type of price
competition imbedded in them. The model also predicts that an increase in the intensity of
price competition will lead some firms to exit and others to merge so that the level of
concentration tends to increase.

Sutton maintained that the negative relationship market size and market
concentration may fail to hold in endogenous sunk costs industries. He argued that in
industries where the bulk of the sunk costs are advertising and R&D expenditures, the lower
bound concentration does not converge to zero as market size increases. In this model, it is
assumed that advertising is used to influence consumer’s willingness to pay a higher price.
The model shows that as the market size increases, firms will invest a larger amount of
resources in advertising or R&D, thus increasing the level of sunk costs. As the sunk costs
increases, the minimum efficient scale will increase and there will be less incentive for firms
to enter even though market size has increased. The consequence is that the market
concentration does not necessarily decrease with an increase in market size.

The model predicts that actual market concentration may not attain the lower bound
concentration if entry and exit is impeded. Sutton claimed that the observed high market
concentration in countries with relatively large market sizes is due to the entry condition in
those countries. He identified a first mover advantage as one factor that inhibits entry in these
countries. Since the presence of entry barriers will push the actual concentration level away
from the lower bound, the deviation of the actual concentration level from the lower bound
curve can be used as a measure of the entry condition and therefore, as a measure of the lack
of competition arising from entry conditions. In this sense the lower bound concentration can
be used as the benchmark from which the intensity of competition can be measured.

The lower bound concentration model has been successfully used to explain the
competitive effect of market expansion and of changes in policy regimes. Lyons et al. (2001)
used the model to analyse the effect of European market integration on the intensity of
competition in the European Union. The model has also been used to analyse the effect of
regime change in competition policy. Symeonidis (2000) used the model to explain why
1
The price competition can be Cournot, Bertrand, or collusive.

5
concentration ratios in U.K manufacturing industries increased following the imposition of
the 1956 Restrictive Trade Practices Act which abolished cartels. Driffield (2001) used the
model to test the hypothesis regarding the competitive effect of foreign direct investment in
the U.K. He showed that larger foreign direct investment in the UK had pushed the
concentration ratio closer to the lower bound, indicating that foreign direct investment had a
positive effect on domestic price competition.

III. Lower bound concentration as measure of competition

Sutton (1991) models the relationship between market size, sunk cost, price
competition and market concentration in the following manner. He assumes that entry and
exit are totally free and that firms make decision in a multi-stage game. In the first stage firms
decide whether or not to enter the market given the size of entry costs (which is treated as
sunk cost), the market size, and the knowledge of the intensity of price competition in the next
stage. In the second stage firms that decided to enter determine the level of output that will
maximize their profit. In industries with endogenous sunk costs the game is played in three
stages. In this industry firms choose the level of endogenous sunk cost in the second stage and
determine profit maximizing output in the third stage. The number of firms entering in the
first stage is solved by backward induction.

Consider the exogenous sunk costs industry. Let there be N firms that decide to enter
the first stage. In the second stage, firms that decided to enter determine their profit
maximizing output and price. The equilibrium outcome in the second stage can be represented
by the gross profit function Π (Ν, S,α), where S and α are shift parameters indexing market
size and the “intensity of competition”. The intensity of competition captures the pricing
strategy of firms, which will in turn be influenced by institutional factors such as trade and
competition policies and the nature of the goods being produced. 2 The model assumes that the
gross profit declines with the number of firms ( Π N 〈0 ), increases with market size ( Π S 〉0 ),

and decreases with the intensity of price competition ( Π α 〈0 ).

The number of firm entering in the first stage (market concentration) can be solved by
backward induction. Assuming symmetric firms, the number of firm can be solved from the
free entry condition:

Π (Ν ,S,α) = F (1)

2
For example, homogenous goods tend to have tougher price competition as they have a high elasticity
of demand.

6
Where F is the size of exogenous sunk cost. Given the size of sunk cost and the assumption
about the toughness of competition, let the equilibrium number of firms entering in the first
stage be N*. From equation 1, a functional relationship between the market size and the
equilibrium number of firms can be established. If the market concentration is defined as 1/N*
the relationship between market size and market concentration curve can be written as

C = 1 / N* = f ( S | α , F ) (2)

Sutton maintains that the relationship occurs at the lower bound since a market
concentration that is lower than the lower bound (N < N*) implies negative profit for every
firm and therefore cannot be supported as an equilibrium. The properties of the gross profit
function imply that an increase in market size, ceteris paribus, will have to be matched by an
increase in the number firms (a fall in market concentration) for the equilibrium condition to
hold. Therefore, in equilibrium there will be a negative relationship between market size and
market concentration.

Using the equilibrium condition 1, the following properties of the lower bound
concentration can be established. First, an increase in the intensity of price competition (α)
will reduce market concentration, since an increase in the intensity of price competition,
ceteris paribus, will reduce profit and therefore be matched by a fall in the number of firms to
keep the equilibrium condition intact. A consequence of this property is that a change in trade
and competition policy regime that increases the price competition should increase the lower
bound concentration.

The second property is that an increase in the magnitude of entry cost (sunk cost) F
will increase concentration. An increase in the sunk cost, ceteris paribus, will reduce gross
profit and therefore will have to be matched by a fall in the number of firm (an increase in
market concentration) to keep the equilibrium. It follows from this property that an increase in
the minimum efficient scale due to a larger setup costs (sunk costs) will increase the lower
bound concentration.

The above predictions do not always apply to industries with large advertising and
R&D sunk costs. These two sunk costs are endogenously determined because an increase in
market size will push firms to increase their advertising and R&D expenditures in order to
capture a larger share of the market. Since higher sunk costs imply higher entry costs, the
implication is that entry will not necessarily occur when market size increases. As long as the
increase in profit due to the increase in market size is relatively larger than the increase in
advertising costs, firms will enter in response to an increase in market size. The implication is

7
that a negative relationship between market size and market concentration will prevail. This
situation will generally occur when the market size is relatively small. However, as the market
becomes relatively large the endogenous sunk costs effect will eventually take over the
market size effect, and the negative relationship between market concentration and market
size will then cease.

The lower bound concentration can be perceived as the benchmark concentration. It is

the market concentration that will prevail under free entry. The actual market concentration
may deviate from the lower bound due to entry restrictions. These restrictions may be created
by entry-deterring strategic behaviour such as investment in excess capacity (Spence 1977;
Dixit 1980), raising rivals’ costs (Salop and Scheffman 1983; Salop and Scheffman 1987),
and predatory pricing. The restriction may also be created directly by governments through
legal entry restriction, or indirectly as a result of trade and industrial policy. These policies
may arise due to market failure consideration, but it can also arise as result of rent seeking
activities (Salop, Scheffman et al. 1984; Grossman and Helpman 1994; Hillman, Long et al.
2001). In addition, in countries where state enterprises play an important role in the economy,
the entry barriers may be erected as a policy to protect these enterprises.

In developing countries entry barriers may also be influenced by the presence of

multinational corporations. There are two ways that multinational corporations can influence
domestic entry conditions. First, competition among developing countries for foreign direct
investment may lead these countries to offer market protection as an incentive. Similarly,
multinationals may require some kind of market protection in return for their investment in
these countries. In this respect the presence of multinationals tends to be associated with
higher entry barriers. On the other hand, the presence of multinationals can discipline the anti-
competitive conduct of domestic oligopolies. In this respect, multinationals tend to reduce
domestic entry barriers (Driffield 2001).

The deviation of the actual concentration from the lower bound can be used as a
measure of competition which results from the entry conditions. If the deviations can be
estimated, we can identify the factors that determine their variability across industries and
across time. In particular, we would be interested in looking at how rent seeking activities,
government protection for state enterprises, and strategic behaviour affects entry conditions
and competition across industries and across time.

The model allows us to separate market concentration into elements that are
determined by structural variables such as market size, the magnitude of sunk costs, and the
competition policy regime, and elements that are determined by entry conditions. The former

8
determines the lower bound concentration and the latter determines the deviation of the actual
concentration from lower bound. The sunk cost model of market concentration determination
can be expressed in the following general form:

CR = f (N,F,α) + g(z) (3)

The first term in the right hand side of equation 3 is the lower bound function that is
determined by structural variables. The second term is the entry condition which is
determined by a vector of independent variable z.

IV. Empirical model for lower bound function

Sutton (1991) used a frontier model to estimate the lower bound concentration curve.
The frontier model has been widely used to analyse production and cost function (for a good
review of the frontier function see Kumbhakar and Lovell, 2000). Development of the
econometric method of estimating frontier functions has been presented almost entirely in the
context of efficiency in production and cost. Aigner and Chu (1968) initially introduced the
model as a deterministic model where the frontier curve is determined entirely by a set of
predetermined independent variables. A stochastic model of the frontier function was
introduced simultaneously by Aigner et al. (1977) and Meeusen and Broeck (1977). The
stochastic frontier model allows the frontier curve to be determined by a random variable in
addition to the predetermined independent variables.

The stochastic frontier model has the following general form :

y it = f ( xitk ; β ) ⋅ exp( vit ) exp( u it ) (4)

where i is index of observation, t is time index, and k is index of the independent variables.
The variation of y across observations is influenced by two elements: the first element is the
deterministic function f(xk;β) which is determined by the magnitude of a set of predetermined
variables xk and the unknown parameters β=(β0, β1,...,βk) which can be statistically estimated.
The parameters reflect the effect of the independent variables xks on the dependent variable y
and are common to all observations. The second element is the stochastic component which
consists of exp(vit) and exp(uit) and is assumed to be specific to observation i at time t. The
random variable vit accounts for measurement errors and other random factors, whereas uit

9
measures the deviation of the actual dependent variable y from its frontier. 3 The deviation
ratio can be written as the following

DRit = exp( uit ) =

f ( xit ; β ) ⋅ exp( vti )
yit

A log linear specification of the stochastic frontier model can be written as

ln( yi ) = β 0 + ∑β ln( x ji ) + vi + ui
k
(5)
j =1
j

The term ui is non-negative since the frontier function is the lowest attainable value of y for
observation i given the values of the independent variables for that observation. Unlike ui the
error term vi is not restricted to be non negative. Assumption regarding the distribution of uit
and vit will be given in the next section.

The model described by equation 5 can be estimated using cross-section data, time
series data, or panel data. Sutton (1991) conducted a cross-country study of the same (similar)
group of industries to estimate the lower bound concentration function. He used similar
industries because technology and the pricing behaviour in these industries were expected to
be the same across countries, such that the differences in the long run concentration level
across countries can be explained by differences in market size. Therefore, the concentration
curve estimated using this data should be the lower bound concentration curve for these
industries. Sutton asserted that the deviation of each country’s actual concentration from the
estimated lower bound concentration showed the entry condition in that country. In a country
that had a strict competition policy on entry deterrence activities, the market concentration
level will be very close to the lower bound, and vice versa. Similarly, in a country where
there was a first mover in the industry, or where there were many legal entry restrictions, the
market concentration will deviate further away from the estimated lower bound concentration.

When the lower bound concentration is estimated using cross industry data the term ui
measures the differences in competition intensity across industry. The differences in
competition intensity result from differences in entry conditions across industry. In an
industry where there are no legal barriers to entry, where government imposes a strict
competition policy on strategic entry deterrence activities, and where government adopts a
free trade regime, the level of concentration will be very close to the industry’s estimated
lower bound concentration. On the other hand, in industries where there are many entry

3
In production analysis uit measures the technical inefficiency.

10
restrictions, which may result from strategic behaviour or from government regulation, the
level of concentration will fall far from the estimated lower bound concentration. Therefore,
deviation from the lower bound concentration captures the height of entry barriers and its
implication on competition. The variations of this “deviation” across industry are determined
by factors such as competition policy, trade policy, and the presence of first mover in the
industry. When the lower bound function is estimated using time series data for one industry,
the deviation of actual concentration from the lower bound concentration at each period may
indicate the government entry regulation policies toward the industry for that particular
period.

With regard to Indonesian manufacturing industries, trade policies differ across

industries and across time. For example, the government adopted a protectionist policy before
1987 but the degree protection varied across industries. Between 1987 and 1995 trade was
liberalized but the speed and the degree of liberalization in each period during this time
varied, with most of the liberalization policies occurring between 1989 and 1990. Because of
these features a cross section or a time series analysis will be inadequate to estimate the lower
bound concentration curve. For this reason panel data for the manufacturing industries will be
used to estimate the lower bound concentration.

V. The method of estimating the stochastic frontier function and the deviation
index.

Battese and Coelli (1992) proposed using maximum likelihood to estimate equation 5.
They assumed the noise component vit to be independently and identically distributed normal
with zero mean and variance σv2, that is, it is iid ~ N(0,σv2). It is also assumed to be
independently distributed of the second random variable uit. The latter is assumed to be non-
negative and independently and identically distributed (iid) truncated normal distribution with
mean μ and variance of σu2 (iid ~ N(μ,σu2)) 4 . We assume ui to be systematically influenced by
a set of exogenous variables such as government policies, rent seeking activities, and other
external factors. If we assume the set of explanatory variables are the vector Z, the component
uit can be written as

u it = Zδ + wit (7)

where δ is column vector of unknown scalars and wit is the unexplained component of the
deviation which has an assumed distribution. One method that has been used in the literature

4
Other assumptions regarding the distribution of ui that have been used are exponential distribution and
gamma distribution (see Reifschneider and Stevenson, 1991)

11
to estimate equation 7 (e.g Pitt and Lee, 1981) is a two-stage approach: in the first stage uit is
estimated based on the stochastic frontier function, and in the second stage the estimated uit is
regressed on the set of explanatory variables using OLS. The problem with this method is that
it produces inconsistent estimates of the δs. To overcome this problem Battese and Coelli
(1995) proposed estimating equation 7 and the stochastic frontier equation 5 simultaneously.

Consider the error term uit. Since this term is assumed to be non negative the
unexplained component wit in equation 7 will have the following property:

wit ≥ - Zit δt (8)

To accommodate this property Battese and Coelli assume that wit is distributed as a truncated
normal distribution with zero mean and variance σ w2 , such that the truncation occurs at wit = -

Zitδ. This ensures that the assumption about the distribution of wit is consistent with the
assumptions that uit is a non-negative truncation of N(Zitδ, συ2) .

Once the parameters of the frontier function have been estimated the next step is to
calculate the value of uit from the equation. The deviation of the observed values of the
variable yit from their frontier value can be expressed as the ratio of these two values

exp( xit β + vit )

DRit = = = exp( −uit )
yit exp( xit β + vit + uit )
ŷit
(11)

If the industry has achieved the lower bound concentration then DRit=1. On the other hand
DRit <1 implies that there are entry barriers that inhibit the concentration ratio from moving
toward its lower bound level. The larger the entry barriers the larger will the value of DR be.

VI. Empirical specification for the lower bound concentration model

The first challenge in any empirical work on market structure and competition is to
find the empirical counterpart of the theoretical concepts. Theories seldom give sufficient
guidelines on how concepts should be measured empirically. In the case of market structure
the simplest definition is the number of firms in the market. In empirical work, this concept is
usually translated into the size distribution of firms, measured as seller concentration. The
four measures of size distribution that are widely used in empirical work are: the k-th firm
concentration ratio, the Hirschman-Herfindahl index, the entropy index, and the Gini
coefficient index. Each measurement has its own advantages and disadvantages such that the
decision to choose one measurement over the other is basically determined by its relevance to

12
the objective of the research. For this research we will use the Hirschman-Herfindahl index to
take into account the role of smaller firms in the concentration measure. Furthermore, it is
shown that the results are consistent when a four-firm concentration ratio is used in place of
the Hirschman-Herfindahl index. Both concentration indexes are measured using production
data.

One problem of using either the Herfindahl index as measure of concentration or the
k-th firm concentration ratio is that their values are constrained from above and below. The
Herfindahl index is constrained by zero from below and by ten thousand from above, whereas
the k-th firm concentration ratio has to fall between zero and one. As long as the main
purpose of the study is to estimate the regression parameters this problem is not a big issue. It
becomes an issue when the result is used to estimate the concentration level. In this case there
is no guarantee that the estimated concentration will fall between the upper and lower limits.
To overcome this problem the following log transformation of concentration is used

⎡ ⎤
LHHI it = ln ⎢ ⎥
HHI it
⎣ 10000 − HHI it ⎦
(12)

This transformation will ensure that the concentration measure will always fall between the
upper and lower limit.

The second issue is whether the concentration measures need to be adjusted for
foreign trade. This issue is particularly relevant for Indonesia since it is an open economy and
the period of study involves a period when exports and imports increased sharply due to trade
liberalization. The trade adjusted concentration is obtained using the following formula:

HHI it ( Oit − X it )
HHIADJ it =
( Oit − X it + M it )
(13)

Where O is the total production of industry i at time t, X is the exports of industry i at time t,
and M is imports of industry i at time t.

As explained in the previous section, the structural variables that are of interest to this
model are the size of market, the size of set-up costs (sunk costs), and the toughness of price
competition. In addition we will also include the minimum efficient scale to take into account
the effect of economies of scale. Previous empirical studies have shown that these four
variables are generally statistically related to market concentration, whether the latter is
measured by the four-firm concentration ratio or by the Hirschman-Herfindah index (Curry
and George 1983).

13
 In this study market size will be measured by industry output deflated by the
producer’s price index. However, since the estimation will utilize panel data there is an issue
whether the relationship between concentration and market size is contemporaneous. We
assume that entry does not materialize immediately following a demand shock since it takes
time for new firms to be established. This is particularly true in Indonesia, where new
investment in some sectors of the economy has to be approved by the government. We found
that output with a one year lag has the highest correlation with current market concentration.
For this reason market size enters the model with a one year lag.

The second structural variable is the setup cost. It is the cost associated with acquiring
a plant of minimum efficient scale and it is assumed to be exogenously determined. Theory
predicts a positive relationship between lower bound concentration and the setup costs. To
obtain a proxy for setup costs we use the industry’s aggregate market value of fixed capital.
The value of the industry’s total fixed capital is divided by the number of firms, assuming that
each firm is operating at the minimum efficient scale, to arrive at the costs of setting up a
single firm. In this paper the minimum efficient scale will be measured in terms of labour
employed. The number of firms with minimum efficient scale can be obtained by dividing the
total number of workers employed in the industry by the minimum efficient scale. For this
purpose, the minimum efficient scale has to be estimated first.

There are three approaches that are generally used in the literature to derive the
minimum efficient scale (Waterson 1984): the technological or engineering costs study, the
survivor technique, and the first moment distribution of firm size. The engineering cost
approach would require a survey that is beyond the scope of this research so only the latter
two will be considered. Conceptually the survivor technique is superior to the first moment
distribution approach. The survivor technique identifies the level of output that most firms
gravitate to over a period of time. This output is the minimum efficient scale since firms can
only survive if they operate with this scale in the long run. The problem with the survivor
technique is that it assumes that there is no technology change affecting minimum efficient
scale. This assumption may not be tenable in developing countries undergoing trade
liberalization. Trade liberalization and foreign direct investment generally exposes these
countries to newer and better technology (Tybout 2000), which in turn may change the
minimum efficient scale. Due to these issues the first moment distribution method is
employed to measure minimum efficient scale. Labour is used as the measure of scale and
minimum efficient scale is the number of workers used by firm of median scale. The number
of firm with the minimum efficient scale is obtained by dividing the total number of workers
employed in the industry to the number of workers employed by the median firm.

14
 The minimum efficient scale is also included as one of the independent variables in
the model. It is included to control for variations in the economies of large scale across
industries and across time. However, as economies of scale may arise from high setup costs
there is a possibility that the minimum efficient scale variable is correlated with setup costs.
In fact, when minimum efficient scale is measured by the number of workers employed we
found a strong statistical correlation between this variable and the setup cost variable such
that it creates a multicollinearity problem in the regression. To overcome this problem we
used a more practical approach to measure minimum efficient scale. It was measured by the
relative efficiency of the large firms, defined as the ratio of output (in real terms) per unit of
labour of the largest four firms relative to the output per unit of labour of the median firm.
The method, which is used by Davies et al. 1988, Davies and Geroski (1997) and by
Bhattacharya and Bloch (2000), is based on the idea that if economies of scale exist the
efficiency of large firms will be higher than small firm. The larger the minimum efficient
scale, the higher the relative efficiency of the large firm.

Intensity of competition should capture two aspects of competition. The first aspect is
the type of firm interaction that exists in the industry, that is, whether firms choose price or
quantity as the strategic variable, and their conjectural variation. This aspect of competition
can be measured by the degree of product homogeneity on the assumption that producers of
homogenous goods face tougher price competition because of the high degree of
substitutability of their products, whereas differentiated goods tend to have softer price
competition. Following Bhattacharya and Bloch (2000) the degree of homogeneity is
measured by the proportion of output that is sold as intermediate input. This is based on the
idea that intermediate goods tend to be more homogenous, whereas consumer goods tend to
be differentiated. Since tougher competition implies lower profit, there will be fewer firms in
the market and therefore higher concentration. For this reason we would expect a positive
relationship between this variable and the lower bound concentration.

The second aspect of competition is the general competition environment associated

with government policy toward competition. The two elements of this aspect that are of
interest are the effect of competition law and the effect of the trade and industrial policy
regime on domestic competition. The first element is not relevant to this paper since
competition law was not introduced in Indonesia until 2001. The second element is included
because the imports-as-market-discipline hypothesis predicts that trade liberalization will
have a disciplining impact on anti-competitive behaviour.

The effect of trade and trade policy on competition will be measured in the following
manner. Trade policy regime is measured is by introducing a time dummy variable that

15
separates the period before and after the trade and industrial policy change. The dummy
variable will measure whether there is a shift in the general competitive environment
following a trade regime change. If the import-as-market-discipline hypothesis holds, that is if
price competition increases following trade liberalization, we expect an upward shift in the
lower bound concentration. The effect of trade, particularly imports, at the industrial level is
measured by the level of the effective rate of protection. Technically, the effective rate of
protection measures the percentage of an industry’s value added under trade restrictions
relative to the value added obtained under free trade, and it is calculated based on the input-
output table. Calculations of effective rate of protection for Indonesia for 1987, 1989, and
1995 are available from Fane and Phillips (1991), Wymenga (1991), and Fane and Condon
(1996). The advantage of these three calculations is that they factor in non-trade barriers in
their calculation, which give them a more comprehensive measure of effective rate protection.
The other advantage is that they are calculated using the same methodology which allows
them to be compared to each other.

Since the effective rate of protection is available only for these three years whereas
the analysis uses the period from 1987 to 1995, the effective rate of protection for the
remaining years have to be estimated. For this purpose two possible methods of estimation are
considered. The first method assumes that the effective rate of protections for 1988 are the
same as that of 1987, the effective rate of protections from 1989 to 1994 are the same as 1989,
and for 1995 and 1996 they are the same as the effective rate of protection in 1995. The
second method assumes that the effective rate of protection are different in each year and the
effective rate of protection for periods other than the three years where the figures are
available are obtained by extrapolation. The effective rate of protection for 1988 is obtained
by extrapolation using 1987 and 1989 figures, whereas the effective rate of protection figures
between 1989 and 1995 are obtained by extrapolation using the effective rate of protection in
1989 and 1995. For 1996 we assume that the effective rate of protection is equal to that of
1995. It is decided to choose the second method on the assumption that trade liberalization in
Indonesia occur gradually and therefore, the effective rate of protection also declines
gradually.

The import-as-market-discipline hypothesis predicts that there is positive relationship

between the effective rate of protection and the lower bound concentration since a higher
protection will shield domestic collusive behaviour from import competition. However, we
allow the possibility that imports join the domestic collusion rather than challenge it, or act as
price followers with respect to domestic oligopolist (Urata 1984; Globerman 1990; Feinberg
1996). Under this circumstance we will expect an insignificant or even a negative relationship
between ERP and lower bound concentration.

16
 Although there are strong theories to support the relationship between the lower
bound concentration and the six independent variables mentioned above, there is very little
guide on the functional form of this relationship. A linear relationship between concentration
and the independent variables is assumed in Geroski et al. (1987), Geroski (1991)
Bhattacharya and Bloch (2000), and Bhattacharya (2002); whereas a log linear relationship is
used by Sutton (1991), Symmeonidis (2000), and Lyons et al (2001) for the lower bound
relationship. The latter three use a log linear relationship since the logistic transformation of
the concentration ratio is used as measure of concentration instead of the actual value. For this
paper we follow Sutton’s specification of log linear relationship.

Given the above determinants of lower bound market structure we can write the lower
bound concentration function as

LCR it = β 0 + β 1 DTRDRGM + β 2 LMKSZ it + β 3 LMES + β 4 LSETUPit + β 5 ERPit + β 6 PDRit + vit + u it

(14)

where: i = industry index; t = index of time

LCR = logistic transformation of concentration ratio
DTRDRGM = dummy trade regime (1=after 1989)
LMKSZ = natural log of market size
LMES = natural log minimum efficient scale
LSETUP = natural log setup costs
ERP = effective rate of potection
LPDR = natural log of proportion of output sold as intermediate input

The expected signs of the coefficients are given in Table 1 below.

Table 1 Expected sign of regression

Variable DTRDRGM LMKSZ LMES LSETUP ERP PDR
Expected sign +/− − + + +/− +

The second part of the model is the entry condition element of market concentration.
This is measured by the deviation of actual concentration from the lower bound concentration.
The actual concentration ratio of the i-th industry at the t-th period can deviate from the lower
bound depending on the degree of entry barriers that existed in each industry at a different
time. Entry barriers due to strategic behaviour, a first mover advantage, a monopoly licence,
and government privileges can lead market concentration to deviate from the lower bound.
The deviation is measured by the random factor uit, which is assumed to be systematically
determined by a set of explanatory variables.

17
 The most significant impediment to competition in developing countries is
government-created entry barriers (Singleton 1997). In Indonesia these barriers can be found
for example in the wheat flour industry where only two firms were allowed to operate. The
other government-created barrier is the requirement that all large investments have to be
approved by the government through the Investment Coordination Board. 5 Furthermore, due
to the extensive government role in the economy and an authoritarian political system, entry
can be deterred by influencing government policy. If rent seeking is an effective method for
restricting entry it will have the effect of pushing the actual concentration level away from the
lower bound. However, measuring rent seeking is always the issue in empirical work on this
subject.

In this paper, charity donation expenditures are used as a proxy for rent seeking
activities. Donations and political contributions are identified by Grossman and Helpman
(1994) as the means by which rent seekers influence policy. In the case of Indonesia there is a
possible link between these expenses and the funds raised by charity foundations belonging to
the political elite and policy makers. Charity foundations are popular institutions among
politicians and the ruling elites to generate donations and funds because they have very few
legal restrictions on public accountability, at least during the period of study. Donations will
be a good measure of rent seeking activities in industries where there is a large opportunity
for rent seeking. This will be in industries where government intervention is significant or in
industries that are subject to crony influence. For example, in Indonesia the wood and paper
products industry, the cement industry, and the machinery industry have all been identified as
sector that are sensitive to rent seeking activities (Basri and Hill 1996). Donation expenditures
by large and medium scale manufacturing firms are recorded in the Industrial Statistics
database published by the Indonesian Central Bureau of Statistics.

However, rent seeking is one of many purposes behind donations. The other
objectives of donations include philanthropic objectives, social responsibility, and public
relations. If donations are used for purposes other than rent seeking, then their variation across
firms or across industries should be random and not systematically related to profits or
production. In particular, we will not see a systematic relationship between industry’s total
donation expenditure and the level of entry barriers in that industry. On the other hand, if
donations are used as a rent seeking tool, we will see a positive relationship between the two
variables. Donation expenditures in each industry are measured in real terms (deflated by the
consumer’s price index) per unit of labour employed in the industry.

5
The reason for this regulation is to avoid excess capacity.

18
 The second determinant of the entry condition is foreign investment. Foreign
investment has played a very important role in Indonesia’s manufacturing sector since the late
sixties when the New Order government adopted a very friendly atmosphere for foreign
investment. Although the degree of openness has fluctuated during the thirty years of the New
Order regime, the government’s stance towards foreign investment has always been more
favourable compared to imports. Foreign investors are better equipped to break down
domestic entry barriers compared to new domestic producers, as they usually come with
better technology and efficiency, and greater resources. As a consequence, they have a better
chance of overcoming domestic entry barriers. On the other hand foreign firms may see the
privileges that they get from government incentives as a way to generate monopoly rents from
the local market. Therefore, their presence will not induce more domestic competition but
merely replace the domestic oligopoly. The role of foreign firms is measured by their share in
total industry output. An industry’s foreign firms output is obtained by multiplying the share
of foreign ownership in each firm to the firm’s output and aggregating them for all firms. The
ratio of this output to the total industry output is the share of foreign firms’ output.

Entry restrictions may be imposed by governments to protect state enterprises. State-

owned enterprises play a very important part in the Indonesian economy, not only because
their existence is sanctioned by the constitution, but also because they are used by
government as an “agent” of economic development, playing such role as employment
generator, and as a means for technology transfer. Because of their importance, many state
enterprises are protected by government not only from foreign competition but also from
potential domestic competition. Given this condition it would be reasonable to include the
role of state enterprises in a particular industry as one of the variables determining the ease of
entry. We measure the role of state enterprises by their share in total industry output. This is
obtained in the same manner as the share of foreign firm in total industry output.

The model explaining variation in the degree of competition resulting from the entry
condition is specified in the following form

u it = δ 1 DNTit + δ 2 FORit + δ 3 GOVit + wit (15)

where: i = industry index , t = index of time

DNT = donation expenses
FOR = output share of foreign firms
GOV = output share of state enterprises

The error term wit has the properties explained in previous section. The expected signs of the
coefficients are given in Table 3

19
 Table 3 Expected sign of entry condition regression
Variables FOR DNT GOV
Expected signs +/− + +

The lower bound model applies to exogenous sunk cost industries only, but does not
necessarily apply to the endogenous sunk costs industries. Therefore, the regression has to be
run separately between the exogenous sunk costs industries and the high advertising and R&D
intensity industries.

VII. Data and the empirical results

The data used in this study are mainly taken from the industrial statistics for large and
medium size manufacturing industries published by the Indonesian Central Bureau of
Statistics (BPS). The period covered by study is between 1986 and 1996. The regression
analysis uses industry level data which are based on the four digit Indonesian Standard
Industrial Classification (ISIC) code. Before 1990 the Indonesian manufacturing data
published by the BPS are classified into the four digit ISCI code, and for 1990 and beyond the
data is classified into five digit code. For consistency in estimation, the industry data for the
period 1990 and beyond are reclassified into the relevant four digit code. Supporting trade
data are obtained from BPS trade statistics and the BPS Input-Output Table. The data for
effective rates of protection are taken from Fane and Phillips (1991), Wymenga (1991), and
Fane and Condon (1996).

The software used to run the regression is the FRONTIER 4.1 package. 6 It is one of
the few software packages that allows estimation of frontier functions. The advantage of
FRONTIER 4.1 is that it allows a more flexible assumption regarding the distribution of the
uit component of the error term. In particular it allows us to estimate equation 14 and equation
15 simultaneously. The program follows a three-step procedure to get the maximum
likelihood estimates of stochastic function’s parameters. In the first step the OLS estimates of
the function is obtained. In the second stage a two-phase grid search is conducted with the
starting value set to the OLS estimates. In the last stage, the values selected in the grid search
are used as the starting values in an iterative procedure.

The regression is run on two groups of industries, the low advertising (exogenous
sunk cost s) industries, and the high advertising (endogenous sunk costs) industries. The
classification is based on each industry’s ratio of advertising expenditure to total output.
Unfortunately the data on advertising expenditures are only available from 1992 to 1996. The

6
FRONTIER 4.1 is provided by Tim Coelli from the Efficiency and Productivity Analysis, the
University of Queensland.

20
annual average advertising expenditure to sales ratio for all industry during this period was
0.77%. The lowest annual average advertising ratio, which stands at 0.0001%, is for the re-
milled rubber industries, and the highest advertising expenditure ratio was for the drug and
medicine manufacturing industry with an annual average of 11.31%. Following Sutton (1991
and Symeonidis (2000) we define an exogenous sunk cost industry as one which spends less
than 1% of its output on advertising. According to Symeonidis the 1% cut-off point was
chosen as it is commonly used to classify industries according to advertising or R&D
intensity. Out of the 116 industries there are 98 industries that have an expenditure ratio of
less than 1% and 18 industries that have advertising expenditure ratio more than 1%. The
breakdown of these industries for each group by two digit ISIC is given in Table 4.

Table 4. Breakdown of industry by advertising intensity

ISIC Descriptions Low advertising High advertising
31 Manufacture of food, beverages and tobacco 24 10
32 Textile, clothes, and leather industry 16 -
33 Wood and wood products 10 -
34 Paper, paper products, printing, and publishing 5 -
35 Chemicals, petroleum, coal , rubber, plastic products 10 6
36 Non metallic mineral products 10 -
37 Basic metal industries 1 -
38 Fabricated metal, machinery, and equipment 17 2
39 Other manufacturing 5 -
Total industries 98 18

The summary of the mean and standard deviation of the five independent variables by
industry’s advertising ratio for at the beginning of observation, the end of the observation, and
the period in between, is given in Table 5 below. As shown in the table, the average level of
concentration declines over the period of study and market size has increased. The decline in
the effective rate of protection is the result of trade liberalization that occurred after 1986. The
exogenous sunk cost industry (low advertising) has higher setup costs relative to the
endogenous sunk costs industry (high advertising). The average minimum efficient scale
measure increases from 1987 to 1990 but falls from 1990 to 1996. However, the changes in
the minimum efficient scale are not statistically significant. 7 The average setup costs have
increased over the 10 years period being observed, but the increase is not statistically
significant.

The data also shows that foreign firms tend to have a larger share in the high
advertising industries, whereas government-owned firms tend to be in the low advertising

7
Based on difference of mean test.

21
industries. This is consistent with the fact that most state enterprises are in the intermediate
input industries rather than in consumer products’ industries.

Table 5. The descriptive statistics for four variables by advertising intensity, 1987,1990, and
1996.
Year Advertising CR4ADJ MSIZE MES SETUP ERP1
1987 Low 44,5 (22.13) 2.170.861 4.33 3.667.508 144,8
(3.765.001) (13.33) (11.180.000) (199.09)
High 64,6 (20.93) 2.553.684 3.89 1.271.867 265,8
(5.687.404) (3.75) (1.928.699) (248.10)
Total 47,6 (23.05) 2.230.265 4.60 3.295.770 163,6
(4.093.598) (12.33) (10.900.000) (211.26)
1990 Low 39,74 3.583.172 5.47 8.592.677 161,211)
(21.14) (5.528.048) (15.10) (34.700.000) (199.09)
High 63,21 4.293.751 4.33 1.718.168 285,271)
(20.40) (8.585.111) (5.00) (2.101.654) (248.10)
Total 45,45 3.693.434 5.29 7.525.943 180,801)
(22.67) (6.061.211) (14.01) (32.000.000) (211.26)
1996 Low 38,98 9.062.377 5.19 9.302.766 49,162)
(20.54) (13.500.000) (9.28) (29.700.000) (121.30)
High 62,34 7.687.570 10.82 2.683.190 111,172)
(21.96) (13.100.000 (15.04) (3.130.249) (150.99)
Total 42,61 8.849.045 6.06 8.275.590 58,782)
(22.34) (1.340.000) (10.50) (27.400.000) (127.63)
Note : figures in parenthesis are the standard deviations
1) ERP for 1989, 2)ERP for 1995

The regression analysis is conducted separately for exogenous sunk cost and
endogenous sunk costs. For the exogenous sunk cost industries the regression is run for six
industry groups and for all the total industries. The six industry groups are: food and beverage
(ISIC 31), textile (ISIC 32), wood and paper product (ISIC 33 & 34), chemical products
(ISIC 35), non-metallic mineral products (ISIC 36), and fabricated metal products (ISIC
37&38). For the endogenous sunk cost (high advertising) industries the regression is run for
the total industry due to the limited number of observations, and because we assume that these
industries are bound by their similarity in advertising expenditure rather than by industry
characteristics.

The stochastic frontier model defined by equation 14 and 15 contains six β-

parameters, three δ-parameters, and three additional parameters associated with the

distribution of the error terms vit and uit. The three parameters are: σ u2 which is the variance

of uit , σ v2 which is the variance of vit, and γ which is the ratio of σ u2 to the sum of the two

variances. The twelve parameters are estimated using maximum likelihood function. The
results of the regression are given in Table 6 below.

We begin our analysis by looking at the appropriateness of using the stochastic

frontier function to estimate the lower bound concentration curve for the sunk costs industry.

22
The likelihood ratio test for one-sided error (H0: γ=0) is rejected for all industry groups. This
result indicates that the traditional average response function (i.e obtained by least squares
method) is not an adequate representation of the data. In other word the result shows that the
frontier approach is the correct specification of the relationship between market concentration
and the independent variables. It is a lower bound relationship.

The model predicts that the market size (LMKSIZE) has a negative effect on lower
bound concentration. This hypothesis is supported by the results. The signs of the coefficients
for the market size variable are negative and statistically significant for all industries except
non-metallic mineral products (ISIC 36) and chemical products (ISIC 35). The effect of
market size on lower bound concentration for the latter two industries is positive but it is not
statistically significant. This result has the feature of the endogenous sunk cost industry where
an increase in market size does not effect market concentration. However, since both
industries are low advertising cost industries, advertising expenditures cannot explain this
finding. One possible explanation is that because the setup cost in these industries is large, a
firm will need to get approval from the government, through the Investment Coordinating
Board, to enter or expand. There may be costs associated with “convincing” the Board to
approve the investment, and these costs increase with market size. As a result when market
expands, entry costs increase making it less attractive for potential entrants to enter.

The setup cost (LSETUP) and minimum efficient scale (LMES) are predicted to have
positive impact on the lower bound market concentration. The regression result supports this
assertion. The coefficients for setup cost are all positive and statistically significant. The
relationship between the minimum efficient scale variable and the lower bound concentration
is positive and statistically significant in every case except for non-metal mineral products
(ISIC 36).

The effect of the trade and industrial policy regime, represented by the dummy
variable DTRDRGM, on market concentration is not obvious. The import-as-market-
discipline hypothesis predicted that trade liberalization will cause an increase in the intensity
of domestic price competition. If the hypothesis holds, it will be reflected by an increase in
the lower bound concentration because tougher competition will reduce profit and will force
firms to either exit or merge. On the other hand, because the Indonesian trade liberalization
emphasized the promotion of exports rather than promoting domestic competition, it is
possible that the policy could have little effect on the domestic competitive environment.

The signs of the coefficient for the trade liberalization dummy variable are all
positive except for the the non-metallic mineral products industry (ISIC 36). For industries

23
with positive coefficients the impact of trade regime on lower bound concentration is
consistent with the import-as-market-hypothesis hypothesis and the argument that trade
liberalization industry in Indonesian has implicitly functioned as a competition policy for the
domestic economy (Bird 1999; Hill 1999). However, it is statistically significant only in the
food industry (ISIC 31) and textile industry (ISIC32), indicating that although trade
liberalization has increased domestic price competition, its effect was not statistically
significant in most industries. Again, this finding can be explained by the specific feature of
the Indonesian trade liberalization which emphasized the promotion of exports. Protection for
import substitution industries was maintained as long as this protection did not impede
exports. Therefore, the impact of liberalization on these industries has been limited, as shown
by the insignificant coefficient of the trade regime variable.

The sign of the trade regime coefficient (DTRDRGM) is negative but not statistically
significant for the non-metallic mineral industry (ISIC 36). The negative sign, although
insignificant, deserve some elaboration since it contradicts the theory. It indicates that after
trade was liberalized price competition tended to soften in this industry thus inducing entry
and lowering the lower bound concentration. The likely reason for obtaining this
inconsistency is because the non-metallic industry consists of products that would be
characterized as nontraded goods like cement products. For these industries, import
competition is likely to have little impact on the pricing behaviour of domestic oligopoly.
Furthermore, the government had taken a very lenient attitude towards collusive pricing in
these industries and trade liberalization did not change the government’s stance on this matter.
A good example for this claim is the cement industry where price was effectively set by an
informal cartel organized by the industry trade association. The collusive pricing was
practised with the support of government, and the support continues despite the change in
trade regime (Maarif, 2001). The high profit associated with collusive pricing and high
demand for cement provided a strong incentive for entry which in turn explained the drop in
the level of concentration.

The effect of trade policy on lower bound market concentration is also determined by
the degree of import protection, which is measured by the effective rate of protection (ERP).
The import-as-market-discipline hypothesis predicts that the market power of oligopolies will
fall as they are exposed to greater import competition, which will force some firms to exit and
others to merge. As a consequence we will see the lower bound concentration increase as the
effective rate of protection falls. The result shows that the effective rate of protection has a
negative effect on lower bound concentration for all industries except the textile industry
(ISIC32) and chemical products industry (ISIC35) where it is positive but insignificant. This
result seems to confirm the hypothesis, however all of the coefficients are statistically

24
Table 6. Regression results
Low advertising industries High
VARIABLES Advertising
ISIC 31 ISIC 32 ISIC 33-34 ISIC 35 ISIC 36 ISIC 37-38 Total
(Food) (Textile) (Wood&Paper) (Chemicals) (Non-metal) (Fabr. metal)
-3.3963 1.6864 -0.7956 -0.7093 -1.1568 -0.4555 -1.1386 3.7995
CONSTANT
(-4.7716)** (3.0526)** (-4.3581) ** (-0.7699) (-0.4558) (-0.6577) (-3.7035)** (5.4728)**
0.2535 0.2352 0.0772 0.00950 -0.2845 0.1104 0.0708 0.4905
DTRDRGM
(1.8673)* (2.0163) * (0.7457) (1.0291 (-1.2188) (0.8780) (1.0180) (3.2592)**
-0.1761 -0.3560 -0.2760 0.0579 0.0864 -0.2414 -0.2178 -0.2598
MKSIZE
(-3.7550)** (-11.1669) ** (-14.9929) ** (1.1479) (0.9473) (-4.6253)** (-11.2228)** (-6.0224)**
0.2557 0.8202 0.5288 1.0060 1.2271 3.5765 0.0507 0.0531
SETUP
(4.8884)** (3.8726) ** (12.7038) ** (12.5334)** (9.4241)** (4.5569)** (14.0461)** (5.7458)**
0.3784 0.1552 0.3136 0.0810 -0.0554 0.1025 0.2488 0.2112
MES
(6.5848)** (2.5878) * (3.9822) ** (2.5994)* (-0.6159) (2.2517)* (9.5269)** (4.7736)**
-0.0003 0.0005 -0.0005 0.0001 -0.0024 -0.0003 0.0000 0.0019
ERP
(-0.5557) (0.9914) (-1.7801) * (0.4305) (-1.3524) (-0.7353) (0.2274) (5.6387)**
0.5188 -0.0964 0.1418 -0.5439 -0.8232 0.1706 0.0611 -0.2980
PDR
(5.5045)** (-1.9970) * (8.4421) ** (-2.4911)* (-2.3984)** (2.3054)* (2.1507)* (-5.5536)**
0.0087 -1.7930 0.0287 0.0282 -0.0206 0.0331 0.0294 -0.1033
GOV
(0.6378) (-1.0735) (1.2073) (11.3298)** (-0.3052) (5.3686)** (10.2017)** (-1.9990)*
-0.0138 -0.1536 0.0392 0.0125 0.0242 -0.0105 0.0177 0.0136
FOR (-0.8341) (-1.7404) * (5.9736) ** (2.5365)** (0.4511) (-1.9770)* (4.6630) ** (3.4199)**
0.0062 0.0401 1.7405 0.1394 -6.1162 0.0637 0.0757 0.0440
DNT
(0.2878) (1.1915) (2.9881) ** (1.5319) + (-1.2862) (0.2208) (1.6579)** (2.0348)*
σ 2 = σ v2 + σ u2
2.6101 0.6582 0.8871 0.3830 2.9381 0.1689 1.5147 0.5423
(5.8680)** (3.8683) ** (5.2679) ** (5.4036)** (2.8290)** (5.4728)** (12.2829)** (7.3965)**
γ = σ u2 /( σ u2 + σ v2 )
0.9124 0.5808 0.9999 0.9999 0.9188 0.0192 0.7668 0.0680

-307.00 -133.2119 -131.0824 -51.3249 -120.4287 -50.3976 -1188.8549 -177.7366

LR
12.88* 10.51* 61.92** 76.51** 11.22* 34.67** 133.2935** 8.51*
One-sided test (χ2)
Number in parentheses are t-statistics
**
) significant at 1%, * ) significant at 5%, +
) significant at 6%

25
insignificant except for wood and paper products industry (ISIC 33&34). This indicates that
exposure to import competition has had little impact on domestic price competition. Although
the finding does not find support for the import-as-market-discipline hypothesis, it is
consistent with previous finding by Melo and Urata (1986) using Chilean manufacturing data
which find that import has insignificant effect on price-cost margin.

The second part of the concentration function is the deviation of actual concentration
from the lower bound. It measures the intensity of competition brought about by the entry
condition in manufacturing industries. We use donation expenditures, foreign ownership, and
government ownership as independent variables that explain the observed variation in entry
conditions across industries.

Donation expenditures have a very significant impact on the entry condition of the
wood and paper products (ISIC 33 & 34), and of the chemical products industry (ISIC 35).
The wood and paper products industry in Indonesia is a very lucrative industry as Indonesia,
at least during the period of study, had some market power in the world market for hardwood
products. Entry into this industry depends heavily on obtaining government permits
(concessions) to exploit forest wood. The license carries a very high economic rent and
therefore became the reason for rent seeking activities. The trade association in this industry,
APKINDO, was very influential. Not only did APKINDO have the power to determine price
and allocate market shares through its joint marketing body, it can also advise the government
to revoke producer’s license if it violates the price and marketing arrangement set up by the
joint marketing body. More importantly APKINDO’s decisions are supported by the
government. The result indicates that donation expenditure is an effective tool for this
industry to influence government entry policies in this industry to ensure the economic rents.

Foreign investment is expected to increase domestic competition. The results for the
textile industry (ISIC 32) and for the fabricated metal products industry (ISIC 37-38) support
his hypothesis. In both industries the coefficient of the “foreign-firm share” variable (FOR) is
negative and statistically significant which indicate that a larger presence of foreign firms
tend to bring down entry barriers and thus bring the concentration level closer to the lower
bound. However, for reasons explained earlier, the presence of foreign firms can also reduce
competition if the foreign firms replace the domestic oligopoly rather than discipline their
anti-competitive behaviour. This seems to be the case for the chemical products industry
(ISIC 35) and for the wood and paper products industry (ISIC 34) where the coefficients are
positive and significant. In these industries imports join the pricing strategy of the domestic
oligopoly.

26
 State enterprise plays a very important role in Indonesian manufacturing industry.
Although there are about 500 state enterprises (fully owned or joint venture) in manufacturing
industry, they play a different role in different industries. These enterprises were established
for a number of reasons such as to correct market failure, as a source of government revenue
(particularly those belonging to regional government), and to support the government’s
economic development agendas. Some of the state enterprises are foreign firms that were
nationalized in the 1950s. Depending on the objective of their creation, state enterprises have
different level of importance to the government. Those that are established to support
economic development agendas are generally more important to the government than those
that are established for revenue sources. The importance of state enterprises is measured by
the contribution of these enterprises to the total production in the industry where they are
operating. The majority of state enterprises (approximately 29%) are in the food products
industry (ISIC 31), however the contribution of these enterprises to the total production of this
industry is only six percent. There are 90 state enterprises in the chemical product industry
(ISIC 35) with an average contribution of 15 percent total industry production. The number of
state enterprises in the fabricated metal products industry (ISIC 38) is 45 with an average
contribution of 15 percent to total industry production. On the other hand there are only 2
state enterprises in the metal products industry (ISIC 37) but they contribute about fifty
percent of the industry’s total production. Based on these figures we maintain that the role of
state enterprises in the latter three industries (ISIC, 35, 37 and 38) is more important to the
government than their role in the former. Therefore, we would expect entry protection to be
higher in these three industries.

The result shows that the coefficient for the “state enterprises share” variable (GOV)
is positive and statistically significant in the chemical product industry (ISIC 35) and in the
basic and fabricated metal product industries (ISIC 37 & 38). However, the coefficient of this
independent variable is not significant in the other industries.

A separate regression is run for the high advertising industries since the theory
predicts a different outcome for the relationship between market size and market
concentration in this industry. The result of the one-sided test (H0 : γ=0) is rejected which
means that the stochastic frontier method is more appropriate than the least squares method in
estimating the relationship between the six independent variables and the concentration level.
The theory predicts that in the high advertising (endogenous sunk cost) industry, lower bound
market concentration may not fall with higher market size. There may still be a negative
relationship between the two variables when the market is still relatively small but the
relationship may not prevail as the market size becomes relatively large. The results for the 18

27
high advertising manufacturing industries show that the market size has a negative impact on
lower bound concentration. This indicates that the market size for the advertising intensive
industries is relatively too small for the firms to use advertising as the main tools of
competition. As can be seen from Table 5 the difference in market size between low
advertising and high advertising industries are statistically different, and from 1990 to 1996
the market size of the high advertising industries grew less than the low advertising industries.

The trade liberalization policy significantly increased domestic price competition for
the high advertising industry. This is indicated by the positive and statistically significant
value of the trade regime dummy variable coefficient. The tougher price competition brought
about by trade liberalization forced domestic firms in this industry to either exit or merge and
thus, shifting the lower bound concentration curve upward. However, the coefficient for the
effective rate of protection is positive and significant which shows that lower protection tends
to soften the intensity of price competition, thus reducing the incentive for firms to enter. This
result contradicts the import-as-market discipline hypothesis. One possible explanation for
this contradiction is that the increase in price competition brought about by higher imports is
countered by an increase in advertising expenditure. Since advertising expenditure is a sunk
costs, the increase in advertising implies a smaller number of firms can be supported by the
market.

With regards to the factors that explain the deviation of the concentration level from
the lower bound curve, the result shows that the coefficient of foreign firm’s share and
donations are both positive and significant. Since the deviation measure the degree of entry
barriers, the results indicates that entry barriers tend to be high when there is an extensive
presence of foreign firms. The result also shows that rent seeking, as measured by the size of
donation expenditures, is an effective means for creating high entry barriers. On the other
hand because state own enterprise plays an insignificant role in the high advertising industry,
it is insignificant in creating barriers to entry.

VIII. Conclusion

At the beginning of this paper we asked the question whether trade liberalization can
increase domestic competition. It was argued that due to the specific nature of Indonesian
trade liberalization, that is its emphasis on export promotion, the conventional hypothesis
about the effect of trade liberalization on domestic competition may not apply. It was also
argued that due to the extensive government role in the economy and to the nature of the
political system, rent seeking and the presence of state enterprises could have a strong impact
on domestic competition through their influence on the entry condition. We used the lower

28
bound function to test these hypotheses. The lower bound function, which has been used in
previous studies with similar objective, was estimated using the stochastic frontier approach.

Although the model used in this paper is unconventional, it has successfully

supported the hypotheses proposed in this paper. The results of the regression are consistent
with the prediction of the lower bound theory proposed by Sutton. In cases where the results
are inconsistent with the theory, we have provided a logical explanation for the inconsistency.
With regard to testing the effect of trade liberalization on domestic competition, we have
found support for the proposition that the Indonesian trade liberalization has not contribute
significantly to domestic competition. With regards to the disciplining effect of import we
could not find convincing support for the conventional hypothesis that exposure to imports
competition increases domestic competition. We found that although a reduction in trade
restrictions do generally increase domestic competition, they are not statistically significant.
However, these findings are consistent with the empirical findings of some previous research
on the same issue.

The results also support the assertion that rent seeking and the presence of state
enterprises tend to increase entry barriers. We have found that in industries identified to be
sensitive to rent seeking behaviour the proxy for rent seeking expenditure contributes to
higher entry barriers, and thus soften domestic competition. The results also show that entry
barriers tend to be higher in industries where state enterprises play a significant role. Lastly,
we have also found that the presence of foreign firms do not always contribute to the
enhancement of domestic competition.

The findings in this paper have a number of policy implications for Indonesia. First,
the role of competition policy can not be entirely replaced by trade liberalization since the
idea that international trade induces competition hinges on how trade liberalization policy is
designed. Second, the popular notion in Indonesia that state enterprises are effective sources
of government revenue and therefore should not be privatized has little merit. In a rent
seeking society like Indonesia, state enterprise is more likely to be exploited by rent seekers,
and therefore tend to be very inefficient. Furthermore, state enterprises are inclined to impede
competition. Third, given Indonesia’s history of nepotism, cronyism, and weak governance
problem, less government interference in the economy is always better.

29
 REFERENCES

Aigner, D. J. and S. F. Chu (1968). "On Estimating the Industry Production Function."
American Economic Review 58: 826-839.
Aigner, D. J., C. A. K. Lovell, et al. (1977). "Formulation and Estimation of Frontier
Production, Profit and Cost Functions." Journal of Econometrics 6(1): 21-37.
Bain, J. S. (1956). Barriers to New Competition. Cambridge, Harvard University Press.
Basri, M. C. and H. Hill (1996). "The Political Economy of Manufacturing Protection in
L.D.Cs : An Indonesian Case Study." Oxford Development Studies 24(3): 241-259.
Battese, G. E. and T. J. Coelli (1992). "Frontier Production Functions, Technical Efficiency
and Panel Data: With Application to Paddy Farmers in India." Journal of Productivity
Analysis 3: 153 - 169.
Battese, G. E. and T. J. Coelli (1995). "A Model for Technical Inefficiency Effects in a
Stochastic Frontier Production Function for Panel Data." Empirical Economics 20(2):
325-32.
Bhattacharya, M. (2002). "Industrial Concentration and Competition in Malaysia
Manufacturing." Applied Economics 34(17): 2127-34.
Bhattacharya, M. and H. Bloch (1997). "Specifiaction and Testing the Profit-Concentration
Relationship in Australian Manufacturing." Review of Industrial Organization 12:
219-230.
Bhattacharya, M. and H. Bloch (2000). "The Dynamic of Industrial Concentration in
Australia." International Journal of Industrial Organization 18(8): 1181-99.
Bird, K. (1999). "Concentration in Indonesian Manufacturing, 1975-93." Bulletin of
Indonesian Economic Studies 35(1): 43-73.
Collins, B. R. and L. E. Preston (1969). "Price-Cost Margins and Industry Structure." The
Review of Economcs and Statistics 51(3): 271-286.
Curry, B. and K. D. George (1983). "Industrial Concentration : A Survey." Journal of
Industrial Economics XXXI(3): 203-255.
Davies, S., B. Lyons, et al. (1988). Economics of Industrial Organization. London and New
York, Longman.
Davies, S. W. and P. A. Geroski (1997). "Changes in Concentration, Turbulence, and the
Dynamic of Market Shares." Review of Economics and Statistics 79(3): 383-91.
Demsetz, H. (1973). "Industry Structure, Market Rivalry, and Public Policy." Journal of Law
and Economics 16(1): 1-9.
Dixit, A. (1980). "The Role of Investment in Entry Deterrence." Economic Journal 90: 95 -
106.
Domowitz, I., R. G. Hubbard, et al. (1986). "Business Cycles and the Relationship Between
Concentration and Price-Cost Margin." The Rand Journal of Economics 17(1): 1-17.
Domowitz, I., R. G. Hubbard, et al. (1986). "The Intertemporal Stability of the Concentration-
Margins Relationship." Journal of Industrial Economics 35(1): 13-34.
Driffield, N. (2001). "Inward Investment, Industry Concentration, and the Speed of
Adjustment." Weltwirtschaftliches Archiv 137(2): 193-214.
Esposito, L. and F. Esposito (1971). "Foreign Competition and Domestic Industry
Profitability." Review of Economics and Statistics 53: 343-53.
Fane, G. and T. Condon (1996). "Trade Reform in Indonesia, 1987-1995." Bulletin of
Indonesian Economic Studies 32(3): 33-54.
Fane, G. and C. Phillips (1991). "Effective Protection in Indonesia in 1987." Bulletin of
Indonesian Economic Studies 27(1): 105-125.
Feinberg, R. (1996). "The Impact of Trade Liberalization and Competition Policy on
Domestic Pricing." Challenge 39(4): 40-44.
Geroski, P. A. (1991). Market Dynamics and Entry. Oxford, Blackwell.
Geroski, P. A. and A. Jacquemin (1981). "Import as a Competitive Discipline." Recherches
Economiques de Louvain 47(3-4): 197-207.

30
Geroski, P. A., R. T. Masson, et al. (1987). "The Dynamic of Market Structure." International
Journal of Industrial Organization 5: 93-100.
Ghellinck, E. d., P. A. Geroski, et al. (1988). "Inter-industry Variations in the Effect of Trade
on Industry Performance." The Journal of Industrial Economics 37(1): 1-19.
Globerman, S. (1990). "Trade Liberalization and Competitive Behaviour : A Note Assessing
the Evidence and the Public Policy Implications." Journal of Policy Analysis and
Management 9(1): 80-88.
Grether, J.-M. (1996). Mexico, 1985-1990: Trade Liberalization, Market Structure, and
Manufacturing Performance. Industrial Evoulution in Developing Countries. M. J.
Roberts and J. R. Tybour, Oxford University Press: 260-284.
Grossman, G. M. and E. Helpman (1994). "Protection for Sale." American Economic Review
84(4): 833-850.
Hall, R. (1988). "The relationship between Price and Marginal Cost in U.S. Industry." Journal
of Political Economy 96(5): 921-947.
Hill, H. (1999). "Indonesia's Microeconomic Policy Changes : Industry Policy, Competition
Policy, and Small-Medium Enterprises." The Indonesian Quarterly XXVII(1): 22-33.
Hillman, A. L., N. V. Long, et al. (2001). "Protection, Lobbying, and Market Structure."
Journal of International Economics 54: 383-409.
Kalirajan, K. P. (1993). "Simultaneity Bias : Is It Relevant in the Developing Countries
Industrial Structure-Performance Paradigm?" Journal of Quantitative Economics 9(1):
147-156.
Katics, M. M. and B. Petersen (1994). "The Effect of Rising Import Competition on Market
Power : A Panel Data Study of US Manufacturing." The Journal of Industrial
Economics 42(3): 277-286.
Kee, H. L. and B. Hoekman (2007). "Imports, entry and competition law as market
disciplines." European Economc Review 51: 831-858.
Konings, J., P. V. Cayseele, et al. (2001). "The Dynamic of Industrial Mark-ups in Two Small
Open Economies : Does National Competition Policy Matter?" International Journal
of Industrial Organization 19(5): 841-859.
Kumbhakar, S. C. and C. A. K. Lovell (2000). Stochastic Frontier Analysis. New York,
Cambridge University Press.
Levinsohn, J. (1993). "Testing the Imports-as-Market-Discipline Hypothesis." Journal of
International Economics 35: 1-22.
Lyons, B., C. Matraves, et al. (2001). "Industrial Concentration and Market Integration in the
Europen Union." Economica 68: 1-26.
Maarif, S. (2001). Competition Policy in Indonesia, Komisi Pengawas Persaingan Indonesia.
Martin, S. (1988). "Market Power and/or Efficiency?" The Review of Economics and
Statistics 70(2): 331-335.
Meeusen, W. and J. v. d. Broeck (1977). "Efficiency Estimation from Cobb-Douglas
Production Functions with Composed Error." International Economic Review 18(2):
435-444.
Melo, J. d. and S. Urata (1986). "The Influenced of Increased Foreign Competition on
Industrial Concentration and Profitability." International Journal of Industrial
Organization 4: 287-304.
Pitt, M. M. and L.-F. Lee (1981). "The Measurement of Sources of Technical Inefficiency in
the Indonesia weaving industry." Journal of Development Economics 9(1): 43-64.
Salop, S. and D. T. Scheffman (1987). "Cost-raising Strategies." The Journal of Industrial
Economics 36(1): 19-34.
Salop, S. C. and D. T. Scheffman (1983). "Raising Rivals' Costs." American Economic
Review 73(2): 267-271.
Salop, S. C., D. T. Scheffman, et al. (1984). A Bidding Analysis of Special Interest
Regulation: Raising Rivals' Costs in a Rent Seeking Society. The Political Economy
of Regulation: Private Interest in the Regulatory Process. R. Rogowsky and B.
Yandle. Washington, DC., Federal Trade Commission: 102-127.

31
Schmalensee, R. (1989). Inter-industry Studies of Structure Performance. Handbook of
Industrial Organization. R. Schmalensee and R. D. Willig. Amsterdam, North
Holland: 951-1009.
Singleton, R. C. (1997). "Competition Policy for Developing Countries : A Long-Run, Entry-
Based Approach." Contemporary Economic Policy XV: 1-11.
Spence, M. (1977). "Entry, Capacity, Investment and Oligopolistic Pricing." Bell Journal of
Economics 8: 534 - 544.
Strickland, A. D. and L. W. Weiss (1976). "Advertising, Concentration, and Price-Cost
Margins." Journal of Political Economy 84(5): 1109-1122.
Sutton, J. (1991). Sunk Cost and Market Structure. Cambridge, The MIT Press.
Symeonidis, G. (2000a). "Price Competition and Market Structure : the Impact of Cartel
Policy on Concentration in the UK." Journal of Industrial Economics 48(1): 1-26.
Tybout, J. R. (1996). Chile, 1979-86: Trade Liberalization and Its Aftermath. Industrial
Evolution in Developing Countries. M. J. Roberts and J. R. Tybout, Oxford
University Press: 200-226.
Tybout, J. R. (2000). "Manufacturing Firms in Developing Countries : How Well Do They
Do, and Why?" Journal of Economic Literature XXXVIII: 11-44.
Urata, S. (1984). "Price-Cost Margins and Imports in an Oligopolistic Market." Economics
Letters 15: 139-144.
Waterson, M. (1984). Economic Theory of the Industry. New York, Cambridge University
Press.
Wymenga, P. S. J. (1991). "The Structure of Protection in Indonesia in 1989." Bulletin of
Indonesian Economic Studies 27(1): 127-53.

32