This paper develops a simple theory of the effect of corruption and

Corruption and Energy Efficiency in OECD
Countries: Theory and Evidence*
Per G. Fredriksson*
Southern Methodist University
Herman R.J. Vollebergh
Erasmus University Rotterdam and OCfEB, Rotterdam
Elbert Dijkgraaf
Erasmus University Rotterdam and OCfEB, Rotterdam
This version: June 14 , 2002
The effect of lobby group size on policy outcomes is an unresolved question in the literature. In this
paper we shed new light on this issue by showing that an interaction exists between corruption and
lobby group size. In particular, the theory predicts that the effect of sector size is conditional on the
level of corruption. Moreover, corruption reduces environmental policy stringency. These predictions
are tested using a unique panel data set on energy efficiency for 11 sectors in 14 OECD countries for
years 1982-1997. The evidence supports the hypothesis that the size of a sector is important, in
particular in OECD countries with relatively widespread corruption. Furthermore, corruption explains
differences in energy efficiency across industrialized countries.
Keywords: Energy efficiency, political economy, rent seeking, corruption.
JEL Code: D 78, Q 48.
We would like to thank Leon Bettendorf, Raymond Florax, Daniel Millimet, Qin Wang, participants at the
conference on “The International Dimension of Environmental Policy” in Aquafredda, Italy, October 2001, and
especially Pieter van Foreest for helpful comments on earlier versions of the paper. Vollebergh acknowledges
financial support from the research program ‘Environmental Policy, Economic Reform and Endogenous
Technology’, funded by the Netherlands’ Organisation for Scientific Research (NWO). Peter Mulder and Daan
van Soest were helpful in providing us with much of the data. The usual disclaimers apply.
Correspondence: Herman Vollebergh, Department of Economics, Erasmus University Rotterdam, H
7-23, P.O. Box 1738, 3000 DR, Rotterdam, The Netherlands; Tel: +31-10-4081498; Fax: +31-104089147; Email: [email protected]
I. Introduction
The effect of lobby group size on the ability of special interests to influence policy is an
unresolved issue in the literature. Olson (1965) argues that free-riding problems and coordination
costs of lobbies will reduce the influence of large political groups. Interest groups with a large stake
in the policy outcome can be expected to be more active. The empirical results are far from
inconclusive, however (see Rodrik (1995)). Potters and Sloof (1996, pp. 417-8) point out that “Most
scholars indeed find an increased scope for political influence with higher degrees of concentration,
but there are many that find no effect or even a negative effect.”… and “there seems to be relatively
little direct empirical support for the Olson (1965) influential theoretical study on collective action.”1
A distinct literature discusses the effect of corruption on policy outcomes, pollution, investment, and
economic growth.2 However, these two literatures have expanded in different directions and have not
intersected. In this paper, we seek to fill this void.
Our contribution centers on the interaction between lobby group size and corruption in the
determination of policy outcomes. We use energy policy in industrialized countries as our example.3
We start the analysis by developing a simple model of bribery of the government by worker and
capital owner special interest groups. We build on the menu auction model originating with Bernheim
and Whinston (1986) and Grossman and Helpman (1994) (see also Aidt (1998) and Damania
(2001)).4 As in Laffont and Tirole (1991), bribery is assumed costly to coordinate due to free-riding
problems (see also Eerola (2002)). These coordination costs may for example be due to the size of the
For example, Pincus (1975), Marvel and Ray (1983), Esty and Caves (1983), Kalt and Zupan (1984), Godek (1985),
Walter and Pugel (1985), Gardner (1987), Trefler (1993), Gawande (1997), and Gawande and Bandyopadhyay (2000),
found a positive effect of industry concentration on political influence, whereas Pincus (1975), Salamon and Siegfried
(1977), Finger et al. (1982), Baldwin (1985), and Cahan and Kaempfer (1992) found negative, ambiguous, or insignificant
2 See, for example, Shleifer and Vishny (1993), Mauro (1995), Tanzi (1998), Ades and Di Tella (1999), López and Mitra
(2000), Wei (2000), Fredriksson et al. (2001), and Fredriksson and Svensson (2002).
3 Many casual observers may associate corruption with developing countries. However, recent high-level corruption
scandals in for example Germany and France (embroiling, for example, former Prime-minister Kohl and President Chirac)
indicate that the OECD countries are not immune to this phenomenon. In the United States, the recent collapse of the
major Texas energy firm Enron has focused the attention on energy firms’ open and hidden political activities.
industry sector (our focus) and/or geographical dispersion of firms (see Olson (1965) and Pincus
(1975)). In the first stage of the game, each lobby offers the government prospective bribes in return
for favorable energy policy choices. Energy policy determines the maximum allowed energy use, thus
restricting the productivity of labor and capital. However, greater energy intensity of production
implies that consumers suffer greater environmental damage. The government is assumed to care both
about aggregate social welfare and about bribes. We view the (exogenously given) trade-off between
social welfare and bribes as the level of corruption. In the second stage, the government selects its
optimal energy policy, and receives the corresponding bribes from the two lobbies.
Three main predictions emerge from the model. First, corruption reduces the stringency of
energy policy. Intuitively, corruption shifts the government’s relative weighting from welfare to
bribes, facilitating the purchase of influence by making it cheaper.
Second, increasing costs of coordinating bribery (collective action) for the lobbies cause
energy policy to become more stringent. This is in particular the case where the degree of corruption
is high. The intuition is that the special interest groups are influential where the level of corruption is
high, since buying influence is cheap. An increase in organization costs has a particularly high
marginal effect where the level of influence is high. The coordination costs cause the lobbies to scale
back their bribe offers, which has a high marginal impact where bribery is cheap. The model thus
highlights a new interaction, and brings a new perspective on the effect of collective action costs on
Third, the distribution of the effects of coordination cost on the worker and capital owner
lobbies’ political pressures depends on how energy policy affects the lobby members’ income. These
effects are inversely related, i.e. when the effect of coordination costs on worker lobbying is high
(low), the effect on capital owner lobbying is low (high). Worker and capital owner lobbying are
therefore substitutes.
See Heyes and Dijkstra (2001) for a useful survey of the literature on the political economy of environmental policy.
We test these predictions using panel data on energy intensity of output by sector from OECD
countries for the years 1982-1997. The empirical findings support several of the model’s predictions.
First, corruption strongly reduces energy efficiency in OECD countries. Second, an increase in
coordination costs (measured as an industry sector’s contribution to total value added) reduces the
influence of the capital owners, consistent with Olson (1965). This in particular the case in countries
with relatively low levels of corruption. This is a robust finding. However, our empirical results also
suggest that increased employment (with associated greater coordination costs) may be associated
with less stringent regulations, thus increasing worker influence as the (potential) lobby group size
rises. This effect disappears as corruption rises, however.
It appears that rising coordination costs have different effects on worker and capital owner
lobbying. Moreover, the effects of coordination costs are lower where the degree of corruption is
high. As an explanation of the differing success of worker and capital owner lobby groups, voting
strength is important for the influence of workers. On the other hand, worker rent-seeking (bribery)
may take place in relatively corrupt countries. There, the coordination of bribery becomes a problem
for workers in large sectors.5 In countries with less corruption, worker lobby groups may not suffer
from coordination problems in the environmental policy area. The findings indicate that worker and
capital owner lobbying may be substitutes in the business of energy policy influence seeking and
bribery. We believe this is a novel finding, and may account for some of the ambiguities reported in
the previous empirical literature.
Moreover, to the best of our knowledge little empirical work has been devoted to studying the
effects of corruption on policy outcomes in industrialized countries, and consequently this is a new
contribution to the general literature on corruption. Finally, since energy policy is an important focus
of the current debate (for example in the negotiations on global warming), we believe our findings
may have policy implications. Reforms aimed at reducing corruption would have the indirect benefit
of raising energy efficiency in industrialized countries.
The paper is organized as follows. Section II sets up the model, derives the equilibrium energy
policy, and generates the predictions of the model. Section III describes the empirical model and the
data. Section IV reports the empirical results, and Section V provides a brief conclusion.
II. The Model
In this section, we develop a simple model of the endogenous formation of energy policy and
corruption in a small open economy.6 Each country contains firms that produce a private good, Q, for
a perfectly competitive international market; price equals unity. Production requires inputs of capital,
K, labor, L, and energy, . The capital stock and the labor supply are immobile internationally, while
energy can be imported free of import duties at price p. Energy use is polluting but the damage is
assumed confined to the country using the energy. The production technology exhibits constant
returns to scale, is concave and increasing in all inputs, twice continuously differentiable. It is given
by Q  F ( K , L,  ) , which by linear homogeneity can be rewritten as
Q  Kf (l , ) ,
where l  L / K is the labor-capital ratio (the inverted capital-labor ratio) and    / K ,    ,
   is the energy-capital ratio. The energy-capital ratio is the environmental policy set by the
Because the amount of capital is fixed in each country, this implies that  determines the
aggregate energy use. Implicitly, it also specifies the energy-intensity of production. Suppressing
arguments and using subscripts to denote partial derivatives, the marginal products of capital, energy,
and labor are given by ( f  lf l  f  ) , f  , and f l . The marginal products are diminishing, i.e.,
The lack of “social capital” (mutual trust) in a corrupt country would thus, in fact, mitigate corruption, at least to some
extend. This may be an interesting topic for future research. For a discussion of social capital, see Knack and Keefer
(1997) and Putnam (2000).
f   0 , f ll  0 , and we assume that f l  0 , i.e., an increase in  raises the marginal product of
labor. With constant returns to scale, the size of each firm is indeterminate. The aggregate profit
function of the country’s firms is equal to
  Kf (l , )  Kr  Lw  p,
where r is the cost of capital and w denotes the wage rate. Assuming many small firms, such that r
and w are taken as given, the FOC of (2) with respect to energy use,  , yields f   p. We assume
that the government’s regulation of energy use is binding, such that energy use by firms is restricted to
a quantity lower than implied by the FOC.
There are three types of individuals in this economy: workers, consumers, and capital owners.
Normalizing the population in each country to 1, let  W  L,  S , and  K represent the proportion of
the population that are workers, consumers, and capital owners, respectively. All individuals gain
utility from consuming the good, but consumers in addition suffer damage, D, from the pollution
stemming from the energy input use. This damage suffered by each environmentalist is, for simplicity,
directly proportional to the energy used (and thus the environmental policy set by the government),
i.e. D   . Individuals are assumed to have additively separable utility functions of the following
U j  c j  S D ,
where j  W , S , and K index workers, consumers, and capital owners, respectively, and  S  1 for
environmentalists (0 otherwise).
The income Y S of each consumer is exogenously determined, e.g., gained from employment
in white-collar jobs unaffected by environmental policy. Workers supply one unit of labor each and
the wage equals the gain from employing an additional worker. The marginal product of capital
equals the sum of the marginal product of capital given permitted energy use, plus the additional
The model partially builds on Oates and Schwab (1988) and Fredriksson et al. (2001).
output arising from the increase in allowable energy use, f  . Hence, the returns to capital is given
by f  lf l ( 0).
Capital owners and workers are able to overcome free-rider problems and form a lobby group
(see Olson, 1965). However, the consumers face sufficiently large such costs that they are unable to
overcome the problems associated with collective action. Let both the organized and unorganized
population groups be denoted by a superscript i, i=W,K,S. The organized worker and capital owner
lobby groups are assumed to offer the government a bribe schedule, C i ( ), i  W , K . The bribe
schedule offered by each lobby i relates a prospective bribe to the equilibrium energy policy chosen
by the government. However, we follow Laffont and Tirole (1991) by assuming that corruption
involves coordination (organizational) costs, such that the total cost to lobby i of bribery equals
(1  i )C i ( ) , where we assume that the larger the free-riding problems of lobby group i, the greater
the coordination costs, i (see also Eerola, 2002). For example, the greater the size of the lobby and
the more geographically dispersed, the greater the free riding problems and thus the coordination
costs (see, for example, Olson (1965) and Pincus (1975)).
The net (indirect) utility functions of the worker and capital owner lobby groups are given by
V W ( )  Lf l  (1  W )C W ( ),
V K ( )  K ( f  lf l )  (1   K )C K ( ),
respectively. Both these groups prefer weaker restrictions on energy use. Although not organized in a
lobby group, we express the aggregate utility of the consumers as
V S ( )   S (Y S  D)   S (Y S  K ),
since   D.
The government derives utility from a weighed sum of aggregate social welfare and bribes,
and thus its utility function is given by
V G ( ) 
 aV ( ) 
i  W, K, S
 C ( ),
i  W, K
where a  0 represents the exogenous weight that the government places on social welfare relative to
bribes, and
i W , K , S
V i ( ) represents aggregate social welfare. We interpret the weight a as the level
of corruption of the regime. The greater the influence of lobbying activities (i.e. the smaller is a)
relative to social welfare, the greater the level of corruption. Clearly, in some countries, even
democratic OECD countries, the transfer of funds to politicians is legal. In our view this represents a
form of corruption, and clearly other authors share this perspective. Shleifer and Vishny (1993) and
Bardhan (1997) define governmental corruption as the propensity to sell policies for personal gains in
the form of monetary transfers. Moreover, Schulze and Ursprung (2001) and Fredriksson and
Svensson (2002) view the interaction between lobby groups and the government in the Grossman and
Helpman (1994) model as closely describing corruption since the monetary transfers (bribes) are
aimed at influencing government policy and not elections. The level of corruption in our model is, in
essence, reflected by the government’s willingness to allow lobby groups to influence the
determination of energy policy.
The equilibrium energy policy is determined as the outcome of a two-stage, non-cooperative
game. In stage one the lobby groups offer the government a bribe schedule C i ( ), i=W,K. The
strategy of each lobby is a differentiable function C i :     ; and each lobby offers the
government a monetary reward for selecting the policy . In stage two, the government selects an
energy policy and collects the associated bribe from each lobby. The lobbies are assumed not to
renege on their promises in the second stage.
The lobbies receive payoffs described by
V i :     .7
The subgame perfect Nash equilibrium is defined as a feasible bribe schedule C and an energy policy  such that the
energy policy maximizes the government's welfare V ( ) , taking the bribe schedule as given; and given the
The equilibrium in the common agency model by Bernheim and Whinston (1986), which has
been applied by for example Grossman and Helpman (1994) and Dixit et al. (1997), maximizes the
joint surplus of all parties and is formally equivalent to the Nash bargaining solution. The
characterization of the equilibrium energy-capital standard,  * , can be derived from the following
two necessary conditions (see Eerola (2002) for a complete proof):
 *  arg max V G ( );
 *  arg max{ [V i ( )  (1  i )C i ( )]  V G ( )}, i.
The FOC of condition (C2) implies Vi ( )  (1  i )Ci ( ), which can be substituted into the FOC of
(C1) to yield the characterization of the equilibrium energy policy
 aV ( ) 
i  W , K, S
 1  i
V ( ).
1  i 
i  W, K 
The failure of the consumers to participate in bribery gives rise to a political distortion. From (8) it
follows that their interests are not represented to the same degree as the workers’ and capital owners’
interests. Whereas these latter groups receive a weight of [1 (1  i )  a] , the consumers receive a
weight of only a. As i  , the weight of all groups converge to a, however. Moreover, note that
the greater the organizational costs involved with bribery, the lower the influence of the organized
special interest groups.
To find an explicit expression for the equilibrium energy policy (corresponding to (8)), we
need to find the effects of a change in the energy policy on the gross welfare of the groups. The effect
of energy policy on the workers’, capital owners’, and consumers’ welfare is given by
VW ( )  Lf l  0,
VK ( )  K ( f   lf l )  0,
government's and lobby i’s strategies, lobby j does not have a feasible strategy that results in a net payoff greater than the
VS ( )    S K  0,
respectively. Notice that the effect of energy policy on the welfare of the worker and capital owner
lobby groups are here linearly related. If the wage effect is large (small), the effect on the returns to
capital is small (large). Below, we will see that the effect on income determine the equilibrium
lobbying efforts of the two lobbies.
Substituting (9), (10) and (11) into (8) yields an equilibrium condition for the energy policy
equal to
 1
 1
 a lf l  
 a ( f   lf l )  a
 S  0.
 
 
  
Equation (12) shows the forces on energy policy, in equilibrium. Term A captures the
influence of the worker lobby group, term B mirrors the influence of the capital owner lobby group,
and term C reflects the government’s consideration to consumers’ welfare. All terms multiplied by an
a are included in social welfare, and terms containing i reflect the bribery effort of lobby i. Note that
simple rearrangements of (12) imply that, in equilibrium,  S  f  , i.e. the marginal disutility from
energy related pollution is greater than the marginal productivity of energy. This is due to the lack of
participation by the consumers in bribery activities.
The previous literature has made a distinction between interest group “activity” and “success”
(see the survey by Potters and Sloof (1996)). Let us discuss this in our framework. For example, the
aggregate influence of the worker lobby depends on three factors, in equilibrium. First, the incentive
to offer a bribe (reflected by lf l ) , which can be interpreted as “activity” level. Second, the ability to
offer a bribe [reflected by 1 /(1  W )], ceteris paribus. Third, the degree of government corruption, a,
which is the willingness of the government to sell policy favors. “Success” in influence-seeking is
equilibrium net payoff.
consequently the combination of these three factors, as seen in (12). Our main contribution is the
more precise description of how corruption interacts with lobby group activity (incentives) and ability
(coordination costs) in the determination of the level of success in the policy outcome.
To examine the effect of corruption on energy policy, we assume for simplicity that all thirdorder conditions are approximately zero [assumed also by Laffont and Tirole (1998, p. 66)], and
differentiate (12) to find
 S  f
a  1
 a  f 
1 
which is negative. An increase in corruption (a lower a) raises the distortion in energy policy. This is
simply because the government’s willingness to sell policy favors rises.
We summarize our first testable prediction as:
Prediction 1: Corruption reduces the stringency of energy policy.
Next, we find that the effects of the coordination costs of both capital owners and workers.
These effects are shown to be conditional on the degree of corruption, as reflected in the following
comparative static results:
1  K
1  W
 f  lf l 
 1
 a  f 
1 
 0,
 0.
lf l
 1
 a  f 
1 
Rising coordination costs for the lobbies cause energy policy to become more stringent, because the
bribe offers are scaled back. Moreover, the effect is greater where the level of corruption is great. In
such situations, the policy is relatively more distorted by bribery, and thus coordination costs have a
greater impact on the overall pressure on energy policy from the lobbies. An increase in the marginal
cost of bribery will therefore have a greater marginal (negative) impact. When corruption is virtually
absent, on the other hand, bribery has a small impact on policy. Consequently, changes in the
collective action costs are less important for the policy outcome in this case. Note also that the effect
of coordination costs is non-linear since i appears on the right hand side.
The main prediction that emerges is:
Prediction 2: Lobby group coordination costs cause energy policy to become more stringent, in
particular when the degree of corruption is high.
Finally, (14) and (15) reveal that the effect of the two groups’ coordination costs are interrelated and proportionate to the lobbying incentives reflected in the numerators. Ceteris paribus, when
the wage effect lf l (in the numerators) is large, the effect on energy policy stringency of increased
worker (capital owner) coordination costs is high (low). The crucial determinant is the allocation of
the income effect of energy policy. If a lobby carries a large burden from stricter energy policy,
coordination costs will also have a greater marginal effect on the lobbying effort. There is
consequently a negative interdependence between the lobby groups in their response to greater
coordination costs.
We summarize in the following prediction:
Prediction 3: In equilibrium, the effect of coordination costs on energy policy lobbying depends on
the distribution of the direct effect of energy policy on lobby group welfare. When the effect of energy
policy on labor income is relatively large (small), the effect on capital owner income is relatively
small (large), ceteris paribus.
Prediction 3 implies that worker and capital owner bribery are substitutes in the political
pressure for less stringent energy policy.
Our theoretical framework yields three main hypotheses, which we test empirically in this section: (i)
Does corruption reduce the stringency of energy policy? (ii) Do lobby group coordination costs cause
energy policy to become more stringent, in particular where the degree of corruption is high? (iii) Are
worker and capital owner bribery substitutes?
Econometric model
To test the main predictions of the model we analyze variations in energy policy stringency at
the sector level within and between countries. We fit the following model:
Yijt = cij + it + Zijt + Xijt + ijt
where Yijt is the sector specific energy stringency measure for sector i in country j at time t, cij are
time-invariant country specific sector fixed effects, it are sector specific time trends, Zijt represents
the matrix of explanatory variables, Xijt is the matrix of sector or country specific control variables,
and ijt is the normally distributed error term corrected for cross-sectional heteroskedasticity (with ijt
 N(0, ij). Vector  is the set of parameters of interest and  the vector of control parameters.
In order to test whether the explanatory variables significantly differ from zero, we start from
the presumption that sectors are non-homogeneous. The variation in variances across different sectors
across countries is considerable. Therefore we employ a Generalized Least Square (GLS)
specification to correct for the presence of cross-section heteroskedasticity.8 The sample variancecovariance matrix is obtained from a first-stage OLS regression and is iterated until convergence.
Finally, we exploit the panel structure of our data for several robustness tests, both in the country and
sector dimensions. We first discuss our data, its sources and limitations, and next address the links
between theory and our estimation equation.
Data sources and Measurement Issues
Energy consumption, in particular the consumption of fossil fuels, has been a major focus in
environmental policy over the last decades. We chose to use sector level energy efficiency, i.e.
physical energy units per unit of the value added of the sector as our dependent variable.9 We are
restricted to using energy efficiency rather than energy per unit of capital due to data limitations. In
addition this measure has the advantage of: (i) it reflects the regulatory strategies used by OECD
countries (physical energy constraints, efficiency standards, and energy taxes); (ii) is influenced by
corruption; (iii) is available as a large panel data set, (iv) exhibits sufficient industry and country
variation. By restricting energy use per unit of output, firms internalize several air pollution
externalities that are directly related to the combustion of fossil fuels, including smog, acid rain (SO 2
and NOx), and climate change (CO2). Energy intensity per unit of output is the ultimate outcome of
policies that seek to constrain energy use for environmental reasons.
Our unbalanced panel combines OECD sectoral energy use data and economic data from the
Intersectoral Data Base (ISDB) (see the Appendix for details). We are restricted to sector level data.
There is little reason to employ an error component or random effects model as the number of statistical units is large
and sector data are not drawn from a random set as the sample is closed (see Matyas and Sevestre, 1996). Testing for the
equality of variances across the different sectors clearly rejects the assumption of homogeneity employed by the ordinary
OLS estimator. Due to the relative short time dimension we estimated the model using a deterministic trend. Sensitiveness
for estimation in first differences, however, is very limited.
9 Energy efficiency regulations are usually enforced in negotiations between regulator and firms. According to Leveque et
al. (1996), the evaluation of available technologies for energy efficiency improvements on the basis of Best Available
Technologies Not Entailing Excessive Costs (BATNEEC) at the plant level leaves much room for regulatory capture.
Recent climate change policy proposals typically focus on relative caps. An emission or energy constraint can then be
combined with an output subsidy. Finally, by looking at the variation in energy intensity measured by the physical amount
of energy use (tons of oil equivalent) per unit of value added for each sector, our measure is not obstructed by changes in
the sector’s energy price composite.
However, compared to most cross-country studies in this area, our approach has the advantage of
capturing more variation across the estimation units. Compared to the earlier literature, our data set is
particularly large (and focused).10 A sufficient degree of variation across sectors and countries help us
control for important unobservable factors that may influence our main regressors.11 [TABLE 1 and
Explanatory variables
Corruption is difficult to measure. The widely used Transparency Index (CORRUPTION)
from Transparency International (2000) is a good proxy for corruption in both the legislative process
and the enforcement of environmental policy (see, for example, Persson et al., 2000). The index
allows for variation in corruption across rich countries, and, what is very important for our purpose,
also varies over time because indicative measures are available for two periods before 1996. The
CORRUPTION data shows that Denmark, Finland, Norway and Sweden have become less corrupt
over the time period 1982-1997, whereas Belgium, France, Japan and the USA have become more
corrupt. In order to construct a smooth estimate of the long-term trend in corruption we interpolated
the data using the Hodrick-Prescott filter.12
In order to measure the coordination costs of the worker and capital owner lobbies we focus
on sector size, following Olson (1965) and our theory. As a measure of the capital owner lobby’s
coordination costs, we use an industry sector’s contribution to total value added (VALUEADDED%).
For given firm size, larger sectors represent a greater number of firms and greater coordination
Note, however, that we are unable to instrument for corruption using legal origin (used previously in the literature) as
legal origin does not change over time.
11 The data comes from the following seven manufacturing sectors: food, textiles, wood, paper, non-metallic industries,
basic metal industry, manufacturing and transport equipment, and four other sectors: construction, transport, agriculture
and commercial services. See Table 1. The ten countries are: Belgium, Canada, Denmark, Finland, France, Italy, Japan,
Netherlands, Norway, Sweden, UK and USA, and the years are 1982-1997. The level of energy efficiency is particularly
high in the transport sector, the basic metal industry and non-metallic mineral products industry, whereas the commercial
service sector and construction have the lowest levels. Furthermore, the intra-sectoral variation is considerable as well.
12 The Hodrick-Prescott filter is normally used to smooth macro-economic time series to obtain a smooth estimate of the
underlying long-term trend. The smoothing proces minimizes the variance of the of the series around the smoothed series
problems. Our theory also predicts that larger sectors exert greater bribery efforts since a small
change in policy has a greater impact on the returns to capital. They have larger economic stakes, see
for example Potters and Sloof (1996).
As a measure of worker influence and free-riding problems, we use the share of all workers
employed in a particular sector (EMPLOYMENT%). From our theory, larger sectors also have more at
stake in the aggregate, increasing the bribery effort. However, while worker voting power likely play
a role in democracies, our theory ignores voting issues. Workers in larger sectors have greater
influence at the voting booth. In order to capture the non-linearities discussed by the theory, we
include the squared terms VALUEADDED%2 and EMPLOYMENT%2.
As an additional measure of the stakes and lobbying incentives involved, in the robustness
check we furthermore interact both EMPLOYMENT% and VALUEADDED% with a measure of a
sector’s relative stake in energy policy (ENERGY%), where ENERGY% is the share of total energy
used by a sector.
Control variables
Given that environmental quality is a normal good, demand will increase with income, which
we measure by income per capita (GDPPC). Following Antweiler et al. (2001), we assume that
energy policy is likely to respond slowly to pollution problems, and we therefore use the average
GDP per capita of the previous 3-years.
Differences in the structure of energy production across countries, as well as shifts in
production structure through price or technological effects, are likely to affect energy efficiency. We
include domestic energy prices for heavy fuel oil (OILPRICE) and electricity (ELECTRICITYPRICE).
Evidence suggests that the existence of domestic energy resources is associated with lower energy
prices (OECD, 1999). The price of heavy fuel oil should capture common trends in both the oil and
using a penalty parameter on the second difference of the smoothed series. As our objective is precisely opposite, we
choose very low values of this parameter to create a linear trend as close as possible to the given data points.
natural gas markets, while electricity prices usually reflect country specific circumstances.13
Furthermore, we add factor prices such as mean hourly wage of production workers (WAGES) and
cost of capital (CAPITALPRICE). Changing (relative) input prices might induce substitution between
production factors. We expect substitution between a capital/energy composite on the one hand, and
labor on the other hand as capital is usually found to be complementary to energy (see Antweiler et
al., 2001). Whereas we expect OILPRICE and ELECTRICITYPRICE to take a negative sign because
higher energy prices should lead to substitution away from energy as a factor of production, we
expect WAGES and CAPITALPRICE to have the opposite effect on ENERGYEFFICIENCY.
Finally, we use sector specific fixed effects to allow for uncontrolled variables.14,15 We also
allow sectors to shift towards new techniques over time, i.e. we control for endogenous technological
change by adding sector specific trends. Table 2 provides descriptive statistics [TABLE 2 HERE].
Basic findings
Table 3 presents the main results from our estimations [TABLE 3 HERE]. In Model 1 we begin
by including the variables of primary interest, except the interaction terms. This enables us to better
understand the effect of the interactions that emerged in our theory. CORRUPTION has the expected
positive sign and is significant at the 5% level, implying that greater corruption lowers energy
efficiency. Moreover, consistent with our theory, VALUEADDED% has a significant negative sign,
indicating that industry (capital owner) lobby groups encounter increasing coordination problems as
the size of the industry sector increases. However, the marginal impact of coordination costs on
We also experimented with indices reflecting indigenous energy production variables, measured as the share of total
energy use coming from domestic fossil fuels, nuclear power, and green sources (biomass, hydro, wind, and sun).
Consistent with OECD (1998), we find that none of them are important. Results are available upon request.
14 We extensively tested our model with and without country specific as well as sector specific fixed effects, in particular
because variation of corruption is limited. These results indicate that country specific sector fixed effects are to be
preferred (F-test). Results are available upon request.
energy efficiency is declining as VALUEADDED%2 has a significant positive sign, and this also lends
support to our theory. However, EMPLOYMENT% has a significant but positive sign. Apparently,
workers do not suffer from coordination problems in their political activities, contrary to our, and
Olson’s (1965), theory. Instead, the voting power of workers in large sectors may explain this finding.
CORRUPTION*VALUEADDED% is not significant, and the size of the VALUEADDED% coefficient
remains similar to in Model 1. Second, CORRUPTION*EMPLOYMENT% has a significant and
negative coefficient. In addition, the estimated coefficient on EMPLOYMENT% roughly doubles
compared to in Model 1 and still significantly positive. This implies that workers in large sectors have
a smaller impact on energy efficiency in more corrupt countries due to coordination costs. Consider
what our estimates imply, for example, for Italy with the maximum score on CORRUPTION (6.6). A
ENERGYEFFICIENCY by a small amount, the effect is equal to 3.62-0.49*6.6-0.02*(4.7)2 = -0.06.
Recall that our theory suggests that the marginal impact of a reduction in lobbying due to
coordination costs is greater when political influence is great. Our results suggest that political
coordination (collective action) costs may be a greater problem for worker lobby groups in highly
corrupt societies, consistent with our theory.16 Third, the importance of including the interaction terms
is also reflected in the fact that the size of the CORRUPTION coefficient rises strongly as we move
from Model 1 to 2.
In Model 3, we instead add interactions with ENERGY%, which captures the relative
importance of the energy use by a sector. In Model 4 all interactions are included, and this is our
preferred model. The interactions with ENERGY% yield no robust results, although it appears in
For example, we want to capture variations due to different production processes in different countries. For instance,
the paper industry in Finland uses raw pulp, which is more energy intensive than recycled fibers used by the Dutch paper
Model 4 that workers and capital owners again have the opposite effect as sector energy use rises. In
Model 4 CORRUPTION*VALUEADDED% becomes significant at the 1% level, still with a positive
sign. The significant coefficients imply that a unit increase in VALUEADDED% (evaluated at the
sample means) causes a decline in energy use (a reduction of the value of ENERGYEFFICIENCY) by
3.23 units (–4.88 + 0.34*2.1 + 0.20*4.67 = -3.23). Capital owners still suffer from coordination
problems in highly corrupt countries, but the problems are less acute. For capital owners, increased
corruption increases the ability to gain influence.
Model 4 also implies that a unit increase in EMPLOYMENT% (evaluated at the mean) causes
an increase in energy use by 3.88 - 0.02*(4.7)2 – 0.55*2.1 – 0.14*4.67 = 1.63. However, the effect is
less pronounced in more corrupt countries. Evaluated at two std. deviations above the sample means
for corruption, the effect declines to 0.20 (3.88-0.02*(4.7)2 –0.55*4.7 – 0.14*4.7 = 0.20).
Corruption lowers the stringency of energy policy, consistent with our theory. This is
particularly the case in capital intensive sectors, i.e. with high VALUEADDED% and low
EMPLOYMENT%. As an example, the coefficients in Model 4 suggest that a unit increase in the level
of corruption (evaluated at the mean) raises ENERGYEFFICIENCY (i.e. increases energy use) by
(2.82 +0.34*5.0–0.55*4.7 =) 1.94 units.
Concerning the control variables in Table 3, GDPPC has a surprising significant and robust
positive sign across models, implying that richer countries permit greater energy use per unit of
output. CAPITALPRICE and ELECTRICITYPRICE are significant with the expected positive signs in
two models each, and WAGES only in Model 4.17 The controls reflecting factor prices thus appear less
robust than our main variables, indicating political (and income) variables are relatively important for
energy efficiency.
Although not included in our theoretical framework, this result could in addition be due to a lower level of mutual trust
(social capital) among (potential) lobby group members in highly corrupt societies. Lack of social capital would actually
reduce the negative effects of corruption. This is a topic for future research.
In sum, we find that corruption reduces energy efficiency. The direct effect on a sector’s
energy use of larger (potential) worker lobby groups is positive, but the effect of greater capital owner
lobby groups is negative. However, we find evidence that corruption is highly important for the
impact of these political forces. Corruption reduces the influence of the worker lobby, but increases
the influences of the capital owner lobby. We believe this is a new finding in the literature. It may (at
least partially) explain the ambiguous findings in the previous literature.
Robustness analysis
Table 4 presents a robustness analysis which builds on Model 4 from Table 3 [TABLE 4
HERE]. Here we separate between energy intensive sectors and energy extensive sectors. The energy
intensive sectors are: basic metal industry, non-metallic industry, and transport (see Table 1 and
Figure 1). Behavior may potentially differ considerably. We also present results also without Italy,
which by far has the highest level of corruption in the sample.
We first discuss the results including Italy (Models 5 and 7). Our results are substantially
stronger using the energy intensive data in Model 5. The variables of importance remain significant
with the same signs as in Model 4, Table 3, but the size of the coefficients differ widely between
samples. It appears that energy policy in energy intensive sectors is substantially more affected by
corruption and coordination problems.
However, the results excluding Italy (Models 6 and 8) suggest that Italian data do influence
some of the results. The main effect of excluding Italy from the energy intensive sample is that
CORRUPTION becomes insignificant in Model 6, and the interactions with CORRUPTION change
signs (but remain highly significant). Also, contrary to the remaining models, workers now suffer less
coordination problems in more corrupt countries, whereas capital owner lobbying efforts suffer more
coordination problems in more corrupt countries. The latter finding is consistent with our theory. The
It is also interesting to note that our results seem to challenge the findings of Antweiler et al. (2001) on energy-capital
complementarity. Higher (relative) electricity prices tend to induce a reduction of energy intensity. A higher price of
reduction in the amount of variation may influence our results, for example the direct effect of
However, the interactions with CORRUPTION do exhibit sufficient variation to
yield significant effects in our data. The results excluding Italy (for the energy intensive sectors, in
particular) show that the effect of corruption is strongly conditional on EMPLOYMENT% and
VALUEADDED%. Note also that using only the significant coefficients in Model 6, the effect of a
unit change in corruption at the sample means equals 33.9*2.0 - 29.58*2.2 = 2.7, implying that the
aggregate effect of corruption is to raise energy inefficiency slightly also in energy intensive sectors
outside Italy.18
The changes in the results are less drastic when Italy is excluded from the energy extensive
sub-sample. Most variables are very similar in size and significance level as in Model 4 in Table 3.
CORRUPTION is now significant only at the 10% level, however, and the size of the coefficient is
reduced by almost two thirds.
In sum, we find that the level of energy intensity of a sector is important for the strength of the
effect of corruption and coordination costs. However, the direction of most of the effects remain the
same in the different samples.
Finally, among the control variables, we notice that GDPPC is now significant only in the
energy intensive sample, and CAPITALPRICE takes a significant and (expected) positive sign in all
This paper develops a simple theory of the effect of corruption and lobby group coordination costs on
energy policy. The theory predicts that corruption reduces the stringency of energy policy, in
capital leads to lower energy efficiency. These results should be interpreted with caution, however.
18 In this calculation we use the mean of EMPLOYMENT% (2.0) and VALUEADDED% (2.2) of energy intensive sectors
excluding Italy.
particular in large sectors that face great coordination costs. Moreover, lobby group coordination costs
cause energy policy to become more stringent, in particular when the degree of corruption is high.
We test the predictions of the theoretical model using panel data from 14 OECD countries,
and the empirical results lend general support to the theory. First, corruption increases energy waste
by reducing the stringency of energy regulations. Second, capital owner lobbying becomes less
successful in larger sectors, but the effect is reduced in highly corrupt countries. On the other hand,
worker lobbying is more successful in large sectors, unless the country is highly corrupt.
These findings are novel, to our knowledge. They also appear to have important policy
implications. For example, compliance with the Kyoto Protocol may be facilitated by reforms aiming
at reducing corruption in OECD countries.
Ades, A. and R. Di Tella (1999), “Rents, Competition and Corruption,” American Economic Review
89: 982-993.
Aidt, T.S., 1998, “Political Internalization of Economic Externalities and Environmental Policy,”
Journal of Public Economics 69(1), 1-16.
Antweiler, B. Copeland and S. Taylor, 2001, “Is Free Trade Good for the Environment?,” American
Economic Review,
Baldwin, R.E. (1985), The Political Economy of U.S. Import Policy, Cambridge, MA: MIT Press.
Bardhan, P., 1997, “Corruption and Development: A Review of Issues.” Journal of Economic
Literature 35(3), 1320-46.
Bernheim, B.D. and M.D. Whinston, 1986, “Menu Auctions, Resource Allocation, and Economic
Influence.” Quarterly Journal of Economics 101, 1-31.
Cahan and Kaempfer (1992)
Dixit, A., G.M. Grossman, and E. Helpman, 1997, “Common Agency and Coordination: General
Theory and Application to Government Policy Making.” Journal of Political Economy 105,
Eakins and Spick
Eerola, E., 2002, “Forest Conservation – Too Much or Too Little? A Political Economy Model,”
mimeo, University of Helsinki.
Eliste, P. and P.G. Fredriksson (2002), “Environmental Regulations, Transfers, and Trade: Theory and
Evidence,” Journal of Environmental Economics and Management 43(2): 234-50.
Esty and Caves
Finger, J.M., H.K. Hall, and D. Nelson (1982), “The Political Economy of Administered Protection,”
American Economic Review 72: 452-66.
Fisman, R. and Gatti, R. (2002), “Decentralization and Corruption: Evidence across Countries,”
Journal of Public Economics 83: 325-45.
Fredriksson, P.G., J.A. List, and D.L. Millimet (2001), “Bureaucratic Corruption, Environmental
Policy and Inbound US FDI: Theory and Evidence,” forthcoming, Journal of Public
Fredriksson, P.G. and J. Svensson (2002), “Political Instability, Corruption and Policy Formation:
The Case of Environmental Policy,” forthcoming, Journal of Public Economics.
Gardner, B.L. (1987), “Causes of U.S. Farm Commodity Programs,” Journal of Political Economy
95: 290-310.
Gawande, K. and U. Bandyopadhyay (2000), “Is Protection for Sale? Evidence on the GrossmanHelpman Theory of Endogenous Protection,” Review of Economics and Statistics 82(1): 13952.
Godek, P.E. (1985), “Industry Structure and Redistribution Through Trade Restrictions,” Journal of
Law and Economics 28: 687-703.
Grossman, G. and E. Helpman, 1994, “Protection for Sale,” American Economic Review 84(4), 83350.
Heyes, A. and B. Dijkstra (2001), “Interest Groups and the Demand for Environmental Policy”, in T.
Tietenberg and H. Folmer, eds., International Yearbook of Environmental and Resource
Economics 2001/2002, 2001, Edward Elgar, Cheltenham.
Kalt and Zupan
Knack, S. and P. Keefer (1997), “Does Social Capital have an Economic Payoff? A Cross-Country
Investigation,” Quarterly Journal of Economics 112(4): 1251-88.
Laffont, J.-J. and J. Tirole, 1991, “The Politics of Government Decision-Making: A Theory of
Regulatory Capture,” Quarterly Journal of Economics 106: 1089-1127.
Laffont, J.-J. and J. Tirole, 1998, A Theory of Incentives in Procurement and Regulation, Cambridge:
MIT Press.
Leveque, F. (ed.), 1996, Environmental Policy in Europe, Cheltenham (Edward Elgar).
López, R. and S. Mitra (2000), “Corruption, Pollution and the Kuznets Environment Curve,”
Journal of Environmental Economics and Management 40(2): 137-50.
Marvel, H.P. and E.J. Ray (1983), “The Kennedy Round: Evidence on the Regulation of International
Trade in the United States,” American Economic Review 73(1): 190-97.
Matayas and Sevestre
Mauro, P., 1995, “Corruption and Growth,” Quarterly Journal of Economics 110: 681-712.
Oates, W.E. and R.M. Schwab, 1988, “Economic Competition Among Jurisdictions: Efficiency
Enhancing or Distortion Inducing?” Journal of Public Economics 35(3), 333-54.
OECD, 1999
OECD, 2000a, Energy Balances, Statistical Compendium, ed. 01, CD-ROM, Paris.
OECD, 2000b, Intersectoral Database (ISDB), Paris.
Olson, M., 1965, The Logic of Collective Action (Harvard University Press, Cambridge, MA).
Persson, T., G. Tabellini, and F. Trebbi (2000), “Electoral Rules and Corruption,” mimeo, Institute for
International Economic Studies, Stockholm University.
Pincus, J.J., 1975, “Pressure Groups and the Pattern of Tariffs,” Journal of Political Economy 83:
Potters, J. and R. Sloof, 1996, “Interest groups: A survey of empirical models that try to assess their
influence,” European Journal of Political Economy 12, 403-42.
Pugel, T.A. and I. Walter (1985), “U.S. Corporate Interests and the Political Economy of Trade
Policy,” Review of Economics and Statistics 67: 465-73.
Putnam, R.D. (2000), Bowling Alone: The Collapse and Revival of American Community, New York:
Simon and Schuster.
Ray, E.J. (1991), “Protection of Manufacturers in the United States,” in D. Greenaway, R.C. Hine,
A.P. O’Brien, and R.J. Thornton, eds., Global Protectionism, New York, N.Y.: St. Martin’s
Rodrik, D. (1995), “Political Economy of Trade Policy,” in G.M. Grossman and K. Rogoff, eds.,
Handbook of International Economics, Vol. III, 1457-94, Amsterdam, New York and Oxford:
Elsevier Science Publishers, North-Holland.
Schulze, G. and H. Ursprung, 2001, “The Political Economy of International Trade and the
Environment.” In Günther Schulze and Heinrich Ursprung, eds., International Environmental
Economics: A Survey of the Issues. (Oxford University Press, Oxford).
Shleifer, A. and R.W. Vishny, 1993, “Corruption.” Quarterly Journal of Economics 108(3), 599-617.
Tanzi, V. (1998), “Corruption Around the World: Causes, Consequences, Scope, and Cures,” IMF
Staff Papers 45(4):
Transparency International (2000),
Trefler, D. (1993), “Trade Liberalization and the Theory of Endogenous Protection: An Econometric
Study of U.S. Import Policy,” Journal of Political Economy 101(1): 138-60.
Vollebergh, H.R.J. and E. Dijkgraaf, 2001, “A Note on Testing Environmental Kuznets Curves with
Panel Data”, Fondazione Eni Enrico Mattei (FEEM) Research Memorandum 63.2001.
Wei, (2000)
Table 1. ISIC Industries
Manufacturing industries
Food, Beverages and Tobacco
Textiles, Apparel and Leather
Wood, Products and Furniture
Paper and Paper Products, Printing and Publishing PAP
Non-Metallic Industry
Basic Metal Industry
Transport Equipment
Agriculture, Forestry and Fishing
Other industries
61+62+63+72+81+82+83 Commercial Services
Notes: * Denotes sectors classified as Energy Intensive. See also Figure 1.
Table 2. Descriptive Statistics
Mean Std. Dev. Maximum Minimum
Table 3. Results
Adjusted R2
Notes: Generalized LS regressions (cross section weights); dependent variable is ENERGYEFFICIENCY; Standard errors
in parenthesis. ***[**](*) denotes significance at the 1[5](10) percent level. All estimations use sector specific fixed
effects as well as sector specific trends.
Table 4. Robustness Analysis
Adjusted R2
Energy Intensive
Incl Italy
Excl Italy
Energy Extensive
Incl Italy
Excl Italy
Notes: Generalized LS regressions (cross section weights); dependent variable is ENERGYEFFICIENCY; Standard errors
in parenthesis. ***[**](*) denotes significance at the 1[5](10) percent level. All estimations use sector specific fixed
effects as well as sector specific trends.
Figure 1. Average Energy Efficiency by Sector
Energy efficiency
Appendix Sources and Construction of the variables
Source: OECD/IEA Energy Balances and IDSB/STAN
Energy use of sector i in country j relative to its own value added (constant 1990$ against PPP)
Source: Transparency International (
Transformation: values in this paper = 10 – original values; we interpolated the scores for 1986 and
1987 and 1993-1995 from the data for the periods 1980-1985 and 1988-1992 and the figure for 1996.
Interpolation based on Hodrick-Prescott filter.
Percentage of employees of a particular sector i relative to total number of employees in country j
Percentage value added of a particular sector i relative to total value added of all sectors in country j
Source: .....
GDP data per capita are GDP at market prices (constant 1990 US $ against PPP) divided by
population for each country in each year. Our GDP measure is based on the average GDP in the
previous three years.
Source: OECD/IEA Energy Balances
Percentage energy use of industrial sector i in total final energy consumption (TFC) industry in
country j (including feedstock)
Source: OECD/IEA Energy Prices and Taxes
Price of High Fuel Oil Industry (constant 1990 US $ against PPP per Ton)
Source: OECD/IEA Energy Prices and Taxes
Industrial electricity price (constant 1990 US $ against PPP per kWh)
Sector specific wages constructed from the compensation of employees (constant 1990 US $ against
PPP) divided by the number of employees in sector i
Source: IMF Statistics
Country specific real interest rates (GDP deflated) in percentages.
Related flashcards

17 Cards

Marxist theory

16 Cards

Economic systems

38 Cards

Create flashcards