ESSAYS IN CLIMATE POLICY AND INTERNATIONAL TRADE
by
Jong Duk Kim

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Economics
2012

ABSTRACT

ESSAYS IN CLIMATE POLICY AND
INTERNATIONAL TRADE
by
Jong Duk Kim
The dissertation studies the impact of the climate policy in the context of international
trade.
The first chapter of the dissertation studies how the initial allocation of permits could be
determined in an emissions trading system when the location choice of firms is possible. In
the traditional theory, the distribution of permits does not matter in the equilibrium decisions because the relocation of firms or industries to countries with less stringent regulation
is not considered. The theoretical model of the first chapter emphasizes the importance of
permit allocation when the relocation of firms can hurt the welfare of regulated economy
and hence the government may try to prevent the welfare loss by allocating more permits
to firms whose relocation can hurt the economy more. Then, using the data from the EU
emissions trading system (EU ETS) the theoretical predictions are confirmed empirically.
Historical emissions, the level of profits, and fixed costs are the key components in the initial
permit allocation as economic theory predicts. However, estimation results show that the
industrial competitiveness, measured in net imports, does not show a significant influence
on the permit allocation.
The second chapter revisits the pollution haven hypothesis with the data from the EU
emissions trading system. Traditional tests for the pollution haven hypothesis have used
general environmental spending measures such as pollution abatement cost, which does not

properly contain marginal cost implications. The second chapter provides the reason why
permit price interacted with emissions intensity is a better measure for the regulatory stringency theoretically. Empirical results using the data from the EU emissions trading system
add supporting yet weak evidence that environmental regulations has some modest adverse
impact on international trade.
The reason for using the emissions trading system as an emissions reduction tool is that
it guarantees the cost-effectiveness under certain conditions. The third chapter examines
the cost-effectiveness of the permit market by empirically testing the independence property
between the initial permit allocation and the post-trading equilibrium decisions in the EU
ETS. It is argued that in the case of the EU ETS transaction cost is the most relevant
factor that can hurt the independence property. Theoretical model based on Stavins (1995)
predicts how transaction cost affects the post-trading equilibrium. Empirical tests show
that the EU ETS is in the process of settling down; first two years of the EU ETS show a
significant dependence between the permit allocation and the equilibrium decisions, but the
dependence vanishes as the system matures over time.

Copyright by
Jong Duk Kim
2012

To the memory of my father, who passed away in my first graduate year

v

ACKNOWLEDGMENTS

I believe that writing a dissertation is a very solitary process. Without a great support from
others around me, it would be a much more painful journey and take much longer to finish
the dissertation. For that matter, I have to say that I have been very lucky to have such
an exceptional dissertation committee. I would like to deeply thank Professor Jinhua Zhao
for his great guidance and tremendous patience as a major advisor. He always guide me
through murky unorganized ideas to clear firm understanding in economics. Throughout the
dissertation process he has taught me how to think as a serious economist. I owe sincere
and earnest thankfulness to Professor Soren Anderson. At the various stages of papers, he
has read poorly written drafts carefully and given so many valuable comments in detail. I
have no doubt that without the guidance of Professor Anderson my dissertation would much
less understandable. I also would like to thank Professor Susan Chun Zhu for many critical
comments and support. I also owe Professor John Hoehn, the outer committee member, a
debt of gratitude. Discussion with Professor Hoehn is always intellectually enlightening and
helps me understand my dissertation in a broader picture. Last but not least, I would like
to express my deep appreciation about the quiet yet never-resting support from my family.

vi

TABLE OF CONTENTS

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

ix

LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

xi

CHAPTER 1 PROLOGUE . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1

CHAPTER 2 WHAT DETERMINED PERMIT ALLOCATION IN THE
EU ETS? POLITICAL RHETORIC VS. ECONOMIC THEORY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 A Conceptual Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1 Stage 1: Location Choice of a Domestic Firm under the Regulation .
2.2.1.1 The Profit under the Emissions Trading Regulation . . . . .
2.2.1.2 The Profit in the Pollution Haven . . . . . . . . . . . . . . .
2.2.1.3 The Minimum-Required Permits: Keeping firms under the
Regulation . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.2 Welfare Loss and Government’s Selection of Firms . . . . . . . . . . .
2.3 The Data of the EU Emission Trade Scheme . . . . . . . . . . . . . . . . . .
2.3.1
The EU Emissions Trading System (EU ETS) and the National
Allocation Plans . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4 Empirical Analysis for the Initial Permit Allocations . . . . . . . . . . . . .
2.4.1 Empirical Model Specification for Phase I Permit Allocations . . . .
2.4.2 Empirical Results in Phase I Permit Allocation . . . . . . . . . . . .
2.5 Empirical Analysis for Phase II Allocation . . . . . . . . . . . . . . . . . . .
2.5.1 Empirical Model Specification for Phase II Allocation . . . . . . . .
2.5.2 Empirical Results for Phase II Allocation . . . . . . . . . . . . . . .
2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
APPENDIX A . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
A.1 Industrial Classification Table . . . . . . . . . . . . . . . . . . .
A.2 Summary Statistics . . . . . . . . . . . . . . . . . . . . . . . . .
A.3 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
A.4 Permit Allocation Decision under Cournot Competition . . . . .
A.4.1 The Location Choice of Firms . . . . . . . . . . . . . . .
A.4.1.1 Profit under the Emissions Trading System . .
A.4.1.2 Relocation to a Pollution Haven . . . . . . . . .
A.4.1.3 Firm’s Location Choice and Minimum-Required

vii

5
5
10
11
11
14
15
17
20
20
22
23
23
29
33
33
38
43

. . . . . . . 47
. . . . . . . 47
. . . . . . . 48
. . . . . . . 49
. . . . . . . 50
. . . . . . . 51
. . . . . . . 51
. . . . . . . 55
Permits . . 57

CHAPTER 3 THE IMPACT OF EMISSIONS TRADING
FLOWS: EXPERIENCE OF THE EU ETS
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 A Conceptual Model . . . . . . . . . . . . . . . . . . . . .
3.2.1 Supply-side Equilibrium . . . . . . . . . . . . . . .
3.2.2 Demand-side Equilibrium . . . . . . . . . . . . . .
3.2.3 Permit Market . . . . . . . . . . . . . . . . . . . .
3.2.4 Permit Price, Emissions Intensity and Trade Flows
3.3 Empirical Model and Estimation . . . . . . . . . . . . . .
3.3.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . .

ON TRADE
. . . . . . . . . . 59
. . . . . . . . . . 59
. . . . . . . . . . 64
. . . . . . . . . . 64
. . . . . . . . . . 67
. . . . . . . . . . 70
. . . . . . . . . . 71
. . . . . . . . . . 77
. . . . . . . . . . 80
. . . . . . . . . . 82
. . . . . . . . . . 89

APPENDIX B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
B.1 Summary Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
B.2 The Heterogeneity of Sectoral Emissions Intensity of the EU ETS sectors . . 94
B.3 The comparison of ETS and non-ETS sectors . . . . . . . . . . . . . . . . . 95
B.4 Country Classification (GNI per capita) . . . . . . . . . . . . . . . . . . . . 96
B.5 Robustness Checks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
CHAPTER 4 MARKET EFFICIENCY IN THE EU EMISSION TRADING SCHEME . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2 Theoretical Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.4 Identification Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.4.1 Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.4.2 The Control for Endogeneity of Permit Allocation . . . . . . . . . . .
4.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

101
101
105
109
111
111
112
115
126

APPENDIX C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
C.1 Summary Statistics 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
C.2 Summary Statistics 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
C.3 Seeming Correlation between the Initial Allocation and Equilibrium Decisions

129
129
130
131

CHAPTER 5 EPILOGUE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

viii

LIST OF TABLES

2.1

Growing carbon markets around the world . . . . . . . . . . . . . . . . . . . . .

6

2.2

Phase I permit allocation in the EU ETS . . . . . . . . . . . . . . . . . . . . . .

30

2.3

Phase II permit allocation in the EU ETS (sector) . . . . . . . . . . . . . . . . .

39

2.4

Phase II permit allocation in the EU ETS (installation) . . . . . . . . . . . . . .

41

A.1 Industrial classification correspondence table . . . . . . . . . . . . . . . . . . . .

47

A.2 Summary Statistics

48

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

A.3 Definitions of variables

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

49

3.1

Net imports from large trading partner groups . . . . . . . . . . . . . . . . . . .

84

3.2

The impact of emissions trading on net imports with smaller trading partners .

85

B.1 Summary statistics (net imports mil. e(base year=2000)) . . . . . . . . . . . .

92

B.2 Summary statistics (Explanatory variables) . . . . . . . . . . . . . . . . . . . .

93

B.3 Summary statistics of net imports before ETS (ETS sectors, mil e) . . . . . . .

95

B.4 Summary statistics of net imports before ETS (non-ETS sectors, mil. e) . . . .

95

B.5 Net imports from large trading partner groups II . . . . . . . . . . . . . . . . .

99

B.6 Net imports from sub-group trading partners II . . . . . . . . . . . . . . . . . . 100
4.1

Baseline regressions in phase I . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

4.2

Independence property with respect to production at sector level . . . . . . . . . 118

4.3

Independence property with respect to emissions

4.4

Independence property emissions at installation UK

4.5

Annual independence property at installation level (UK) . . . . . . . . . . . . . 123

C.1 Summary statistics at sector level

. . . . . . . . . . . . . . . . . 119
. . . . . . . . . . . . . . . 121

. . . . . . . . . . . . . . . . . . . . . . . . . 129

ix

C.2 Summary statistics at installation level (UK) . . . . . . . . . . . . . . . . . . . . 130

x

LIST OF FIGURES

2.1

Two stage permit allocation game

. . . . . . . . . . . . . . . . . . . . . . . . .

12

2.2

Phase II permit allocations to permit holding positions 1 . . . . . . . . . . . . .

34

2.3

Phase II permit allocations to permit holding positions 2 . . . . . . . . . . . . .

35

2.4

Phase II permit allocations to permit holding positions 3 . . . . . . . . . . . . .

35

3.1

Permit price decision

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

72

3.2

Permit price, emissions intensity and production choices . . . . . . . . . . . . .

74

3.3

Relative demand and relative supply . . . . . . . . . . . . . . . . . . . . . . . .

75

4.1

The volatility of permit price (futures contracts)

. . . . . . . . . . . . . . . . . 124

C.1 Independent of or dependent on allocation?: Initial allocation vs. Production

. 131

C.2 Independent of or dependent on allocation?: Initial allocation vs. Emissions
(sector) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
C.3 Independent of or dependent on allocation?: Initial allocation vs. Emissions (inst.)133

xi

Chapter 1
PROLOGUE

One of the biggest challenges in climate change protection is the fact that nations have strong
incentives for free riding. Even though serious actions are required to prevent the further
progress of climate change, it is undeniable that global cooperation against climate change
has been lacking, especially when each country around the world faces harsh competitions
in the global market. Two largest emitters in the world, the United States and China,
have refused to participate in global cooperations to reduce carbon emissions such as Kyoto
protocol. However, in other parts of the world, some countries have start to implement an
emissions trading system. A notable example is the EU ETS (EU Emissions Trading System)
in which twenty seven European countries agreed to take an action on climate change and
to implement an emissions trading system since 2005. There have been arguments that the
regulations on carbon emissions can generate adverse economic impact on the economy of
adopting countries because there exists a large imbalance in the regulatory stringency across
countries.1 Then, the theme question throughout this dissertation is what is the economic
impact of the emissions trading system as a climate change protection policy in the competitive global market. Furthermore, since the EU ETS has already started the implementation,
another pursued question is what the EU ETS has taught us about the permit allocation,
the economic impact and the cost-effectiveness of the emissions trading regulation.
The first chapter asks how to design the emissions trading system in terms of permit allocation. More specifically, this chapter studies why grandfathering permit allocation method
can be theoretically more justified even when an auctioning method or emissions taxes are
1 See Jaffe,

Peterson and Stavins (1995),and Copeland and Taylor (2003). Also, the special
issue of Climate Policy (volume 6, number 4, 2006) focuses entirely on the competitiveness
topics.
1

available. The principal reason is to prevent the relocation of industries to pollution havens
seeking better profit opportunities. The relocation of industries and hence carbon leakage is
one of the key threats to the success of emissions trading. The relocation of firms or industries can hurt the economy in terms of tax revenues and employment. More importantly, if
relocated firms or industries emit more in the strictly less unregulated countries, the effort
to reduce emissions in regulated countries will be crowded out. Chapter 1 provides a conceptual model that explains how permits should be allocated to minimize the welfare loss
from the overall relocations of industries when the relocation of firms is unavoidable to some
extent. In addition, empirical analysis is performed to find how consistent and convincing
the theoretical predictions are compared with the actual permit allocation in the EU ETS.
In the second chapter, one of the long-lasting questions in environmental economics is
tackled: does emissions trading systems cause adverse impact on trade flows? The second
chapter questions the quality of environmental spending data used in most research testing
pollution haven hypothesis. Theoretically clear message has not empirically shown corroborating evidence that environmental regulation or regulatory stringency causes adverse impact
on trade flows. One of the culprits suspected is the quality of data used in research. Corresponding to that criticism, the second chapter rather uses permit price from the EU ETS
and emissions intensity, which have better theoretical supports. Heckscher-Ohlin type general equilibrium model incorporated with permit market is built to see the relative impact
of the emissions trading on regulated industrial sectors to non-regulated sectors. The conceptual model gives the clear marginal cost implications derived from the changes of permit
price. Furthermore, the importance of emissions intensity as an aptitude of an industry to
the regulation is emphasized. For the same permit price, each industrial sector with different emissions intensity may feel differently about the regulatory stringency and hence their
decisions on dealing with the regulation may differ too. Therefore, it is argued that the interaction between permit price and emissions intensity plays an important role to the optimal
2

decisions of regulated industries. Furthermore, Chapter 2 asks about the relative impact of
the emissions trading regulation between regulated and unregulated industrial sectors in the
general equilibrium framework. With these new and theoretically supported variables and
data from the EU ETS, the relative changes in trade flows of regulated industrial sectors are
estimated with various trading partner groups.
Then, one of the main issues is transaction costs in permit trading. Chapter 3 tackles that
problem in the EU ETS. The reason policy-makers prefer emissions trading over emissions
tax is that flexible permit price makes emissions trading more cost-effective than emissions
tax if the permit market is efficient. Otherwise, operating emissions trading becomes questionable as an emissions reduction policy in a sense that the implementation of the emissions
trading regulation causes many adverse economic impacts. For example, in Chapter 2 the
adverse impact of the emissions trading regulation on trade flows is studied. In the third
chapter, the question about the cost-effectiveness narrows down to whether transaction cost
has a significant impact on the post-trading equilibrium decisions. It is also discussed why
transaction cost is the most relevant factor in the case of the EU ETS among many others
that can hurt the permit market efficiency. A conceptual model based on Stavins (1995)
explains how transaction cost affects the post-trading equilibrium decisions. Then, using the
data from the EU ETS the impact of the initial permit allocation on output and emissions
is empirically tested. The important empirical issue in testing the influence of permit allocation on equilibrium decisions is also addressed in Chapter 3. Since grandfathering is used
as the permit allocation method in the EU ETS, the allocation decisions are made based
on economic and political factors, which may entail biased results. Without controlling for
those factors, it is argued that the empirical study about the impact of permit allocation
may face endogeneity problem. In order to control for the endogeneity problem in the permit
allocation, the empirical specification from the first chapter is adopted.
The EU emissions trading system is by far the largest multi-national multi-sectoral emissions
3

regulation in the world. There has already been a big argument whether phase I of the EU
ETS has not worked as it was expected to. One thing sure, though, is that the success in
implementing the EU ETS may encourage other countries to adopt similar carbon reduction
systems.

4

Chapter 2
WHAT DETERMINED PERMIT ALLOCATION IN THE EU ETS?
POLITICAL RHETORIC VS. ECONOMIC THEORY

2.1

Introduction

Emissions trading system (ETS) is nowadays the most important market-based method to
curb anthropogenic carbon emission in many regions of the world because of its well-known
cost effectiveness. There are five pollution permit markets and at least two outstanding
project-based mechanisms in the world. In 2009, the size of the entire carbon market reached
8,700 million metric ton of CO2 (MtCO2), which is worth more than US$ 140 billion (See
Table 2.1). However, with the growing popularity of the emissions trading system, there are
also growing political voices about industrial comparative disadvantage and carbon leakage
from the countries that implement environmental regulations. Basically, these concerns arise
because emissions trading systems elevate the cost burdens of regulated firms in the participating regions and only a few countries have adopted equivalent environmental regulations
(See Table 2.1). These concerns have led most of emissions trading system to start with the
grandfathering permit allocation method so that governments can distribute more permits
to firms or industrial sectors which are considered as the most damaged. However, allocating
more permits cannot save firms or industrial sectors from losing their cost-competitiveness.
Furthermore, it is well-known that auctioning permit allocation is theoretically better because it prevents heavy emitters from making windfall profits and the government revenue
from selling permits. Then, the questions are when and why grandfathering permit allocation can be theoretically justified, what factors should be considered in the initial permit
allocation, and whether we can find evidence about the strategic permit allocation in the Eu5

ropean Union emissions trading system (EU ETS). However, the purpose of this paper is not
to compare permit allocation systems, but focuses on grandfathering permit allocation itself.

Table 2.1: Growing carbon markets around the world
2008
Volume
Value
base unit
mil.
US$ milMtCO2e
lion
Permit Markets
EU ETS (Europe)
3,093
100,526
GGAS (Australia)
31
183
CCX (US)
69
309
RGGI (US)
62
198
AAU (Europe)
23
276
Subtotal
3,278
101,492
Spot & Secondary Kyoto Offsets
Subtotal
1,072
26,277
Project-based Transactions
Primary CDM (Europe)
404
6,511
JI (Europe)
25
367
Voluntary Market (Europe & US) 57
419
Subtotal
486
7,297
Total
4,836
135,066

Volume
mil.
MtCO2e

2009
Value
US$ million

6,326
34
41
805
155
7,362

118,474
117
50
2,179
2,003
122,822

1,055

17,543

211
26
46
283
8,700

2,678
354
338
3,370
143,735

Source: World Bank, and Bloomberg New Energy Finance and Ecosystem Marketplace for
data on the voluntary market. Recited from State and Trends of the carbon market 2010
(World Bank).
Note: mil. MtCO2e = million metric tons of CO2 equivalent. EU ETS = EU emissions
trading system. GGAS = New South Wales greenhouse gas reduction scheme. CCX =
Chicago climate exchange. RGGI = Regional greenhouse gas initiative. AAU = assigned
amount unit (Kyoto carbon credit). CDM = clean development mechanism. JI = joint
implementation.
The answer for the first question, when and why grandfathering permit allocation can be
theoretically justified, starts with a brief literature review. Croker (1966) and Dales (1968)
suggested a market system for transferable emission permits based on the idea of Coase

6

(1960). Emissions trading system works in the following way. In a given compliance period,
firms that exceed individual CO2 emissions targets designated by their government have
to buy permits from firms that emit less than their designated targets. At the end of the
compliance period, the aggregate volume of emissions should be equal to the initial overall
permit allocation and hence the emissions target of the economy is met. Rigorously proven
by Montgomery (1972), initial allocation does not affect the equilibrium behavior of firms
under certain conditions (i.e. no transaction cost, no market power, no market uncertainty)
and achieves the emission reduction target at the minimum cost. Therefore, the distribution
of permits is not an important issue. One of the key assumptions that make permit allocation irrelevant is that every emitter faces the same emissions regulation and no one can
walk out of the regulation. This is the point where the reality deviates from the traditional
theory. There is the imbalance of the regulatory stringency across countries and firms can
move from one country to another seeking better profit opportunities. The implementation
of the emissions trading system implies higher marginal costs of regulated firms and hence
deteriorates their cost-competitiveness in the global market. If the regulation hurts the profitability of firms very much, then regulated firms can relocate to other countries with lax
emissions regulations. Since the relocated firms will emit more than they used to under the
regulation, this phenomenon of carbon leakage can crowd out the emission reduction effort.
In addition, the relocation of firms entails the welfare loss of the economy. Thus, one of
the primary actions governments should consider to take is how to prevent carbon leakage
effectively. This paper argues that that can be achieved using the grandfathering permit
allocation method.
Then, the second question slides in: what factors are important in the initial permit allocation? The answer is related to how to minimize the welfare loss from the relocation of
firms. Three critical factors are suggested in this paper: the level of emissions, profitability and mobility. First, the basic principle of the grandfathering permit allocation is that
7

permits should be allocated in proportion to the level of emissions. This base rule may be
able to make the complaints from industries or political interest groups tone down about
the allocation rules. However, it does not guarantee to minimize the welfare loss from the
relocation. The second and third factors are the ones that allow the government to minimize
the welfare loss. The profits of firms are part of the welfare of an economy and hence the
relocation of high profit firms hurts the economy more than low profit firms. Keeping high
profits firms under the regulation by allocating more permits is beneficial to the economy.
It is more likely that the government allocate more permits than the base line allocation
calculated from the level of emissions. On the other hand, there are some firms under the
regulation whose mobility is relatively low. Reasons that make firms less mobile are fixed
cost, transport cost and market accessibility (Ederington, Levinson and Minier (2005)). As
to less mobile firms, even if they are heavy emitters with high profits, the government will
allocate less permits than the base line allocation calculated from the level of emissions.
Saved permits will be allocated to other mobile firms. In all, the level of emissions is the
base line allocation and based on the profitability and mobility of firms the permit allocation is adjusted. By doing so, the government can allocate permit as many firms as it can
under the constraint of emissions target. This paper argues that the welfare loss from the
relocation of firms can be minimized.
The third question is related to finding empirical evidence from the experience of the EU
emissions trading system. In other words, this study tests the significance of three theoretically suggested factors empirically and the results are consistent with the theoretical
expectation. Furthermore, even though the theoretical model developed in this paper suggests three factors that should be considered in the initial permit allocation, there is another
factor strongly urged by political interest groups or public mass media: the industrial comparative (dis)advantage. The comparative disadvantage of an industrial sector, measured in
net imports, is not a problem that can be resolved by the permit allocation. It is because
8

the increase in the marginal cost is not avoidable once the emissions trading regulation is
adopted. Thus, even though allocating more permits to weak (less cost-competitive) industrial sectors is politically popular, it may not be an economically sound decision. The
empirical results also confirm that the industrial competitiveness does not have an significant
impact on the initial permit allocation.
The first two questions are theoretical. This study uses a two stage non-cooperative game
scheme. In the first stage, the government starts the emissions trading regulation and decides how many permits will be allocated to each firm based on the welfare changes from the
regulation. In the second stage, firms decide whether to stay under the regulation or to move
to another country with lax or non-existing emissions regulation based on the profits they
can make in each location. Through these steps, the minimum-required permit allocation
equation that can keep a firm under the regulation is obtained. There are previous literature
related to the environmental policy and the location choices of firms. Most of them focus on
the governments’ competition in local pollution standards or taxes in non-cooperate game
settings. (i.e. Markusen, Morey and Olewiler (1995), Barrett (1994), and Ulph (1992)).
Even though this study adopts a similar non-cooperative game setting, it is different from
these papers. First, there is no competition between governments. The interest of this study
lies on the case where a government operates a unilateral emissions regulation via emissions
trading. Second, the emissions regulation is performed with emissions trading rather than
emissions taxes or standards. Different policies have different implications and not many
papers have studied emissions trading in this setting. Third, permit allocation has not been
talked in the setting of location choice problem. Since Montgomery (1972) proved the irrelevance between the initial permit allocation and post-trading equilibrium under the closed
economy assumption, the permit allocation has not been discussed much. Fowlie (2010) discusses a dynamic updating of permit allocation based on output level. She addresses a rebate
rate at which output-based permit allocated is updated. She investigates how the welfare
9

varies at different levels of the rebate rate and shows that the benefit of output-based updating may exceed the costs when domestic producers compete with firms in less stringently
regulated countries. However, she does not explain how the rebate rate is determined. In
other words, attention is not paid to individual permit allocation decisions.
There are some potential criticisms about this study. Since this study adopts a gametheoretic partial equilibrium model, the general equilibrium effects are not thought through.
For example, Goulder, Parry, Williams III and Burtraw (1999) argue the pre-existing environmental policies can eliminate the cost-advantage of the permit market and Hanemann
(2010) argues that emissions trading needs to be accompanied by complementary regulatory
measures.
This paper proceeds as follows. In the next section, a theoretical model for permit allocation
is developed in the non-cooperative game setting and then an equation for minimum-required
permit allocation is derived. In section 3, the EU Emission Trading Scheme is briefly introduced and data set is explained. In section 4, based on the permit allocation equation,
empirical analyses are performed to see whether our predictions can be confirmed in the
initial allocation of the EU ETS. In section 5, empirical analyses are performed to see how
existing and new information obtained through phase I affects the second phase permit allocation given initial allocation. Then, empirical results are provided. Finally, section 6
concludes.

2.2

A Conceptual Model

The conceptual model adopts a two stage non-cooperative permit allocation game between
the government and regulated firms. In the first stage, permit allocation decision to each
firm is made based on the welfare changes from their relocation to other countries. The
government has a bigger incentive to allocate more permits as the welfare loss from the
10

relocation of a firm gets larger. In the second stage, firms choose their location based on
the profits they can make in each location. Under the regulation, firms have to face the increase of marginal cost due to the emissions trading regulation. In another country with no
environmental regulation, known as the pollution haven, firms can avoid the marginal cost
increase. However, they have to face a new fixed relocation cost from the relocation. The
highest profit among other locations plays an role as reservation profits or the opportunity
cost. Throughout the two stage process, the equation for minimum-required permits that
are required to keep a firm under the regulation is derived. In the usual manner, the game
is solved backward from the second stage to first stage. The following Figure 2.1 exhibits
the structure of the game.

2.2.1

Stage 1: Location Choice of a Domestic Firm under the Regulation

2.2.1.1

The Profit under the Emissions Trading Regulation

There is country R among a large number of countries. There is a continuum of polluting
industrial sectors in country R, indexed by w ∈ [0, M ]. Each industrial sector is represented
by one firm. There is no transport cost between countries and hence every firm in the same
industrial sector faces the same demand function (the preference of consumers around the
world is assumed to be the same). Firms serve domestic market first. If the level of the
production exceeds the consumption of domestic market, then they also serve other markets
in the world. Country R decides to implement the emissions trading system as a carbon
emissions regulation. Other countries have no regulation on emissions.1 The target of carbon
¯
emission of country R, E, is exogenously determined by an international treaty or a social
1 The

model can be easily extended to the case of differential regulatory stringency across
countries.
11

Figure 2.1: Two stage permit allocation game
Note: In the first stage, the government decides whether it allocates the minimum-required
permits to keep a firm under the regulation based on the welfare loss from the relocation of
the firm. Hence, if the government decides to allocate the minimum-required permits, then
profits in A and B are the same. In the second stage, firms decide whether to stay under
the regulation or to move to the pollution haven based on the profits they can make in each
location. If the permits are allocated, then the domestic firms will stay because the profits
are the same in each location. If the permits are not allocated, then the domestic firms will
choose the location based on the profit comparisons in each location.
agreement. The target is also the amount of permits country R can issue. The government
decides to use grandfathering as the permit allocation method. Hence, firm i in industrial
sector ws is endowed with Ai units of permit to pollute for free. Then, given the (expected)
permit price, τ , the profit function of each firm under the regulation is as follows. For
simplicity, sector index ws is dropped unless the omission causes confusion.

πi = p(X)xi − Ci (xi ) − Bi (ai ) + τ (Ai + ai − ei (xi )),

(2.1)

where xi is the output of firm i in a regulated sector; X is the world output of sector s

12

(X =

K
k=1 xk ),

where K is the number of firms in sector ws ; p(X) is the output price

corresponding to world output; Ai and ai denote the number of permit allocated and abatement of firm i, respectively; and ei (xi ) denotes the level of emission, which depends on the
production of firm i. Profit πi of firm i is composed of four parts; sales revenue at the world
price p, the production cost, C(·) (C > 0 and C

≥ 0), pollution abatement cost, B(·)

(B > 0 and B ≥ 0), and compliance cost of emissions trading system. On the other hand,
there are other firms in sector ws in other countries, which do not have any environmental
regulations. Then, profits of firm j ∈ ws that are located in non-regulated countries are as
follows:

πj = p(X)xj − Cj (xj ),

(2.2)

where X is the world production of good s and xj denotes the amount of production by
firm j in sector s in another country with no environmental regulation. The production cost
structure of firm j is the same with firm i.
Given the permit price and decisions of other firms in sector ws , firm i makes the optimal
decision on output and pollution abatement. For firm i the first order conditions are as
follows.
dX
∂Ci
∂e
dπi
= p(X) + p
xi −
−τ i =0
dxi
dxi
∂xi
∂xi

(2.3)

∂πi
∂B
= − i + τ = 0,
∂ai
∂ai

(2.4)

and

where the magnitude of

dX
dxi

depends on market structures. Using the concept of conjec-

tural variation we can include many different and familiar models in the same frame. For
example,

dX
dxi

= 1 implies Cournot competition, whereas
13

dX
dxi

= 0 implies Bertrand

competition. Furthermore, if treated as parameters, the conjectural variations can describe
the idea of varying degrees of competition.2 The optimal output and abatement of firm
i under the emissions trading are denoted as xET and aET , where superscript ET de¯i
¯i
note emissions trading. In addition, equilibrium outputs of individual firms are denoted
¯
as xET = (¯1 , .., xET , .., xK ). Then, the equilibrium price and corresponding profit to the
¯
x
¯i
optimal choices is denoted as pET and πi (¯ET ), respectively.3
¯ ET x

2.2.1.2

The Profit in the Pollution Haven

Country H, a pollution haven, has no environmental regulations and so regulation-related
compliance costs do not exist. However, the migration to country H from country R causes
relocation fixed cost, Gi , to the migrating firm i. Fixed cost G includes not only moving
costs but other costs when a firm moves to other country; the disposal cost (the cost of
building new facilities less any salvage value in the regulated country), the cost caused by
the difference of judicial system, languages, or cultures, the cost to new market research,
and so on. Therefore, we understand G as a permanent increment of fixed costs when a firm
moves to another country.4 Then, the corresponding profit function when firm i moves to a
pollution haven becomes:

πi = p(X)xi − Ci (xi ) − Gi ,

(2.5)

Hence, we can obtain the maximum profit firm i can earn if it migrates to country H
2 For

more explanations, see Dixit (1986) and Kamien and Schwartz (1983)
specifically, the output price p is determined as follows.

3 More

∂Ci
∂xi |xET
¯i

∂e

+τ ∂xi | ET
¯
i x
i

ET
1
1 − ET dXET
dxi

−1

.

ET

pET

=

denotes the price elasticity when firm

i is under the regulation.
4 There are other factors that make firms less mobile such as transport cost or agglomeration economies. See Ederington et al. (2005).

14

from the first-order condition as follows.
dX
dπi
∂Ci
= p(X) + p
xi −
=0
dxi
dxi
∂xi

(2.6)

From the first order condition when firm i is located in the pollution haven the optimal
output xP H is derived, where superscript P H denotes pollution haven. Then, the equilib¯i
rium price and corresponding profit to the optimal choices is denoted as pP H (X P H ) and
πi H (¯PH ), where X P H is the aggregate output and xPH is K-tuple equilibrium outputs of
¯P x
¯
individual firms (K is the number of firms in sector ws ) when firm i is located in the pollution
haven. Note that the output prices are different depending on the location choice and the
market structure. Since the relocation from the regulated country to non-regulated country
changes the marginal cost structure of firm i, the output prices vary depending on the market structure. For instance, in the Cournot-type competition with a small number of rivals,
the relocation of firm i changes the output price significantly. However, as the number of
rivals increases and so the market structure converges to the perfect competition, the impact
of the relocation on output price becomes less significant. Similarly, in the Bertrand type
competition, the price changes by the location choice of firm i depend on how homogeneous
goods are produced in the industrial sector firm i belongs to.

2.2.1.3

The Minimum-Required Permits: Keeping firms under the Regulation

ET
In our framework, firm i will decide its location by comparing profit πi it can earn under
P
the emissions trading system with the reservation profit, πi H , when it migrates to country
ET
P
H with no emission regulations. Therefore, if πi ≥ πi H , then firm i has no reason to

change its current location; otherwise, it will change its location to the pollution haven,
ET
P
country H. If πi < πi H and the government wants to keep firm i under the regulation

for some reason, then the government can allocate permits that equalizes the profits in both

15

countries. Then, by comparing profits at each location we can obtain the minimum-required
permits to keep firm i in the regulated country R. The minimum-required permits are the
number of permits that equalizes the profit that firm i can earn in country R with the profit
it can earn in the pollution haven.

1
¯
[ˆP H xP H − pET xET − Ci (¯P H ) + Ci (¯ET )]
p
¯i
ˆ ¯i
xi
xi
Ai =
τ
aET
1
ET Bi (¯i ) − τ + e(¯ET ) − G
+
ai
¯
xi
i ,
τ
aET
¯i
where

pET
ˆ

=

∂Ci
∂xi |xET
¯i

∂e
+τ ∂xi | ET
¯
i xi

1−

1 dX ET
ET dxET
i

(2.7)

−1

is the price under the regulation

and
pP H
ˆ

=

∂Ci
∂xi |xP H
¯i

1−

1 dX P H
P H dxP H
i

−1

is the price under no regulation.

The minimum permit requirement is the minimum number of permits that can deter
the relocation of firms and so keep firms under the regulation if necessary. The followings
show how minimum-required permits vary when each component of the formula changes.
The first term in the square bracket states that as the production margin of firm i (total
revenue less variable production costs) in the pollution haven gets larger relative to that
in the regulated country, the firm will be assigned more permits. Higher unit abatement
cost relative to permit price also brings more permits to firms. The third term tells that
as expected emission under the emissions trading system gets larger, the firm will get more
permits. On the other hand, as the fixed cost of a firm grows and so the mobility of a firm
gets lower, there is a less chance to get more permits.

16

2.2.2

Welfare Loss and Government’s Selection of Firms

This section explains how the government selects firms to keep under the regulation. The
basic principle of permit allocation is to minimize the welfare loss per permit from the reloca¯
tion of firms. In other words, permits should be allocated conditional on emissions target E.
The government first selects firms whose relocation causes the larger welfare loss per permit,
gives the minimum-required permits to those firms, and make their relocation unnecessary.
Hence, the following explains how to measure the welfare loss from the relocation of firms
and how to minimize the welfare loss per permit given the minimum-required permits.
This paper assumes that the overall emission target is determined by an international treaty
(or binding social agreement) and so a government cannot decide the total number of per¯
mits at its own discretion. Furthermore, the emission reduction target, E, is stringently
determined so that it is a binding constraint. Then, some firms confronting the hike of cost
burden may want to relocate their production facilities in other countries with no environmental regulations. What government of country R can do now is to screen firms, assess their
social values and decide which firms to keep and which firms to let go. The whole process
starts with the welfare function of country R. In other words, the decision on whether a
government allocates the minimum-required permits depends on the welfare impact a firm’s
relocation can causes.
As usual, the social welfare function is the sum of producer surplus and consumer surplus.
In addition, environment-caring country R suffers from the global carbon emission, E GE .
Hence, the environmental damage from the global carbon emission should also be considered.
Note that carbon emissions is a global emission so that emission from wherever in the world
affects the welfare of country R. Then, the social welfare of country R will be as follows:

W = (producer surplus) + (consumer surplus) − (damage from global emission)

17

More specifically, the benchmark social welfare across all polluting industrial sectors when
no firms move to pollution haven is:

u(x(w)∗ET ) − p(w)ET x(w)∗ET (dw) − D(E GE ),

π(w)ET dw +

W =
w

(2.8)

w

where u is the utility function of the representative consumer and D(·) is the damage function
from carbon emission (D > 0 and D ≥ 0). x(w)∗ET denotes the equilibrium consumption
when industrial sector w is under the emissions trading. p(w)ET is output prices when firms
in country R are under the regulation. Now, under emissions trading, we think of the case
ET
P
that firm i changes its location to country H when πi < πi H . In the case that firm i

moves to pollution haven, the new social welfare will be:

Wi =

w

π(w)ET dw − πi (ws )ET +

u(x(w)∗ET ) − p(w)ET x(w)∗ET (dw)

(2.9)

w

+ u(xi (ws )∗P H ) − p(ws )P H xi (ws )∗P H − D(E GE − e(X(ws )∗ET ) + e(X(ws )∗P H ))
In words, if firm i in sector ws moves to a pollution haven, the following changes occur.
First, the profit of firm i no longer belong to the welfare of country R. πi (ws )ET is the
welfare country R would lose if firm i moved to the pollution haven. Second, there will
be a change in consumer’s surplus caused by output price change from p(X(ws )ET ) to
p(X(ws )P H ). x(ws )∗P H denotes the new equilibrium consumption of industrial sector ws .
Third, the output of industrial sector ws that firm i belongs to will change and hence the
global emission will change as well. (−e(X(ws )∗ET ) + e(X(ws )∗P H )) is the net changes in
global emissions.5
5 It

is assumed that emissions among industrial sectors are independently decided and so
there is no emission change from other industrial sectors due to the relocation of firm i if all
others are equal.

18

Formally, we can define the welfare loss, li , caused by the migration of firm i in the following
way.

ET
li ≡ −(Wi − W ) = πi + ∆Di − ∆Si ,

(2.10)

where ∆Di and ∆Si denote the difference of environmental damage and the change in
consumer surplus caused by the relocation of firm i.6 Once firm i moves to the pollution
haven, the profit country R used to have is no longer in the welfare; the environmental damage
changes depending on the changes in emissions level; and consumer’s surplus changes too.
Since the minimum-required permits to keep each firm under the regulation are known
to the government, the goal of the selection process will be to minimize the total welfare
loss per permit allocated. The government ranks each firm by the welfare loss per permit
caused by the relocation and then allocate the minimum-required permits first to the firm
that causes the highest welfare loss per permit. Then, the government allocates minimumrequired permits to the firm that cause second highest welfare loss per permit. This process
¯
carries on until the total number of permits allocated reaches the emissions target, E. In
other words, by giving permits first to the firms that can cause larger loss per permit and so
preventing them from the relocation, the government can minimized the total welfare loss.
¯
Λ(A) is defined as the value of prevented-welfare-loss by distributing permits. Formally, the
objective function of the government selection process is to

Λ = max
wj

1[wj ] ·

l(wj )
¯
A(wj )

dwj

(2.11)

subject to

l(wj ) > 0 and
6 See

¯
¯
1[wj ] · A(wj )dwj ≤ E,

appendix A.4 for the Cournot case results
19

(2.12)

where I[wj ] is an indicator function of permit allocation decision on industrial sector wj .
I[wj ] = 1 if the government selects industrial sector wj and allocates the minimum-required
¯
permits A(wj ); I[wj ] = 0 otherwise and then no permit will be allocated.
The selected sectors are allocated with the minimum-required permits, which equalize the
profits in each location, and hence the selected firms will not move to the pollution haven
by design. On the other hand, non-selected sectors will not have permits allocated. Not all
non-selected sectors will move to the pollution haven if their mobility is fairly low. Since
there is no hidden information so that the government’s conjecture about permit price is
correct, the whole permit allocation process is conditional on equilibrium permit price.
The objective function tells that the government will try to keep as many firms as the
target number of permits cover and to minimize the welfare loss by giving permits with higher
priority to firms that can cause larger welfare loss. The first constraint (lj > 0) implies that
there is no reason to allocate permits to firms whose relocation actually increases the welfare.
The second constraint states that the total number of permits allocated should not exceed
the emission target.

2.3
2.3.1

The Data of the EU Emission Trade Scheme
The EU Emissions Trading System (EU ETS) and the National Allocation Plans

In order to commit themselves to fight against climate change in a cost efficient way, EU27
countries have agreed to set up an internal market, named EU Emission Trading Scheme
(EU ETS), enabling firms to trade carbon dioxide pollution permits. Since January 2005,
more than 10,000 large industrial plants in the EU have been required to trade permits to
emit CO2 into the air. Industrial sectors covered by the EU ETS include: combustion instal-

20

lations, mineral oil refining, coke ovens, metal ores, iron and steel, cement and lime, glass,
ceramics, and pulp paper and board. The first phase of the EU ETS operated from 2005 to
2007 with 25 starting members first. The second phase of the EU ETS runs from 2008 to
2012, which coincides with the first commitment period of Kyoto Protocol. The empirical
analysis of this study covers the allocation-related data of seven EU ETS sectors in twenty
four member states in both phase I and phase II.
Now, we need to discuss further about the permit allocation in the EU ETS. An emission
target for each installation is designated via so-call National Allocation Plan (NAP) submitted by member states and approved by the European Commission. The NAP of each
member state must stipulate the total quantity of permits that a member state intends to
issue during that phase and how it proposes to distribute those permits among participating
installations. The main permit allocation method is grandfathering. For the first phase,
member states should allocate at least 95% of the permits free of charge. For the second
phase, member states should allocate at least 90% of permits free of charge.
Annex III to European Union (2003) contains criteria relating to the national allocation
plans. The relationships between these criteria can be revealed by categorizing them various
ways. The Directive explicitly states that the national allocation plans may contain information on the manner in which the existence of competition from countries or entities outside
the European Union will be taken into account. In addition, the Directive also states that if
competition from outside the European Union is taken into account in the national allocation
plans, the criterion should only be applied in determining the quantity of permits allocated
at activity level, without changing in the total quantity of permits. In other words, if a government concerns about the competitiveness of industrial sectors, then without changing the
total quantity of permits of the economy it can adjust permit allocation among participating
sectors and installations.

21

2.3.2

Data

The data of this study cover seven industrial sectors in twenty four European countries participating in the EU emissions trading system. Three layers of information (sector, firm and
installation) are also included. The phase I permit allocation from 2005 to 2007 is investigated. Historical data for explanatory variables range from 1999 to 2002. Unlike sectoral
data, most of firm and installation level data are rare before 2005. By far, the UK is the only
country that has kept and publicized historical data before 2005 in the installation level.
Basically, permit allocation data can be found in the National Allocation Plans. The EU’s
Community Independent Transaction Log (CITL) has also kept detailed data for individual
installation allocation and actual (verified) emission since the EU ETS started in 2005. For
explanatory variables, historical emission data at the sector level have been kept by the
United Nations Framework Convention on Climate Change (UNFCCC) since 1990.7 Most of
sector level industrial data other than emissions are obtained from Eurostat, which is the
official institute managing statistics in European countries. Eurostat keeps gross operating
surplus, production level, fixed cost variables (gross investment in tangible goods and net
investment in tangible goods), and number of employees. This paper uses net import (import
7 In accordance with Articles 4 and 12 of the Climate Change Convention,

and the relevant
decisions of the Conference of the Parties, countries that are Parties to the Convention submit national greenhouse gas (GHG) inventories to the Climate Change secretariat. These
submissions are made in accordance with the reporting requirements adopted under the
Convention, such as The UNFCCC Reporting Guidelines on Annex I Inventories (document
FCCC/SBSTA/2004/8) for Annex I Parties and Guidelines for the preparation of national
communications for non-Annex I Parties (decision 17/CP.8). The inventory data are provided in the annual GHG inventory submissions by Annex I Parties and in the national
communications under the Convention by non-Annex I Parties.
The GHG data reported by Parties contain estimates for direct greenhouse gases, such as:
Carbon dioxide (CO2); Methane (CH4); Nitrous oxide (N2O); Perfluorocarbons (PFCs); Hydrofluorocarbons (HFCs); Sulphur hexafluoride (SF6) as well as for the indrect greenhouse
gases such as SO2, NOx, CO and NMVOC. Currently, only CO2 among these greenhouse
gases are traded in the EU ETS.

22

less export) as the measure for the industrial competitiveness following literature. However,
there is a discussion about what is the proper measure for the industrial competitiveness (See
Jaffe et al. (1995) and OECD (1993) for further discussion.). At installation level, the data
availability other than emissions is very limited. Carbon Market Data database provides the
link between installations and firms.8 It also provides ownership information in each firm.
Then, there is a difficulty matching the EU ETS sector classification with other existing
industrial classification. This paper follows European Environmental Agency (2007), which
studies the classification between the EU ETS sectors , the Common Reporting Framework
of the Intergovernmental Panel on Climate Change (IPCC) (IPCC (1996); Intergovernmental Panel on Climate Change (2006)), and International Standard Industrial Classification
(ISIC). According to European Environmental Agency (2007), the sum of emissions in the
GHG inventory from the relevant CRF categories is always higher than the verified emission
from the EU ETS because inventory includes all plants and does not use any threshold criteria for the inclusion of installations as in the EU ETS. The calculated share of the ETS
total in the CRF ranges from 75% to 96% and the average share of the EU ETS total in the
CRF total emissions is 85.4% for EU23. The concordance among classifications are shown
in Appendix A.1.

2.4
2.4.1

Empirical Analysis for the Initial Permit Allocations
Empirical Model Specification for Phase I Permit Allocations

Each participating country in the EU ETS already has its National Allocation Plan with
detailed (mostly verbal) explanations. The intention of this study is to pull out rather consistently agreeable general allocation rules that can embrace heterogeneous national allocation
8 http://www.carbonmarketdata.com/

for more information.

23

plans. The theoretical model shows that the cores of permit allocation narrow down to three
components: the level of emissions, profitability, and mobility. Net import and the number of
employees are included to measure the industrial competitiveness and the potential political
pressure from the interest groups. Three different layers of data (sector, firm and installation level) are used for the empirical analysis. Since the data availabilities of covariates
are different at data levels, the empirical specifications at each level differ slightly. The first
regression model, implied by the theoretical model provided in equation (2.7) in section 2.2,
is to investigate how initial permit allocation decision at sector level is made in the EU ETS.
This equation implies that the projected emissions under the regulation, profits, and unit
abatement should increase permit allocation. On the other hand, higher fixed costs hinder
firms’ mobility in the regulated sectors from moving to other less regulated countries and so
should be associated with fewer permits. Besides these theoretically-derived factors, there
are other control variables to consider. Typically discussed variables are the current comparative (dis)advantage of an industrial sector and political pressure from interest groups. For
such considerations, net import and the number of employees at sector level are included for
competitiveness and political influence, respectively. In the meanwhile, detailed abatement
costs at sector level are not available so that they are not included. The following is the
empirical model with sector level data for participating members.
(sector level)

S
S
P ERM ITsc = µs + µc + β1 HISTsc + β2 P ROF ITsc

(2.13)

S
S
S
+ β3 N ET IMsc + β4 F IXEDsc + β5 EM Psc + εsc ,

where superscript S on the coefficients indicates that this regression is run at sector level
and subscript s and c indicate industrial sector and country, respectively; the dependent
variable (P ERM ITsc ) is the number of permits allocated to industrial sector s in country c
24

in phase I;9 µs and µc are sector- and country fixed effects; HISTsc denotes average annual
historical emissions; P ROF ITsc denotes average annual historical profit level; N ET IMsc
denotes average annual historical net imports; F IXEDsc average annual historical fixed
costs; EM Psc denotes average annual historical employment size; and finally εsc is the error
term. The period of historical data ranges from 1999 to 2002.
The first explanatory variable (HISTsc ) is average annual historical emission of sector s in
country c in 1999 to 2002. The measurement unit of historical emission is 1 million MtCO2e.
From the minimum-required permit allocation equation (2.7), sectors that are expected to
S
emit more under the regulation have more permits allocated. Hence, the coefficient β1 is

expected to be positive. P ROF ITsc is the average annual production margin (total revenue
less total variable costs) measured by gross operating surplus.10 The base unit is 1 million
euros. Note that production margin does not include fixed costs, which is treated as an
independent factor to influence the permit allocation. The level of profits represents the potential welfare loss caused by the relocation of firms in certain sectors when all other things
are equal. Hence, governments that try to minimize the welfare loss from implementing the
emissions trading will distribute more permits to sectors that earn high profits and hence
S
β2 is expected to be positive.

Fixed costs (F IXEDsc ) proxy for the mobility of industrial sectors; relocation of production
facilities may be harder for firms in industrial sectors with high fixed costs. Since the relocation of firms in the regulated sectors can reduce the benefits such as profits and employment
in the country and since carbon leakage through the relocation can undermine the goals of
the emissions trading, the prevention of carbon leakage is important and so the governments
have a good incentive to distribute more permits to “footloose” sectors. Therefore, the co9 the

measurement unit of permit is 1 million MtCO2e. MtCO2 = metric ton of carbon
dioxide equivalent
10 For detailed definitions of variables and data, see appendix B

25

efficient of fixed costs, β4 , is expected to be negative. Demeaned interaction term between
P ROF IT and F IXED is included to see how governments distribute permits to high (low)
profit industries with high (low) fixed costs. For instance, industrial sectors with high profits
will have more permits allocated. But, among those sectors with higher fixed cost will have
relatively fewer permits allocated. Regarding data, investment in tangible goods is used for
fixed costs. The base unit of investment in tangible goods is e1 million.
Net imports (N ET IMsc ), which are imports minus exports indicating the competitiveness
status of sector s in country c before the EU ETS. Net imports in the regression model
are not directly implied from the theoretical model. However, it is considered in permit
allocation for more political reasons than other variables. Large net import of an industrial
sector means that that industrial sector do not have the comparative advantage in the global
market. Allocating more permits to less cost-competitive (comparatively disadvantageous)
sector may compensate for the profit loss from the emissions trading regulation, but does not
improve the cost disadvantage caused by the regulation. In other words, permits should not
be allocated in the purpose of improving cost-competitiveness of industrial sectors. However,
in many cases, political interest groups urge to allocate more permits to already weak industrial sectors. Hence, if the government is subdued by the political pressure, it may allocate
more permits to less cost-competitive sectors and hence the coefficient of net import can be
non-negative.
Another variable that is not directly implied by the theoretical model is number of employees. Number of employees is also included in order to control a potential political power
of an industrial sector in permit allocation. Larger number of employees can be not only a
measure for a greater political power but the the size of workers whose jobs might be at stake
because of the emissions trading. In either way the coefficient is expected to be non-negative.
The next empirical model is presented at installation level. Data availability at installa-

26

tion level is even more limited. Installation level emissions data are rare before the EU ETS.
Only the UK keeps historical emissions data at installation level.11 Therefore, the same regression model used in the sector level is not applicable at installation level. However, what
is distinctive at installation level regression from sector level is the ownership information of
installations by firms and hence other aspects of permit allocations from firm characteristics
can be investigated. The following is the regression model at the UK installation level:

I
I
log P ERM ITif s = φs + β1 log HISTif s + δ1 M U LT I IN STif

(2.14)

I
I
+ δ2 M U LT I N ATif + δ3 M U LT I SECTif + εI s ,
if

where superscript I indicates that this regression is run at installation level; subscripts i, f
and s imply installation, firm and sector, respectively; dependent P ERM ITif s denotes
the number of permits allocated to each installation in phase I; φs denotes the sectorspecific fixed effects; HISTif s denotes average annual historical emissions from 1999 to
2002; M U LT I IN STif , M U LT I N ATif and M U LT I SECTif denote dummies indicating
whether an installation belongs to a firm with multiple installations, whether an installation
belongs to a multi-national firm or whether an installation belongs to a firm involved in
multiple sectors; and εI s is the error term. This regression model is for the UK only so that
if
country subscript is dropped and country-specific fixed effect is not included.
The basic regression identification structure is somewhat different from equation (2.13) at
the sector level. The dependent variable is the log of number of permits allocated to installation i that belong to firm f and to sector s in phase I. Note that since there are some firms
that are involved in multi-sector activities, firms are not a strict subset of industrial sectors.
The first explanatory variable is the log of historical emissions at the installation level. As
11 In

the U.K., the Department of Energy and Climate Change (DECC) keeps the installation emission data before 2005, which enables us to perform the regression analysis for phase
I permit allocation at installation level.
27

in the sector level analysis, the coefficient of historical emissions is expected to be positive.
The following three dummy variables are unique at installation level. Each installation is
linked with the firm it belongs to and hence the following information is available: how
many installations each firm owns; where each installation of a firm is located; which industrial sector each installation of a firm is involved in. Based on the information, three
dummies are generated to indicate whether the firm an installation belongs to owns multiple
installations (M U LT I IN ST ), whether the firm an installations belong to is multi-national
(M U LT I N AT ION ), and whether the firm an installation belongs to is involved in multiple industrial sectors (M U LT I SECT ). Among these multi-national dummy is expected to
have a positive coefficient. The reason is that multi-national firms tend to be more footloose
or mobile. They readily have production facilities in other countries and hence it is relatively
easier for multi-nationals to adjust the production level among production facilities if necessary. Therefore, it is expected that multi-national firms have more permits allocated due to
its supposedly high mobility. Two other dummies are expected to have negative coefficients
because multiple installations and multi-activity involvement imply that relocation may not
be easy as a firm. However, the mobility of multi-installation or multi-activity firms also
depend on the size of each installation, how independent (or dependent) each installation to
each other in the same sector, how complementary an industrial activity of each installation
is to other activities in the same firm and the like. Therefore, empirical results need to
be interpreted with these caveats. Since firm and installation characteristics such as multiinstallation, multi-nation and multi-sector are not changing over time, subscript indicating
periods are dropped.

28

2.4.2

Empirical Results in Phase I Permit Allocation

Table 2.2 provides the first phase permit allocation estimation results. The sector level
results are shown in columns (1), (2) and (3), whereas installation level result for the UK is
shown in column (4). It turns out that empirical outcomes confirm many predictions from
the conceptual model.
For the sector level analysis, the number of permits allocated in phase I, P ERM IT1 ,
is the dependent variable. First, empirical results show that higher annual average historical emission unambiguously leads to higher permit allocation. In sector level regressions
(columns (1), (2) and (3)), the coefficients of annual average historical emission (HIST0 )
are about 1.2 at 1% statistical significance. An additional historical emission leads to increase permit allocation by 1.2 MtCO2 in phase I.12 Since the permit allocation decisions are
made based on the projection of historical emission in each industrial sector, the coefficient
larger than 1 is reasonable.
In section 2.2, it is argued that the relocation of sectors with high profits to another country
is a big loss of the social welfare and hence governments have an incentives to distribute more
permits to sectors with high profits in order to minimize the welfare from the relocation. The
regression results confirm that theoretical prediction. P ROF IT0 in columns (1), (2) and (3),
measured by gross operating surplus, has positive coefficients at 5% statistical significance.
An extra euro increase in profit (gross operating surplus) leads to increase permit allocation
by about 0.0031 to 0.0052 tons in the first phase.
As theoretical model predicts, the mobility of industrial sectors plays an important role in
permit allocation; firms with a large investments in fixed physical structures are expected
not to be very responsive to stricter environmental regulations in respect of location choice
and hence it is reasonable to think that governments distribute more permits to firms in
12 MtCO2

= metric ton of carbon dioxide

29

Table 2.2: Phase I permit allocation in the EU ETS

historical emission (HIST0 )

(1)
1.263***
(0.059)

P ERM IT1
sector level
(2)
(3)
1.239***
1.173***
(0.067)
(0.065)

log (HIST0 )
gross operating surplus
(P ROF IT0 )
net import (N ET IM0 )
gross investment in
tangible goods (F IXED0 )
net investment in
tangible goods
(P ROF IT0 ) × (F IXED0 )

No. of employees (EM P0 )

0.984***
(0.011)
0.0052**
(0.0022)
0.0005
(0.0006)
-0.02***
(0.006)

0.0048**
(0.0018)
0.0008
(0.0006)

0.0031**
(0.0013)
-0.0002
(0.0004)
-0.026***
(0.0041)

-0.019***
(0.007)

0.0002***
(0.00005)

0.0002***
(0.00006)

1.0e06***
(3.4e-07)
0.0002***
(0.00005)

multi-inst. dummy
multi-nation dummy
multi-sect. dummy
constant
country FE
sector FE
firm FE
R2
Obs.

log (P ERM IT1 )
installation level (UK)
(4)

1.199
(8.350)
yes
yes
0.9878
106

-3.781
(8.823)
yes
yes
0.9866
98

7.726
6.050
yes
yes
0.9906
106

-0.005
(0.081)
0.069*
(0.040)
0.034
(0.026)
-0.053
(0.116)
yes
yes
0.9801
602

Note: Robust standard errors in parentheses; Significance is noted at 10% (*), 5% (**), and
1% (***). The base unit of permit and historical emission is 1 million MtCO2e. The base
unit of gross operating surplus, gross (net) investment in tangible goods is e1 million. The
measurement unit of the number of employees is 1,000.

30

“footloose” industrial sectors and distribute fewer permits to less mobile firms. Furthermore, since carbon leakage by the relocation of production facilities can significantly crowd
out the effort put on emission reduction, it is crucial for governments to prevent carbon leakage by distributing more permits to mobile firms. This argument is empirically confirmed.
In all sector level regressions, firms with high investments in fixed physical structures have
fewer permits distributed in the first phase. Fixed costs, measured in gross investment in
tangible goods and net investment in tangible goods, have negative coefficients that are at
1% statistical significance. An additional euro in fixed cost leads to decrease permit allocation by about 0.02 tons in columns (1) and (2). In column (3), the demeaned interaction
term between the profit level and fixed costs is included. The signs and magnitudes of main
effects are consistent with those without the interaction term. The coefficients of profits and
fixed cost are 0.0031 and -0.026 at 1% statistical significance, respectively. The coefficient
of demeaned interaction term is 1.0e-06 at 1% statistical significance. It implies that high
profit sectors for the same mobility have more permits allocated. In addition, profitability
may weigh more than mobility in permit allocation because even if the mobility is very high,
governments may not give more permits to firms in low profit sectors.
Net imports are used for a measure for industrial competitiveness. For sector level regressions net imports in columns (1), (2) and (3) have mixed signs of coefficients, none of which
have statistical significance. In other words, the competitiveness is not counted as an important factor in permit allocation of the EU ETS. This is theoretically convincing; the permit
allocation is irrelevant to the cost-competitiveness.
Combined with the insignificant results of net imports, high significance of fixed costs in
permit allocation decision implies that carbon leakage is a bigger concern than protecting
readily less competitive sectors. As explained, allocating more permits does not improve the
cost-competitiveness but just compensate for the profit loss from the regulation.
Last but not least, number of employees (EM P0 ) is included to control for the political
31

power of each sector and the results are intriguing. In sector level regressions, the number of
employees in columns (1), (2) and (3) has positive coefficients at 1% statistical significance.
The magnitude is about 0.0002; an additional 1000 employees in an industrial sector leads
to increase permit allocation by about 0.0002 tons to that sector.
The UK installation level regression is provided in column (4). The main focus of installation level regression is to find out the impact of firm characteristics on permit allocation
at installation level. The dependent variable is now log (P ERM IT1 ). The first variable is
average annual historical emission and the coefficient is 0.984 at 1% statistical significance.
One percent increment of historical emission implies about 0.984% more permits in phase I in
the UK. Now we look at the influence of firm characters on permit allocation at installation
level. First, the coefficient of multi-installation dummy is -0.005 at no statistical significance.
In other words, if an installation belongs to a firm that runs more than one installation, then
it tends to have -0.5% fewer permits in phase I. The coefficient of multi-nation dummy is
0.069 at 10% statistical significance. Installations run by multi-national firms tend to have
6.9% more permits allocated. The outcome is interpreted in respect of mobility of firms.
Multi-national firms readily have production facilities in multiple countries so that they can
adjust production levels among production facilities if necessary and can expand the production by adding production equipment or updating existing facilities. In other words, carbon
leakage is more likely to happen by multi-national firms. Then, governments are expected to
distribute more permits to multi-national firms or installations run by multi-national firms to
curb carbon leakage. Finally, the coefficient of multi-sector dummy is 0.034 at no statistical
significance. An installation run by a firm that is involved in more than one industrial sectors
has 3.4% more permits allocated. In the previous section, it is argued that installations run
by firms that operate multiple installations in one or more industrial sectors are likely to
have fewer permits allocated than installations run by firms that operate a single installation
in one sector. The reason is that the mobility of firms operating multiple installations in
32

multiple sectors tends to be low. However, the regression result in the UK permit allocations
does not send very clear message about it.

2.5
2.5.1

Empirical Analysis for Phase II Allocation
Empirical Model Specification for Phase II Allocation

The purpose of this section is to compare permit allocation in phase II with permit allocation
in phase I. Since the permit allocation decision for phase II was determined in 2006, in the
middle of phase I, important information such as overall permit holding position of the first
phase was not available. Furthermore, national allocation plan of each country reveals how
permits are allocated in phase I and hence many installations or firms may have taken some
actions to have more permits allocated in phase II. This is why data from phase I may not
be ideal for phase II permit allocation in the EU ETS. Therefore, the same historical data
set for the phase I permit allocation are used for the second phase permit allocation decision.
The regression models for the phase II permit allocation are also intended to investigate the
inter-phase permit allocation change. Hence the empirical specification includes data from
2005 of phase I and tries to see how updated information obtained during phase I affects the
phase II permit allocation given the first phase permit allocation decisions. Fowlie (2010)
addresses a rebate rate at which output-based permit allocated is updated and points out
that dynamic updating of permit allocation has the potential to mitigate competitiveness
adversity and carbon leakage but can undermine the cost-effectiveness of permit market.
Questions related to the second phase permit allocations are: whether the permit allocation
rules in the first phase still hold in the second phase permit allocation; and whether updated
information in permit holding position has an influence on permit allocation in phase II.
Before answering the questions with regression analysis, Figure ?? shows simple correlations

33

at installation level between (a) the log of phase II permit allocation and the log of 2005
permit allocation of phase I; (b) the log of phase II permit allocation and the log of 2005
verified emission of phase I; and (c) the log of phase II permit allocation and the ratio of
verified emission to permit allocation in 2005.

y: log (P ERM IT2 ), x: log (P ERM IT2005 )
Note: For the interpretation of the references to color in this and all other figures, the
reader is referred to the electronic version of this dissertation.
Figure 2.2: Phase II permit allocations to permit holding positions 1

From Figures (2.2) and (2.3), we can see that both permit allocation and verified emission in 2005 are positively correlated with phase II permit allocation. In terms of percentage
change shown in Figure 2.4, permit allocation in phase II is positively correlated to the ratio
of verified emission to permit allocation in 2005.
To answer the first question whether the permit allocation rules in the first phase are
still valid in the second phase permit allocation, we use the same regression equation used
for the phase I permit allocation in equation (2.13). The only difference is that the depen-

34

y: log (P ERM IT2 ), x: log (V ERIF2005 )
Figure 2.3: Phase II permit allocations to permit holding positions 2

y: log (P ERM IT2 ), x:

V ERIF2005
P ERM IT2005

Figure 2.4: Phase II permit allocations to permit holding positions 3

35

dent variable is now the permit allocation in phase II, P ERM ITsc,2 . To capture the effect
of performance in 2005 on the phase II permit allocation, both verified emission and the
number of permits allocation in 2005 are considered in the same regression. However, the
framework of regression models is still the same with that for the first phase. The following
is the regression model for phase II permit allocation with data from 2005.

S
S
P ERM ITsc,2 = µs + µc + γ1 V ERIFsc,2005 + γ2 P ERM ITsc,2005

(2.15)

S
S
S
+ γ3 N ET IMsc,2005 + γ4 P ROF ITsc,2005 + γ5 F IXEDsc,2005
S
S
+ γ6 EM Psc,2005 + ηsc,2 ,

where superscript S implies that this regression is run at sector level; subscripts s and c
denote sector and country, respectively; the last subscripts ‘2005’ and ‘2’ denote year 2005
of phase I and phase II of the EU ETS, respectively; at sector level the dependent variable is permit allocation in phase II, P ERM ITsc,2 ; V ERIFsc,2005 and P ERM ITsc,2005
denote the reported verified emissions and number of permits allocated in 2005 of phase I;
N ET IMsc,2005 , P ROF ITsc,2005 , F IXEDsc,2005 and EM Psc,2005 denote net imports, profS
its, fixed costs, and the size of employment, respectively; and ηsc,2 is the error term.

Permit allocation in 2005, P ERM ITsc,2005 , is included to control the information contained
in the first phase permit allocations. In addition, to test whether permit shortage in 2005 affects the phase II allocation the case that the coefficient of 2005 permit allocation is restricted
to negative one is considered. Hence, we observe the sign and magnitude of coefficient of
V ERIFsc,2005 − P ERM ITsc,2005 . Updated information of other sector-level variables in
2005 is included too; they are profit level (P ROF ITsc,2005 ), net import (N ET IMsc,2005 ),
fixed costs (F IXEDsc,2005 ) and the number of employees (EM Psc,2005 ). Country and sector fixed effects are also included. Hence, the impact of changes in 2005 on the second phase
36

permit allocation can be investigated.
Next, installation level regression model follows in equation (2.16). Since the EU ETS
started, permit allocations and verified emissions data have been available at installation
level in all ETS participating countries. Therefore, all installation data of European member
states that have participated in the EU ETS from phase I are used.13

I
I
I
log P ERM ITif sc,2 = φs + φc + γ1 log V ERIFif sc,2005 + γ2 log P ERM ITif sc,2005

(2.16)
+ λI M U LT I IN STif + λI M U LT I N ATif + λI M U LT I SECTif
1
2
3
I
+ ηif sc,2 ,

where superscript I implies that this regression is run at installation level; subscripts i, f ,
s, and c denote installation, firm, sector and country, respectively; last subscripts ‘2005’
I
and ‘2’ denote 2005 and phase II of the EU ETS, respectively; P ERM ITif sc,2 is the de-

pendent variable denoting the number of permits allocated in the second phase; φs and
φc are sector-specific and country-specific fixed effects, respectively; V ERIFif sc,2005 and
P ERM ITif sc,2005 denote reported verified emissions and number of permits allocated in
2005 of the first phase; M U LT I IN STif , M U LT I N ATif and M U LT I SECTif denote
dummies indicating installations of firms with multiple installations, installations of multiI
national firms, and installations of firms involved in multiple sectors; and ηif sc,2 is the error

term. The dependent variable and explanatory variables, verified emissions and the number
of permits allocated in 2005 are in the form of logarithm so that we can see the percentage
13 Among

EU27 members, Bulgaria and Malta did not join the first phase of the EU ETS.
Liechtenstein and Norway are not EU27 members, but they participate in the EU ETS from
phase II. Therefore, these four countries are not included in the analysis of phase II permit
allocation.
37

changes of permit allocations in the second phase corresponding to the variations of verified
emissions and permit allocations in 2005. Furthermore, to see the impact of permit holding positions on permit allocations in the second phase the log ratio of verified emissions
and permit allocations in 2005, log

V ERIFif sc,2005
P ERM ITif sc,2005

is included. As in the first phase

permit allocation regression, M U LT I IN ST , M U LT I N AT ION , and M U LT I SECT are
included to see how the characteristics of firms that each installation belongs to have influences on permit allocations in the second phase.

2.5.2

Empirical Results for Phase II Allocation

Table 2.3 provides the estimation results of permit allocations at sector level in phase II based
on the models discussed in section 2.5.1. Columns (1) and (2) use the same explanatory
variables with the phase I permit allocation estimations, which are mostly average annual
historical data from 1999 to 2002. On the other hand, data used in columns (3) and (4) are
from the first year of phase I (2005).
Average annual historical emission prior to the EU ETS (1999 to 2002) is still the critical
determinant of phase II permit allocation. An additional ton of historical emission leads
to increase permit allocations by about one more permits in phase II. The magnitudes of
coefficients in phase II are lower than those in phase I from around 1.2 to 1, which may imply
that the overall target becomes more stringent in phase II. Both coefficients in columns (1)
and (2) have statistical significance at 1% level. In column (3) verified emission in 2005
of phase I has a positive coefficient. An additional ton verified emissions in 2005 leads to
increase phase II allocation by about 0.594 more permits. In other words, industrial sectors
that have emitted more in phase I are likely to have more permit in phase II. In column (4)
the impact of permit shortage in 2005 (V ERIF2005 − P ERM IT2005 ) on permit allocations
38

Table 2.3: Phase II permit allocation in the EU ETS (sector)

country dummy
sector dummy

permit allocation in Phase II (sector level)
(1)
(2)
(3)
(4)
0.957***
1.057***
(0.080)
(0.035)
0.594*
(0.338)
0.190
(0.315)
-1.643
(1.018)
0.001
0.0027*
(0.001)
(0.0015)
0.003*
0.010*
(0.002)
(0.005)
-0.587
-0.359
(0.442)
(0.483)
-0.004
0.226
(0.044)
(0.148)
-0.007*
0.003
(0.004)
(0.003)
0.0024
-0.011
(0.0036)
(0.012)
-9.3e-06
-0.00003
(0.00007)
(0.00005)
-0.00002
0.00005
(0.00006)
(0.0002)
-1.1e-06***
(2.2e-07)
-8.2e-07**
8.9e-07
(3.4e-07)
(6.4e-07)
1.099
-0.664
-1.096
13.454
(1.502)
(2.169)
(1.779)
(8.638)
yes
yes
yes
yes
yes
yes
yes
yes

R2
Obs.

0.9835
104

historical emission
emissions in 2005
Phase I allocation
permit shortage
(V ERIF05 − P ERM IT05 )
avg. historical profits
(P ROF IT0 )
profits in 2005
(P ROF IT05 )
avg. historical net imports
(N ET IM0 )
net imports in 2005
(N ET IM05 )
avg. historical fixed cost
(F IXED0 )
fixed cost in 2005
(F IXED05 )
avg. historical employment
(EM P0 )
employment in 2005
(EM P05 )
P ROF IT0 × F IXED0
P ROF IT05 × F IXED05
constant

0.9888
104

0.9339
132

0.8603
132

Note: Robust standard errors in parentheses; Significance is noted at 10% (*), 5% (**), and 1%
(***).

39

in the second phase is estimated. The coefficient is -1.643 with no statistical significance. As
in phase I, the level of profits has a positive impact on permit allocations in phase II as well.
An additional euro in profits leads to increase permit allocation in phase II by 0.001 and
0.0027 more units in columns (1) and (2), respectively. The latter result is also statistically
significant at 10% level. The level of profits in 2005 has positive coefficients in both columns
(3) and (4). The magnitudes are 0.003 in column (3) and 0.01 in column (4); an additional
euro in profits leads to increase the permit allocation by 0.003 more permits in column (3)
and to 0.01 more permits in column (4). These results are consistent with the theoretical
prediction and the permit allocation decision in phase I. Fixed cost also shows a consistent
impact on permit allocation in phase II. The coefficient of F IXED0 is -0.007 in column (1),
which is statistically significant at 10% level. In other words, An additional euro in fixed cost
leads to decrease 0.007 fewer permits in phase II. Since fixed cost represents the mobility
of a sector, the result is consistent with theoretical prediction and the permit allocation in
phase I. In column (2) with a demeaned interaction term between profits and fixed cost,
fixed cost as a main effect has a positive yet not statistically significant coefficient. However,
the coefficient of interaction term is significantly negative. In words, the effect of profits is
weaker for firms with high fixed cost in phase II permit allocation. The interaction term
between profits and fixed costs in column (3) also has a negative coefficient, -8.2e-07, at 5%
statistical significance. Again, low mobility modifies the values of high profit sectors and
hence governments have allocated fewer permits. As in the first phase permit allocation, net
imports as a measure for the competitiveness of sectors do not show any significant impact on
permit allocation in the second phase, either. Throughout the analysis, net import appears
not an important factor determining permit allocations. Even though the cost disadvantage
of regulated sectors caused from the emissions trading looms large in political debates, governments rather follow economic reasonings than political rhetoric and see carbon leakage
as a real and more urgent problem of implementing the emissions trading. The number of
40

employees as a measure for potential political pressure has positive coefficients with high
statistical significance in phase I. However, it is not an important factor any more in the
second phase. The coefficients have no statistical significance in columns (1) through (4).
In contrast to the first phase permit allocation, the number of employees does not seem to
play an important role in the second phase.

Table 2.4: Phase II permit allocation in the EU ETS (installation)
I
log P ERM IT2

log (V ERIF05 )
I
log P ERM IT05

(1)
0.483***
(0.040)
0.471***
(0.040)

V ERIF

05
log P ERM IT
05

multi-inst. dummy
multi-nation. dummy
multi-sect. dummy
constant
country dummy
sector dummy
firm dummy
R2
No. of Obs

(installation level)
(2)

0.470***
(0.076)
-0.681***
(0.148)
0.248***
(0.073)
-0.011
(0.073)
1.745
(0.727)
yes
yes
yes
0.2216
4178

-0.069*
(0.037)
0.064**
(0.029)
-0.059**
(0.030)
-0.002
(0.156)
yes
yes
yes
0.9092
4178

Note: Robust standard errors in parentheses; Significance is noted at 10% (*), 5% (**), and
1% (***).

The next Table 2.4 provides the estimation results of permit allocation at installation
level using data from 2005, which cover all participating member countries in Europe. The
41

I
dependent variable is the log of phase II permit allocation, log P ERM IT2 . Hence, the

percent variations of permit allocation in phase II is of interest. The first explanatory variable
is the log of verified emission in 2005. The coefficient is 0.482 at 1% statistical significance;
one percent increment in verified emission in 2005 leads to 0.482% more permits on average.
I
The next variable is the log of permit allocation in 2005, log P ERM IT05 . The coefficient

is 0.471 at 1% statistical significance. One percent increment of annual permit allocation
in phase I leads to 0.0471% more permits in phase II. Unlike sector level analysis, percent
changes using the logarithmic form show that both verified emission and permit allocation in
2005 have substantial influences on phase II permit allocations. Another interest is also on
the relationship between permit shortage in phase I and permit allocation in phase II, which
is represented by the logarithmic ratio of verified emissions to permit allocation in 2005,
I
I
log V ERIF2005 /P ERM IT2005 . The coefficient of permit shortage in phase I is 0.470 at

1% statistical significance. In other words, one percent increment in permit shortage implies
0.47% increase in phase II permit allocation. This results implies that there is an update in
permit allocations in phase I based on the performance in phase I.
Variables indicating firm characteristics provide consistent outcomes in both phases. The
coefficient of multi-installation dummy is -0.681 at 1% statistical significance. In words,
if installation belong to firms that run multiple installations, then they have -6.81% fewer
permits allocated. One possible explanation argued in section 2.4.1 is in the perspective of
mobility of firms. This paper argues that firms with low mobility have fewer permits allocated. For firms with multiple installations across multiple industrial sectors, the relocation
may not be an easy decision. Each installation of a firm may be strategically located in a
certain region and linked with other installations. For similar reasons, it is expected that the
coefficient of multi-sector dummy is negative and the regression results in column (1) and
(2) confirm the prediction based on mobility. The coefficient of multi-sector dummy is -5.9

42

at 5% statistical significance in coulmn (1), whereas it is -0.011 at no statistical significance
in column (2). In contrast to dummies discussed above, multi-nation dummy has positive
coefficients. The magnitudes are 0.064 and 0.248 in column (1) and (2), respectively. Each is
statistically significant at 5% and 1% level, respectively. Hence, if firms are multi-national,
then they tends to have 6.4% or 24.8% more permits allocated depending on specifications.
These results are in line with the result of phase I permit allocation. It is argued that the
mobility of multi-national firms is higher than single-national firms because multi-national
firms already have production facilities in other countries.

2.6

Conclusion

This Chapter studies how permit allocations in the EU emissions trading system are affected
by factors that are related to industrial competitiveness and carbon leakage have been sought.
Even though the permit allocation is irrelevant to the cost disadvantage, grandfathering permit allocation is much preferred in the real world and it is argued that industrial sectors at
stake should be saved by allocating more permits. First, a two stage non-cooperative permit
allocation game model that explains grandfathering permit allocation decisions is built considering the possibility that firms can choose locations based on the profit comparisons. An
equation for the minimum required permits to keep each firm under the regulation is derived.
A firm is expected to get more permits when a firm earns higher profits in a pollution haven,
when it has high unit abatement cost compared to permit price, when its emission under
the emissions regulation are expected to be higher, and when it has lower fixed relocation
costs. Then, the governments with unadjustable stringent target emissions select firms to
keep based on the welfare impact (per permit) from the relocation of each firm.
With the data from the EU ETS, the predictions about the initial permit allocation decisions
in the EU Emission Trading Scheme are empirically tested. Three different layers of data
43

from the EU ETS are covered: installation level, firm level, and sector level. The overall
results can be summarized as follows. First, average annual historical emissions (prior to
the EU ETS) are the most decisive factor that explains permit allocation at any analyses.
Second, profit levels also have significant positive impact on permit allocation in the initial
permit allocations. Since the profits of firms are part of the social welfare, the relocation of
high profit firms hurts the social welfare larger. Hence, relatively more permits are allocated
to high profit firms. Third, industrial sectors with high fixed costs have relatively fewer
permit allocated. Fixed costs of firms are a factor that deters the relocation of firms. Therefore, there is less incentive for a government to distribute more permits to sectors insensitive
to stringent environmental regulations. Fourth, most arguably, net imports as a measure
for competitiveness do not seem to matter much in the initial permit allocation. From the
results of empirical analyses, it appears that what matters more in permit allocation decision is carbon leakage rather than the adverse change of industrial competitiveness. Fifth,
installations that belong to multinational firms tend to have more permits allocated, while
multi-installation firms receive fewer permits. It is argued that the result is part of mobility
argument. Since multinational firms readily have production facilities in other countries,
they are more footloose than single-national firms and hence fewer permits are expected. On
the other hand, firms with multiple installations are more likely the ones that spend high
fixed cost and hence their mobility is limited. Fewer permits are expected, too
In summary, the results suggest that countries implementing emissions trading system use
the permit allocation to keep foot-loose firms. The empirical results confirm theoretical
predictions and provide supporting evidence that the participating governments to the EU
ETS weigh more on keeping firms under the regulation and welfare loss in consequence than
saving less competitive sectors in the initial permit allocations. In other words, the level of
emissions, profitability and geographical mobility play important roles in permit allocation.
Considering that keeping firms under the regulation is one of the biggest challenges to the
44

successful implementation of the emissions trading system to curve global anthropogenic carbon emission, the evidence we found in the case of the EU ETS is reasonable and convincing.
In fact, both economists and environmentalists have been skeptical about the effectiveness
of the emissions trading with grandfathering permit allocation. It has been argued that
auctioning may be a better method for permit allocation. However, the drastic cost increase
may cause firms to move to other countries with no environmental regulations and then
carbon leakage can happen eventually. If so, carbon leakage is a serious problem and can
crowd out all the efforts made through the emissions trading regulation. Furthermore, tax
revenue and employment are also affected by the relocation of firms. Then, in terms of
the social welfare, keeping firms under the regulation by giving more permits may not be
so defective. Many emissions trading systems around the world start with grandfathering
permit allocation with some political reasons. However, in effect, it is not all political but
actually helps reduce global emissions. At least in the starting stage grandfathering permit
allocation is not so erroneous a decision.

45

APPENDIX

46

Appendix A

A.1

Industrial Classification Table
Table A.1: Industrial classification correspondence table

EU ETS
1.combustion installations
2.mineral oil refineries
3.coke ovens
4.metal ore
5.production of iron and
steel
6.production of cement and
lime

UNFCCC
1.A.1.A
1.A.1.B
1.A.1.C
not classified
1.A.2.A and
2.C.1
2.A.1
and
2.A.2

ISIC 3.1
4010
2320
2310
2710

NACE 1.1
E 40.11
DF 23.2
DF 23.1
DJ 27.1

2694

DI 26.5

7.manufacture of glass

2.A.7.1, 2.A.3
and 2.A.4
not classified
1.A.2.D

2610

DI 26.1

2100

DE 21

8.manufacture of ceramic
9.pulp, paper and board

SITC rev. 3
351
334, 335, 342, 344
325, 335.21
67
273.24,
278.23,
661.11,
661.12,
661.13,
661.21,
661.22,
661.23,
661.29
651.95, 664, 665,
773.22, 813.91
251.2, 251.3, 251.41,
251.42, 641, 642,
659.11, 892.81

Source: World Bank, Eurostat, European Environmental Agency (2007), IPCC (1996) and
Intergovernmental Panel on Climate Change (2006)

47

A.2

Summary Statistics
Table A.2: Summary Statistics

Variable
permits allocated in Phase I (inst.)
verified emissions in Phase I (inst.)
permits allocated in Phase II (inst.)
multi-inst. dummy
multi-nation dummy
multi-sector dummy
average imports (’99-’02)
average exports (’99-’02)
average net imports (’99-’02)
(net import/production) (’99-’02)
average imports (’05-’07)
average exports (’05-’07)
average net imports (’05-’07)
(net import/production) (’05-’07)
gross operating surplus (’99-’02)
gross investment in tangible goods
(’99-’02)
net investment in tangible goods
(’99-’02)
No. of employees (’99-’02)
gross operating surplus (’05-’07)
gross investment in tangible goods
(’05-’07)
net investment in tangible goods
(’05-’07)
No. of employees (’05-’07)

Obs
11154
11554
11084
5663
5663
5663
175
175
175
175
175
175
175
175
175
175

Mean Std. Dev.
Min
Max
566148.5
2735348
0 9.24e+07
533958.8
2754169
0 9.04e+07
858927.6
3750598
0 1.35e+08
.945
.228
0
1
.522
.499
0
1
.441
.497
0
1
1025.1
1886.4
0
11553.7
1206.6
2259.2
.149
14437.7
-176.9
1308.9 -8216.3
4399.8
-.016
.313
-6.548
9.088
1543.1
2839.2
3.281 23870.23
1873.6
3366.5
.299 26546.34
-331.7
1490.0 -10071.5
4730.67
.020
1.116
-1.540
36.519
4966.1
4971.9
-163.9
13397.4
2043.4
2173.1
0
6633.5

175

2105.819

2108.699

-35.65

6132.12

175
175
175

71155.36
6562.81
2744.9

64850.2
6893.3
2607.3

0
-.233
0

218204.8
18742.4
6586.97

175

2554.3

2487.8

-.4

6280.3

175

64903.49

61637.5

0

206427.7

48

A.3

Definitions
Table A.3: Definitions of variables

Variable
permit

base unit
1 million
MtCO2

Definitions and Descriptions (data source)
Average number of permits allocated in a given Phase. EUA
(the title of permit used in the EU ETS) is equivalent to one
ton of carbon dioxide. (CITL)
historical
1 million Average annual emission before the EU ETS. Countries that are
emission
MtCO2
Parties to the Convention submit national greenhouse gas inventories to the Climate Change secretariat based on the UNFCCC
Reporting Guidelines. (UNFCCC)
verified emis- 1 million Average amount of emission in a given phase. The EU ETS resion
MtCO2
quires all annual emissions reports and monitoring to be verified
by an independent accredited verifier. (CITL)
net imports
e1 million Import less export scaled by the level of production. (Eurostat)
gross operat- e1 million The surplus generated by operating activities after deducing laing surplus
bor input has been recompensed. Essentially GOS is gross output less the cost of intermediate goods and services, and less
compensation of employees. (Eurostat)
investment
e1 million Investment in tangible goods includes gross investment in land,
in
tangible
gross fixed capital formation in existing and new buildings and
goods
structures and machinery and equipment. (Eurostat)
No. of em- 1000
Total number of persons who worked in or for the concerned
ployees
workers
unit during the reference year. (Eurostat)
multi-inst.
Multi-installation dummy indicates whether an installation is
dummy
run by a firm that has more than one installation. (Carbon
market data)
multi-nation
Multi-nation dummy indicates whether an installation is run by
dummy
a multi-national firms. (Carbon market data)
multi-sector
Multi-sector dummy indicates whether the activities of a firm is
dummy
whether an installation is run by a multi-sector firm. (Carbon
market data)
Source: Structural and Demographic Business Statistics 2009 (OECD). The Definitions of
Characteristics for Structural Business Statistics (EUROSTAT). Carbonmarketdata (2010).

49

A.4

Permit Allocation Decision under Cournot Competition

Here, more explicit story on permit allocation can be generated by assuming market structures. The questions we are intended to answer through this process are; how does the
permit allocation change depending on market structure?; how would the changes of permit
price τ affect the location choices of firms and the government permit allocations?
The followings are simplifying assumptions;
1. a linear demand function p = d − bX is assumed, where d and b are constant.
2. the same linear production cost function and linear emissions function in the same
sector s : C(xi ) = cs xi and e(xi ) = es xi
3. There are N firms in sector s in the world. Firm i is the firm under the emissions
trading initially, whereas firm j is the representative firm that is located in another
country with no environmental regulations.
4. Linear environmental damage function D = D(E GE ) = hE GE is assumed, where E GE
denotes global emissions and h denotes a damage coefficient..
5. Income is big enough so that with quasi-linear utility function there is no income effect
caused by output price change.
6. Emission target in country R is exogenously given by the social agreement.
7. If profits are the same in both the regulated (home) country and the pollution haven,
firms choose to stay at home country.

50

A.4.1
A.4.1.1

The Location Choice of Firms
Profit under the Emissions Trading System

The market structure of sector s is Cournot oligopolistic competition. Profit function of firm
i in sector s under the emissions trading system is as follows.

πi = p(X)xi − Ci (xi ) − Bi (ai ) + τ (Ai + ai − ei (xi )),
where X =

N
k=1 xk

(A.1)

is the aggregate supply of sector s; p denotes the corresponding output

price; C(xi ) is total production cost when xi is produced; B(ai ) is abatement cost when the
amount of ai is abated; τ is exogenous permit price; Ai is the number of permits allocated
to firm i; and e(xi ) is the amount of CO2 emissions when xi is produced. For simplicity
linear production cost function is assumed. Also, the level of emissions is determined linearly
to the level of production. On the other hand, abatement cost is an increasing function of
abatement (B > 0 and B > 0). First order conditions of firm i in sector s under the
emissions trading system with respect to the production and abatement are:
∂Ci
∂e
dπi
= p(X) + p xi −
− τ i = d − bX − bxi − ci − τ ei = 0
dxi
∂xi
∂xi

(A.2)

dπi
∂B
=− i +τ =0
dai
∂ai

(A.3)

and

Production and abatement are not affect each other. Reaction function of firm i under
emissions trading system with N − 1 rivals is:

ET
Ri = xi =

d − ci − τ ei − b
2b

51

j=i xj

(A.4)

Reaction function of all the other N −1 firms in sector s under no environmental regulation
is

ET
R j = xj =

d − cj − b

k=j xk

(A.5)

2b

In the matrix form for N firms in the world, we can solve the simultaneous equations to
obtain equilibrium output for each firm.










 1 1/2 1/2 · · · 1/2   x1   (d − c1 )/2b


 
 1/2 1 1/2 · · · 1/2   x   (d − c )/2b
2

 2  


 
.
.  
.

 .
.
.

.
 .  
.


=


 
.
.

  xi   (d − ci − τ e)/2b
.


 
 .  

.
.

 .  
.
.
.
.

 .  


 
1/2 1/2 1/2 · · · 1
xN
(d − cN )/2b

















(A.6)

Using the Cramer’s rule, we can easily obtain the (Nash) equilibrium output of the
individual firm. Then, the equilibrium output for the regulated firm i will be

xET =
i
where d − N (ci + τ ei ) +

1
(N + 1)b
j=i cj

d − N (ci + τ ei ) +

c
j=i j

,

(A.7)

> 0 for positive production level is assumed as well.

For all the other N − 1 firms, the equilibrium output will be:

xET =
j

1
(N + 1)b

where d − N cj + (ci + τ ei ) +

d − N cj + (ci + τ ei ) +

k=i,j ck

c
k=i,j k

,

> 0 is assumed. Summing up across the world

shows the global output of sector s is given as follows:

X ET =

M
l=1

xl =

N d − (ci + τ e) −
(N + 1)b
52

k=i ck

(A.8)

Then, we can obtain the impact of permit price rise on an individual output and total
output. The output changes for the firms in sector s are:
∂xET
Ne
i
=−
<0
∂τ
(N + 1)b

(A.9)

∂xET
e
k
=
>0
∂τ
(N + 1)b

(A.10)

and

As permit price rises, the level of output of firm i under the regulation decreases shown
in A.9. On the other hand, N − 1 rival firms increase their levels of production as permit
price rises as shown in A.10. Hence the overall output changes in sector s in the world will
be:
e
∂X ET
=−
<0
∂τ
(N + 1)b

(A.11)

In words, the increase of permit price harms the firm under the emissions trading the
most and other firms under no regulations share the benefits equally. Finally, the overall
output of the sector s in the world drops as the regulation of emissions trading toughens
up and hence the permit price rises. However, the impact gets blurred as the number of
competitors gets larger and the market gets more competitive.
The corresponding output price when firm i is under the emissions trading system is

pET = d − bX ET =

d + ci + τ e +

k=i ck

(N + 1)

(A.12)

Then the impact of permit price rise on output price will be
e
∂pET
=
>0
∂τ
(N + 1)

53

(A.13)

Since the overall output decreases as the permit price rises, the corresponding output
price rises as well. The profit of firm i under the emissions trading is as follows.

ET = p(X ET )xET − C (xET ) − B (aET ) + τ (A + a − e (xET ))
πi
i
i
i
i i
i i
i
d + c1 + τ e + k=i ck
d − N (ci + τ e) + k=i ck
=
(N + 1)
(N + 1)b

d − N (ci + τ e) +

−c

=

k=i ck

− τe

(N + 1)b

−B(aET ) + τ (Ai + aET )
d + c1 + τ e + k=i ck
(N + 1)

− c1 − τ e

d − N (ci + τ e) +

(A.14)

k=i ck

(N + 1)b

d − N (ci + τ e) +
(N + 1)b

k ck

−B(aET ) + τ (Ai + aET )
d − N (ci + τ e) +

1
=
b

k=i ck

(N + 1)

2

− B(aET ) + τ (Ai + aET )

Put marginal production cost constant at c as assumed in the beginning, the profit of
firm i under the emissions trading system is simplified to;

ET
πi =

1
b

d − c − Nτe 2
− B(aET ) + τ (Ai + aET )
(N + 1)

The effect of permit price change on profit of the firm under the emissions trading is
ET
∂πi
2eN
=−
(d − c − N τ e) + Ai + aET
∂τ
b (N + 1)2

(A.15)

In the monopoly (N = 1), the effect of permit price change on monopoly firm under the
emissions trading is
ET
∂πi
∂τ

=
N =1

1
e (d − c − eτ ) + Ai + aET > 0
2b

54

(A.16)

Hence, if firm i is the monopoly supplier in sector s, then exogenous increase of permit
price also causes monopoly profits to increase. However, as number of firms increases, the
sign becomes ambiguous.

ET
∂πi
∂τ

=
N =∞

lim

N →∞

2ei N
b (N + 1)

2

c
k=i k

+ d − N c i − ei N τ

+ Ai + aET(A.17)

2
= − e (c + eτ ) + Ai + aET
b
Since the first term is negative, the profit depends on how many permits firm i has been
allocated and how much abatement it has done.

A.4.1.2

Relocation to a Pollution Haven

Now, no firm is under environmental regulations because previously regulated firm i now
moves to a pollution haven. Even though the relocation of firm i reduces its marginal cost
from the regulation, it causes extra fixed cost Gi . Additional fixed costs Gi are not just
temporary moving costs. They include the disposal costs of current production facilities
in the regulated country, costs caused by judicial system, language difference, difference in
culture at workplace, costs for new market research and the like. Hence, Gi is permanent
increase to relocation firm i.

πi = p(X)xi − C(xi ) − Gi
Hence, a typical firm in sector s has the following first order condition
∂Ci
dπi
= p(X) + p xi −
= d − bX − bxi − ci = 0
dxi
∂xi

55

(A.18)

Reaction function of each competing firm will be
d − ci − b

P
Ri H = xP H =
i

k=i xk

(A.19)

2b

Solve the simultaneous equation system of reaction functions as done above with matrix
form and we obtain the following Nash equilibrium output of symmetric firm j

xP H =
j

1
(N + 1)b

d − N cj +

c
k=j k

(A.20)

Summing up across countries shows the total world output of sector s.

XP H =

i

xP H =
i

1
Nd −
(N + 1)b

(A.21)

c
i i

The corresponding output price and profit are as follows, respectively.

pP H = d − bX P H =

d + i ci
(N + 1)

(A.22)

and

P
πi H = p(X P H )xP H − C(xP H ) − Gi
i
i

d − N ci +

=

d + i ci
(N + 1)

=

d + i ci
− ci
(N + 1)

1
=
b

d − N ci +

(A.23)

k=i ck

(N + 1)b
d − N ci +

k=i ck

(N + 1)b
k=i ck

(N + 1)

2

− Gi

56

−c

d − N ci +

k=i ck

(N + 1)b
− Gi

− Gi

A.4.1.3

Firm’s Location Choice and Minimum-Required Permits

To decide its location, firm i compares profits from each location. From equation (A.14) and
(A.23), we obtain the profit difference of firm i in each location.

1
ET
P
πi H − πi =
b

d − N ci + −i c−i 2
(N + 1)

−

1
b

d − N (ci + τ e) +
(N + 1)

=

1
b

Nτe
N +1

(A.24)
2

−i c−i

+ B(aET ) − τ (Ai + aET ) − Gi

2 d − N ci + −i c−i − N τ e
+ (B(aET ) − τ aET ) − τ Ai − Gi
N +1

The difference in profit decreases as the number of firms gets larger.

P
ET
∂ πi H − πi
∂N

=−

2τ e
(N + 1)3

2N ci + (N − 1)

c
−i −i

+ (N − 1)d + eN τ < 0
(A.25)

As the number of firms in sector s increases to infinity, the difference in profits converges
to zero.

lim

N →∞

−

2τ e
(N + 1)3

2N ci + (N − 1)

c
−i −i

+ (N − 1)d + eN τ

=0

(A.26)

Finally, the number of permits that can keep this firm in the regulated country is the
one that equalizes the profit from both regulated country and pollution haven.

1
¯
Ai =
b

N ei
N +1

2 d − N ci + −i c−i − N τ ei
aET
+
N +1
τ

B(aET )
−τ
aET

1
− Gi (A.27)
τ

This is the minimum-required permits that keep firms in the regulated country. Three
exogenous variable can change the minimum-required permit allocation; they are, number
57

of competing firms (N ), emission intensity (e), and permit price (τ ). For simplicity, the
constant marginal production cost is assumed. Since we assume that all competing firms
produce positive outputs, (d − ci − N τ e) > 0. First, as the number of competitors increases
in sector s in the world, the benefit from moving to the pollution haven drops. Hence, the
incentive to the relocation is irresolute and the the number of minimum required permits
decreases.
¯
2
e
∂ Ai
=−
((N − 1)d − (N − 1)c + N τ e) < 0
∂N
b (N + 1)3

(A.28)

Emission intensity also plays an important role to determine the minimum required permits.
¯
N
∂ Ai
=
(2d − 2c − N τ e) > 0
∂e
b (N + 1)2

(A.29)

As emission intensity, e, increases, the cost burden of firms under the emissions trading
also increases. Hence, there are more chances for them to relocate production facilities to
the pollution haven. Therefore, more permits will be allocated to firms with high emission
intensity.
¯
∂ Ai
N 2 e2
−
=−
∂τ
b(N + 1)2

B(aET )aET − Gi
τ2

(A.30)

The sign is ambiguous. The first term is definitely negative, whereas the sign of the
second term depends on the sign of B(aET )aET − Gi . If B(aET )aET − Gi > 0 , then
¯
∂ Ai
< 0. Hence, the increase of expected permit price leads the permit allocation to
∂τ
firm i to increase as well. Otherwise, the effect of permit price on the number of minimum
required permit is ambiguous.

58

Chapter 3
THE IMPACT OF EMISSIONS TRADING ON TRADE FLOWS:
EXPERIENCE OF THE EU ETS

3.1

Introduction

With increasing popularity of emissions trading as a tool to curb CO2 emissions in policy discussions, the discussions sometimes lead to an actual implementation of the emissions trading
system. As emissions trading systems are adopted, there is also growing concern about competitive implications of environmental regulation. The empirical findings are mixed but most
papers are about regulation of other pollution instead of carbon. Recently, the special issue of Climate Policy (volume 6, number 4, 2006) focuses entirely on the competitiveness
topics related to the EU ETS. So far research on carbon regulation is relatively lacking and
furthermore the evidence on the effects of carbon regulation is very limited. This chapter
provides another piece of evidence in the context of carbon regulation with theoretically
more supported variables and newer data from the EU ETS.
Brunnermeier and Levinson (2004) surveyed a group of papers that have researched about
the impact of environmental regulations and their stringency on trade flows. Even though
with the explorations of new data and advanced econometric techniques recent research
starts to find rather significant impact of environmental regulations on trade flows (Becker
and Henderson (2000); Greenstone (2002); List, McHone, Millimet and Fredriksson (2003);
Ederington et al. (2005)), the results of these papers are still arguable and most other papers
regarding this topic have not been so conclusive to overturn existing ambiguity about the
impact. Then, are political disputes against emissions trading exaggerated? Is the impact
of emissions trading implied in economic theory actually negligibly small to worry? Or is

59

there some other empirical problems that should be explained?
This paper argues that one of the most persistent problems that may produce obscure results is quality of data that represent the regulatory stringency. The most widely used data
indicating the stringency of environmental regulations are pollution abatement cost (PAC)
measures of industries. PAC measures are aggregate level data and give a general good idea
about how stringent overall environmental regulations are imposed on industries. At the
same time, using general spending data such as PAC causes many problems in analytical
point of view. First of all, we cannot distinguish the impact of one policy intervention from
another. In terms of policy, empirical results by using overall spending data may not help understand the impact or effectiveness of a particular environmental policy. Second, it is hard
to see the marginal impact of the environmental regulations. Annual environmental spending
data do not necessarily represent the change of marginal cost due to a policy intervention.
Since the comparative advantage of an industrial sector of a certain country is based on relative marginal cost or producer price, empirical results using data of environmental spending
that does not affect marginal cost or producer price can be misleading. Third, environmental spending data inevitably suffer from the endogeneity problem between trade policy and
environment policy. Theoretically, the environmental regulations causes the marginal costs
of firms to rise, harms the competitiveness of industries under the regulation and entails the
carbon leakage - a phenomenon that there is an increase in CO2 emissions in one country
as a result of an emissions deduction by another country with a climate policy. In order
to prevent unfavorable consequences from the environmental regulations, governments have
incentives to harmonize environmental and trade policies strategically. In other words, the
strategic policy harmonization can lead the environmental expenditure to be endogenously
linked with trade flows. For example, a dirty industry heavily exposed to international trade
may get subsidies on environmental spending from the government and so the expenditure
of a firm in that regulated industry on the record may appear larger than it actually is.
60

Then, possibly using overestimated expenditure data may generate a bias toward zero with
the changes of the trade flows of that industrial sector in empirical analyses since the actual
spending is smaller. The bottom line is if there is a policy that may generate comparative
disadvantage in international trade, then the governments tend to address a counter policy
than can neutralize the potential adverse impacts. Therefore, spending data such as PAC
cannot differentiate all the complicated interaction between policies properly and hence the
empirical results using those data may be less convincing. Ederington and Minier (2003)
point to the endogeneity as a main problem that causes a bias that can make the effect of
environmental regulations on trade flows ambiguous. Levinson and Taylor (2008) rigorously
explore data problems in so-called pollution haven hypothesis including policy endogeneity.
Results in these papers exhibit a little more supporting evidence to the pollution haven hypothesis but not convincing enough yet. Lastly, the availability and clarity of environmental
spending data are limited. Most of countries do not have relevant data. Even some countries that keep those data often have many missing points or inconsistency problem over
time in data collection. Moreover, detailed sector level data of PAC are really scarce in most
countries.
This study uses the price of emission permit with emissions intensity of each industrial
sector in the EU Emissions Trading System (EU ETS) as a measure for the stringency of
environmental regulations. The EU ETS has a clear policy target and the information about
which industrial sectors are under the regulations is readily available. Permit price with
emissions intensity of each industrial sector has a clear marginal cost implications. The
permit price variations alter optimal levels of output and abatement and so the trade flows
in consequence. Furthermore, each individual government participating in the EU ETS has
no control over the permit price. The total number of permits each government can issue is
strictly bound by the decision of the European Commission. The individual permit allocation to each sector has no marginal impact on equilibrium production or abatement as far
61

as the overall number of permits on the market is not changing and the permit market is
efficient.1 Therefore, there exists intrinsically less danger of endogeneity between environmental policy intervention and trade flows if permit price with emissions intensity instead
of general spending data is used as the regulatory stringency. Finally, permit price data are
easy to get, transparent and reliable. Most of permits in the EU ETS are traded in open
exchange markets such as ECX or Bluenext. Hence, permit price data are available to the
public and transparent.

Another distinctive feature of the paper is that the role of heterogeneous emissions intensities of individual sectors given by production technology is explicitly addressed and
emphasized. Permit price indicates the overall stringency of the emissions trading. However, permit price itself cannot depict optimal decisions of industrial sectors properly. In
other words, even though every emitter faces the same permit price, each individual sector may behave differently in the optimal decision depending on how dirty (or clean) it is.
For the same permit price, dirty industrial sectors produce less (emit less) and abate more
compared to clean sectors. Furthermore the emissions intensity in the same industrial sector is usually assumed to be the same across countries for theoretical simplicity (Antweiler,
Copeland and Taylor (2001); Copeland and Taylor (2003)). However, recent papers have
started to pay attention to international differences in emissions intensity even within the
same industrial sector (Quiroga, Sterner and Persson (2007); Douglas and Nishioka (2010)).
Appendix B.2 exhibits the wide contrasts in emissions intensity of the same EU ETS sectors
in different countries. Since the EU emissions trading system is a multi-country, multi-sector
emission regulation system, heterogeneous emissions intensity may vary across borders even
1 Hahn

and Stavins (2010) point out that there are six conditions under which the market
efficiency may not be reached; they are, transaction costs, market power, uncertainty, conditional permit allocations, non-cost-minimizing behavior by firms, and differential regulatory
treatment of firms
62

in the same sector. Industrial sectors with heterogeneous emissions intensity will take the
regulation differently and hence behave differently. Then the aggregate implications of the
regulation with considering heterogeneous emissions intensity may have different from those
without such considerations. Therefore, it is important to theoretically investigate the impact of emissions intensity in the emissions trading and then apply properly to empirical
analyses. Inclusion of emissions intensity is not only theoretically sound but a good device
to capture cross-sectional features of data for the empirical study performed in this paper.
This study tries to relate the theoretical model with estimation specifications closely. Theoretical structure follows a two-sector Heckscher-Ohlin type general equilibrium model. Each
sector has a different carbon-intensity and only the sector with high carbon-intensity (dirty
sector) is under the regulation. With comparative static analysis, the equilibriums with
and without the emissions trading are compared. Such theoretical framework fits well with
difference-in-difference estimation technique in econometrics. The object of the empirical
analysis is to see the relative impact of the emissions trading on trade flows of regulated
sectors to non-regulated sectors after the implementation of the emissions trading started in
2005. Non-regulated clean (low carbon-intensive) sectors will be a control group, whereas
regulated dirty (high carbon-intensive) sectors will a treatment group. The point that the
EU ETS started, 2005, separates years into two periods: after-policy period and beforepolicy period.
Section 3.2 provides a theoretical background about the emissions trading system incorporated in profits of the industry representative firm and the impact of regulatory stringency
change indicated by permit price and emissions intensity on trade flow. In section 3.3, using data from the EU ETS the impact is investigated. Section 3.4 presents the results of
empirical analysis. Section 3.5 concludes.

63

3.2

A Conceptual Model

This model follows a two sectors (X and Y ), two factors (labor and capital) general equilibriumtype trade model with the assumption of a small open economy. Sector X is the carbonintensive sector, whereas sector Y is clean (low carbon-intensive). This economy operates
a permit market of CO2 emissions, which is an artificially-formed market to curb carbon
emissions. All CO2 emitting firms in sector X are under the regulation of emissions trading;
however, sector Y is not regulated. The introduction of an emissions trading system changes
the production composition of goods; the output of regulated sector X contracts, whereas
that of non-regulated sector Y expands. Given the relative consumer price and the level of
income constant, the changes in production composition leads to the changes in trade flows
of both regulated and non-regulated sectors. Insufficient domestic supply of regulated sector
X due to the regulation is filled with increased net import (import less export). On the other
hand, the expansion of unregulated sector Y leads to the decrease of net import. Following
subsections will discuss more in detail about supply-side and demand-side equilibrium in the
goods and permits markets.

3.2.1

Supply-side Equilibrium

Country R is endowed with fixed inputs, labor (L) and capital (K). Inputs are mobile
between sectors, but cannot cross borders. There are two sectors, X and Y , whose production
technologies exhibit constant returns to scale. Sector X is capital-intensive and pollutionintensive, whereas sector Y is labor-intensive and clean. Since a small open economy is
assumed, relative output price p of good X is assumed to be exogenously determined. Output
price of good Y is normalized to unity. Country R runs an emissions trading scheme with free
permit allocation. Since sector X is pollution-intensive, it is under the regulation. Sector

64

Y is not regulated. A representative (average) firm in sector X in country R is allocated
with A permits to emit carbon dioxide into the air. It is also assumed that the regulatory
stringency of the rest of the world is exogenous and constant. The following is the profit
function of the representative firm in country R.

π = max{px(K, L) − cX (w, r)x(K, L) − D(a) + τ (A + a − θ · x(K, L))},
a,x

(3.1)

where p and τ denote the relative output price of good X and the permit price; x and a are
the firm’s chosen levels of output and abatement; cX (·) is the linear marginal production
cost of sector X, which is a function of wages to labor (w) and rents to capital (r); D(a) is an
abatement cost function (Da > 0 and Daa > 0); A in the last term is the number of permits
allocated; and θ is the emissions intensity (emissions per unit) of the production of x, which
is given exogenously by the production technology of x. Production of x requires labor L
and capital K, whereas abatement does not. In other words, production and abatement
are two separate processes.2 One of the convenient properties of constant returns to scale
technology is that the cost function C X can be written as linear in output and profits can
be maximized with respect to output itself, even though there are multiple inputs.
Since emissions trading creates a production-side distortion, producer price (q = p − θτ ) is
lower than the world output price p due to θ ≥ 0 and τ ≥ 0. Using producer price, the profit
function can be rewritten as follows.

π = max{qx(K, L) − cX (w, r)x(K, L) − D(a) + τ (A + a)}.
a,x

2 There

(3.2)

are other models that production and abatement share exactly the same technology and inputs. See Copeland and Taylor (2003)

65

As we can see, producer price is directly affected by the alterations of permit price (τ ) or
emissions intensity (θ) and so is the profit level of firms in regulated sector X. The following
are the first order conditions with respect to abatement and production:
∂π
= −Da + τ = 0.
∂a

(3.3)

∂π
= (p − θτ ) − cX (w, r) = 0.
∂x

(3.4)

and

The first order condition with respect to abatement indicates that the optimal abatement
level is determined by equating the marginal abatement cost to permit price. If terms are
rearranged in equation (3.3), the following relationship is obtained.

−1
a = Da (τ ).

(3.5)

Hence, the level of abatement is solely a function of permit price and is an increasing
function in permit price. From equation (3.4), the optimal production level is determined
by setting the producer price equal to the marginal production cost. With these equations,
the production side equilibrium conditions are the standard zero profit conditions and full
employment conditions of labor and capital. The zero profit condition for sector coincides
with the first order condition with respect to output. These principles of optimization apply
to both sectors X and Y . Since output price of good Y is normalized to unity, the following
are the zero profit conditions and full employment conditions for an equilibrium.

(zero profits)
q = cX (w, r)

(3.6)

1 = cY (w, r)

(3.7)

66

(full employments)
Y
X
¯ ∂c (w, r) x + ∂c (w, r) y
L=
∂w
∂w
Y (w, r)
X (w, r)
∂c
∂c
¯
x+
y
K=
∂r
∂r

(3.8)
(3.9)

The zero profit conditions are achieved when a producer price is equal to a marginal
input cost in each sector. The full employment conditions tell that this economy hires inputs
exhaustively at the most efficient level. The full employment conditions can be written in
the simple linear form using Shephard’s lemma. Based on these equilibrium conditions, we
can write the relative supply (RS) function of good x to good y as follows.
xs (q, K, L)
RS(q, K, L) ≡ s
,
y (q, K, L)

(3.10)

where superscript s denotes supplies of goods and q = p − θτ is producer price. Then, we
can calculate the partial effect of exogenous variables on the supply of good xs . An increase
in p expands the relative supply ( ∂RS > 0). However, an increase in θ or τ lowers producer
∂p
price and hence contracts the relative supply ( ∂RS < 0 and ∂RS < 0). According to the
∂θ
∂τ
Rybczinski theorem, an increase in factor endowment can either raise or lower the productions of good x and good y depending on their factor intensities. Since sector X is assumed
to be capital-intensive compared to sector Y , an increase in the capital endowment leads the
production of x to rise but the production of y to drop ( ∂RS > 0 and ∂RS < 0).3
∂K
∂L

3.2.2

Demand-side Equilibrium

It is assumed that consumers in country R are all identical. Each consumer benefits from
the consumption of goods and is harmed by carbon emissions. Preferences are assumed to
3 Cole

and Elliott (2003) also argue that pollution-intensive sectors happen to be capitalintensive sectors.
67

be identical and homothetic in consumption of X and Y following the trade literature. In
other words, for the same price consumers spend the exactly the same proportion of their
income on goods X and Y . Unlike producer price, consumer price is not affected by the
implementation of the EU ETS or permit price since consumer price is pinned down by
the world price.(there is no sales tax or other device that can distort consumption in this
economy). The utility function of a representative consumer can be written as follows.

U (x, y, E) = u(x, y) − h(E),

(3.11)

where U , u and h denote gross utility, utility from the consumption of goods x and y, and
disutility from global emission E (h > 0 and h

≥ 0), respectively. By homotheticity

assumption, we can write the indirect utility function as a function of real income (nominal
income divided by a price index) and demand for good x relative to good y is not affected
by the income level. Given output price, income and emission, the relative demand (RD)
function of good x to good y can be written as follows:

RD(p) ≡

bx (p)
,
by (p)

(3.12)

where b denotes the consumption share to overall income, which depends on p. Consumption
share of good x, denoted by bx , is decreasing in its relative output price p. Given relative
output price p an increase in income leads the consumption of good x to rise in proportion
to bx (p). On the other hand, consumption share of good y, denoted by by , is increasing in
relative output price p. Given p, an increase in income leads the consumption of good y to
rise in the proportion of by (p). Therefore, as seen in equation (3.12) the relative demand
(RD) of good x to good y is not affected by the income level.

68

Since world price is assumed to be determined exogenously, the only thing that remains
is to determine the level of income. At the national level, there will be income change from
the implementation of the emissions trading system because the overall emissions target
should be met at the end of the compliance period and the grandfathering permit allocation
method charges nothing by design as far as the target is met. More rigorous explanation
follows. As in the trade literature, the GDP function (aggregate revenue function) approach
is adopted from the standard competitive economy theory. Since each firm’s optimal behavior
maximizes its output value given price, national income reaches the maximum feasible value:

G(p, θ, τ, K, L) = max{(px − θτ )x + y + τ (A + a) : (x, y) ∈ T (K, L)}.
x,y,a

(3.13)

This expression is the sum of the maximum value of net revenue generated by the private
sectors given producer prices and production technology T and the revenue from free permits
and abatement. In order to know the impact of permit price change on national income
(GDP), take a derivative with respect to permit price, which gives the following:
∂G
= A + a − θx.
∂τ

(3.14)

Since each choice variable (x, y and a) is its maximized value by the private sector, the
usual envelope property is applied. The expression in result tells us that the marginal change
in national income (GDP) from the change of permit price is equal to the permit holding
position. In this special case that there is only one regulated sector in a country, Permit
allocation A is equal to the emissions target. In other words, by design the national emissions
target must be equal to the number of permits submitted at the end of a compliance period.
Hence, permit holding position should be neutral at the end of a compliance period and so
there will be no income effect through the permit price change.4
4 In

the case of the EU ETS, from the third phase (2013-2020)the overall emissions target

69

3.2.3

Permit Market

The permit market is an artificially created market to curb the level of emissions to a given
¯
target. Because there is a specific target to achieve, the total supply of permits, A, is set
¯
equal to the emissions target, E. The permit market is modeled based on the EU Emissions
Trading System (EU ETS). Since the determination of permit price is of our interest, the
discussions in this subsection are aimed at the entire EU ETS market. On the other hand,
the discussions about demand and supply equilibriums are made at the national level. The
EU ETS is a joint CO2 emission regulation of EU 27 member countries. The overall emission
target of the EU ETS is established based on the Kyoto Protocol. Emission permits can be
traded and used in any member countries. However, the target should be complied at the
national level. In other words, permit trading across borders is allowed during a compliance
year, but the target should be met by country at the end of the compliance year. This
self-contained system will change from the third phase of the EU ETS (2013-2020) to a
system that the emissions target is determined at the EU level, thereby national target does
not have to be met any more. The determination of emission target is a function of many
components such as overall emissions intensity of the EU (Θ), aggregate social welfare of
the EU (W ), expected future emission level (E e ), political pressure from interest groups and
the like. Under the agreed target, each member state of the EU can allocate permits within
its jurisdiction according to a written national allocation plan. At the end of a compliance
period, each firm has to submit the permits equal to its emissions. The sum of submitted
permits should be equal to the overall permits allocated at the national level. Thus, for N
is determined at the EU level, not at the national level. Then, there will be a potential
income effect depending on the permit holding position of a country.

70

participating firms the following should hold.
N

¯
¯
Aj = A = E.

(3.15)

j

Since the permit supply is predetermined and fixed, the permit price varies with changes
in permit demand. The overall permit demand in EU during a compliance period depends
on various factors such as overall emissions intensity (Θ), output price (p), factor endowments (K, L), the endowments of natural resources of a country, the performance of output
markets, the performance of energy markets, unexpected demand shocks in output markets
and energy markets and the like. Figure 3.1. briefly depicts the permit market and permit
price determination.

3.2.4

Permit Price, Emissions Intensity and Trade Flows

We now investigate the impact of the emissions trading system on trade flows in terms of net
imports. By saying net imports, it is implied that imports take the positive sign, whereas
exports take the negative sign. If the level of domestic consumption at the current consumer
price and income exceeds the level of domestic production at the current producer price, factor endowments and production technology, then this discrepancy will be filled with imports
in the open economy. Furthermore, this economy operates an emissions trading system,
which affects producer price of sector X and the level of income. The trade equilibrium
without environmental regulations is determined by output price p, capital endowment K
and labor endowment L. Under the emissions trading regulation, permit price and emissions
intensity also affects the trade equilibrium.5 Figure 3.2. depicts how trade flows in both
5 Note

that small open economy is assumed and the size of an industrial sector in an ETS
participating country is small relative to the size of the entire EU ETS. Thus, assuming
output price and permit price are exogenous may not be so unreasonable.

71

Figure 3.1: Permit price decision
Note: The permit price (τ ) determination depends on the supply and demand of permits.
The overall EU emissions target Λ is predetermined considering various factors such as
the overall EU emissions intensity (Θ), aggregate welfare of EU (W ), or expected future
emissions level E e . The overall EU emissions target (hence, the overall number of permits)
¯
¯
is determined at E = A. The demand for permits, on the other hand, can change in the
compliance period depending on output price (p), the overall EU emissions intensity (Θ),
capital or labor intensities of regulated sectors, energy prices and unexpected demand shocks.

72

sector X and Y change depending on permit price or emissions intensity.

To fix ideas, let us start with a special case that the free trade equilibrium coincides with
the autarky equilibrium so that we can see the impact of the emissions trading system on
trade flows more clearly. Point C depicts the economy with no environmental regulations.
World price, consumer price and producer price are all equal and the levels of production
and consumption coincide at x0 and y0 . Thus, there is no international trade.
Now country R starts to operate an emissions trading system, with which producer price
changes and international trade occurs. The current economic circumstance forms permit
price at τ . Then, with emissions intensity of sector X, θ1 , producer price moves from q = p
to q1 = (p − θ1 τ ) (point B) along the production possibility frontier. The output decisions of
x and y are made at x1 and y1 . The regulated sector X contracts, whereas the non-regulated
sector Y expands. Since the relative output price is not affected by the regulation, the consumptions of x and y do not deviate from x0 and y0 (point C). Then, this change causes
relative net import of good x to good y to increase.

High emissions intensity amplifies the impact of permit price on producer price and hence
on the level of production. Think of the case that higher emissions intensity than θ1 is given
for the same permit price τ , say θ2 . Then, producer price falls to q2 = p − θ2 τ , at which the
level of production changes from point B to point A. Given relative output price p, permit
price τ and emissions intensity θ2 , the production level of x decreases to x2 and that of y
increases to y2 . Again, the consumption levels of x and y stay still at x0 and y0 . Then,
import level of good x and export level of good y expand.

Figure 3.3 provides another way to look at the change of trade flows using the relative
demand (RD) and the relative supply (RS). As in Figure 3.2, it is assumed that free trade
73

Figure 3.2: Permit price, emissions intensity and production choices
Note: Starting point C is the free trade equilibrium with no regulation. As emissions trading
system runs, the productions of goods x and y change responding to the changes of producer
price from p to q1 (point B). With higher emissions intensity θ2 for the same permit price τ ,
producer price falls to q2 so that sector X contracts more and sector Y expands more (point
A).

74

Figure 3.3: Relative demand and relative supply
Note: With no regulation, the free trade equilibrium forms at point A. As producer price falls
from p0 to q1 , the production composition between sector X and sector Y moves from (x0 /y0 )
to (x1 /y1 ). On the other hand, since consumer price does not change, the consumption
pattern does not change either. Therefore, the horizontal distance between points A and B
shows the increase in relative net import of regulated sector X to non-regulated sector Y .

75

equilibrium coincides with the autarky one. In other words, the RS curve shown in the figure
describes the domestic relative supply as well as world relative supply. Without the emissions
trading regulation, both producer price and consumer price coincide with the world price
p0 (point A). With the implementation of the emissions trading, the relative supply of this
economy moves along the curve to point B; the relative production of sector x to sector y
is declined. However, the relative demand is still on point A because the implementation of
emissions trading does not affect the consumer price. Then, the gap between relative supply
and relative demand is filled with the increase of relative net imports of sector x to sector
y. The change of the relative imports can be described as follows.

RIM (p, θ, τ, K, L) ≡ RD(p) − RS(p, θ, τ, K, L).

(3.16)

Since the identical homothetic preference is assumed in demand side, the output markets are competitive and output prices are given exogenously, this equation tells that the
implementation of emissions trading affects relative imports only through the relative supply. As explained in equation (3.10), an increase in θ or τ lowers producer price and hence
contracts the relative supply ( ∂RS < 0 and ∂RS < 0). Then, the relative import changes
∂θ
∂τ
corresponding to the change of the relative supply.
∂RIM
>0
∂τ

(3.17)

∂RIM
>0
∂θ

(3.18)

These two equations above confirm the findings in Figure 3.3. As the permit price or
emissions intensity increases, the relative net imports of regulated sector decreases.

76

3.3

Empirical Model and Estimation

This section tries to build a sensible empirical model that can test the relative impact of
emissions trading system on trade flows. However, note that this is not an attempt to exactly
translate theoretical suggestions into an empirical model. However, the theoretical build-up
and its results are used as a motivation to derive proper variables that can represent the
regulatory impact of emissions trading system and to compare theoretical predictions with
empirical results.
The theory section shows that permit price interacted with emissions intensity plays an
important role in determining trade flows. In the empirical analysis, permit price indicates
the overall levels of regulatory stringency across periods and emissions intensity provides
the cross-sectional distinctions between industrial sectors. An increase of permit price or
emissions intensity raises the net imports of regulated sectors relative to the net imports
of unregulated sectors. Therefore, this empirical section intends to test the change of net
imports of regulated sectors relative to non-regulated sectors caused by the change of permit
price interacted with emissions intensity after the EU ETS. In order to test this, differencein-difference type econometrics technique is applied in the empirical analysis. Both regulated
sectors and non-regulated sectors are included in the empirical analysis. Then, before-ETS
periods and after-ETS periods are separated and indicated as different period-groups. The
industrial sectors under the EU ETS regulation can be viewed as the treatment group,
while other non-regulated manufacturing sectors can be viewed as the control group. The
level variables (net imports (N I), capital (K), and number of R&D workers (RnD)) are
normalized by the number of workers (L) in an industry. The following is the basic regression
model. Further specifications are shown in tables 3.1 and 3.2.

77

NI g
= β0 + β1
L ict

RnD
g
+ β3 T ARIF Fict
L
ict

K
+ β2
L ict

(3.19)

+ β4 ET Sic + β5 P OSTt + β5 θic + β6 ET Sic × P OSTt
+ β7 ET Sic · θic +

β8 τt

treatment group

after ETS

+ β9 (ET Sic · θic ) × τt + ict ,
treatment group after ETS

where subscripts i, c and t denote industrial sector, country and time, respectively; superscript g on (N I/L) and T ARIF F denotes a trading partner group; dependent variable
NI g
L ict

denotes net imports (per worker) of sector i in country c at time t from trading-partner

group g. For explanatory variables on the right-hand-side, K and L denote capital and the
number of employees and hence (K/L)ict is the capital-labor ratio, which represents physical
capital intensity; RnD denotes the number of R&D personnel and hence (RnD/L)ict is the
g

R&D-labor ratio, which represents human capital intensity; T ARIF Fict denotes tariff rates
imposed on imports from trading partner group g; ET Sic is a dummy that indicates EU ETS
sectors (both before and after the regulation); P OSTt is a dummy that indicates post-ETS
periods; θic denotes emissions intensity; ET Sic × P OSTt is an interaction term indicating
ETS sectors after the ETS; ET Sic · θic is the ETS dummy multiplied by (demeaned) emissions intensity of each sector; τt is permit price at time t, which is zero in the pre-policy
periods; ET Sic × τt denotes the interaction between ETS sectors and post-policy period;
(ET Sic · θic ) × τt denotes the interaction effect between treatment group and post-policy period fully considering sectoral heterogeneity; and finally ict denotes the error term. Country
dummies are included to control for unobserved factors across participating countries. It is
important to mention that empirical analysis differentiates trading partner groups based
on the income level. In the theoretical model in section 3.2, the trading partner has not
been clearly defined. The reason is that the comparative static analysis basically compares
old and new equilibriums when there is an exogenous economic shock or event. All other

78

exogenous variables are not changing between equilibriums. In the theoretical mode of this
paper, therefore, the characteristics of trading partners are described only if necessary. For
example, the environmental regulations of trading partners are assumed to be constant and
hence not mentioned explicitly. However, in reality each individual country has a different
regulatory stringency, which is argued to be correlated with the income level in the literature. The critical component in our theoretical model that generates international trade is
the imbalance of regulatory stringency among trading partner countries. Hence, the empirical analyses in this paper grouped trading partner countries based on the income status or
regions. Subsection 3.3.1 explains trading partner groups in detail.
Throughout the empirical analysis, two questions are expected to be answered. First
question is whether there is a distinctive impact of the emissions trading on trade flows of
the regulated sectors to non-regulated sectors; and second question is how regulated sectors
with different emissions intensity react to the regulation after the EU ETS. The emissions
trading as an environmental regulation is expected to contract the regulated sectors in production and to expand unregulated sectors. Eventually, the changes in production lead to
the changes in trade flows. Overall, the net imports of regulated sectors relative to nonregulated sectors are expected to increase. Furthermore, among regulated sectors, each
sector has different emissions intensity so that their reactions (the change of production)
to the same permit price vary based on their emissions intensities. The coefficient β9 of
(ET Sic · θic ) × τt delivers such information that tells the ‘individualized’ impact of the EU
ETS on regulated sectors after the policy implementation. Based on the argument in section
3.2.4, β9 is expected to be non-negative: the emissions trading is expected to cause regulated
sectors to increase net imports relative to non-regulated sectors and among regulated sectors
high emissions intensity can lead to even larger net imports. There is also a chance that the
impact may not be significant.

79

3.3.1

Data

This study concerns about the impact of EU emissions trading system on trade flows at
sector level. Data covers twenty four participating members of EU. Before-policy periods
range from 2000 to 2003 and after-policy periods from 2005 to 2007 (phase I). Six out of
nine ETS sectors are included as the experiment group; they are, mineral oil refinery, coke
oven, the production of iron and steel, the production of cement and lime, the production of
glass, paper pulp and boards. As the control group, seven other sectors are included. These
other sectors are, manufacture of basic chemicals (ISIC 2410), manufacture of other chemicals (ISIC 2420), manufacture of man-made fibers (ISIC 2430), manufacture of articles of
concrete cement and plaster (ISIC 2695), cutting shaping and finishing of stone (ISIC 2696),
manufacture of other non-metallic mineral products n.e.c. (ISIC 2699), and manufacture of
basic precious and non-ferrous metals (ISIC 2720).
Overall data can be categorized into three groups, trade-related, production-related and
regulation-related. Trade related data include imports, exports and (weighted average) tariff rates. In terms of source of trade data the World Bank, in collaboration with the various
international organizations such as the United Nations Conference on Trade and Development (UNCTAD), International Trade Center (ITC), United Nations Statistical Division
(UNSD) and the World Trade Organization (WTO), has developed the World Integrated
Trade Solution (WITS). This software accesses and retrieves information on trade and tariffs which is compiled by the above international organizations. All of trade related data are
from the WITS system.
The Heckscher-Ohlin type models predict that a country specializes in industries that use
abundant factors of the country intensively and hence production-related data are inputs for
productions. There are three factors considered in this paper: labor, human capital, and
physical capital. The total labor compensations, the total numbers of employees, total R&D

80

expenditures and total numbers of R&D employment are available from the EUROSTAT,
which is the official statistical office of the European Union. The labor share of an industry
(l) is the ratio of total labor compensation to value added of that industry. The physical
capital share is 1 less the labor share (1 − l). Capital-labor ratio is hence the ratio of labor
share (l) to capital share (1 − l) (Romalis (2004)).
Regulation-related data are permit prices and emissions intensity. As mentioned, permit
price is a measure for the overall regulatory stringency of the EU ETS and cost burden of
the regulated industries. EUA (European Union Allowance) is the tradable permit under
the EU ETS and one EUA represents the right to emit one ton of CO2. Nowadays most of
EUAs are traded in open exchange markets such as European Climate Exchange (ECX) in
London or Bluenext in Paris. Various price data are available at those exchange websites.
This study uses futures EUA prices rather than spot EUA prices. The reasons are following.
Trading volume at futures market is almost three times larger than trading volume at spot
market and thus futures prices are treated as more reliable information. In addition, permits
used in phase I of the EU ETS were not allowed to be banked to or borrowed from phase II.
In consequence, spot price of EUA, as phase I came close to the end, converged to zero even
though the regulation was continuously valid in the following phases and the stringency in
the second phase has been toughened up. Hence, it is hard to believe that industrial sectors
behave based on falling spot permit price of phase I as if there is no second or future phase.
Another critical variable regarding the regulation is emissions intensity. Each industry perceives permit prices differently because of its heterogeneity in emission intensity. The permit
price multiplied by the emissions intensity of each industrial sector represents the genuine
‘individualized’ regulatory stringency as suggested in the theoretical model. Emissions intensity of an industrial sector is the ratio of total emissions to real GDP. The EU Community
Transaction Log keeps the verified emissions data of each installation under the EU ETS.
However, the verified emissions of the EU ETS already include abatement and hence do not
81

represent true emissions intensity. Therefore, this paper uses average historical emissions
data (2000 - 2004) obtained from the UNFCCC instead. For the calculation of the real GDP
of each industrial sector, industrial HICP (harmonized index of consumer prices) is used.
Both sectoral GDP and HICP are available from the EUROSTAT.

3.4

Results

Table 3.1 provides regression results on relatively large and more inclusive trading partner
groups that contain very heterogeneous and regionally less specified trading partner countries; they are, outside EU (extra EU), low-middle income countries (LM), least developed
countries (LDC) and developing countries (DEVG). Due to high heterogeneity within a
group, it seems hard to detect any significant impact of the EU ETS on trade flows. In
some cases, results are not quite convincing. The impact of the emissions regulation on
ETS sectors relative to non-ETS sectors is indicated by the coefficients of (ET S × θ × τ ),
which is the interaction term between regulated-sector effect (ET S × θ) and after-policy
effect (τ ). The coefficients of the interaction term show mixed signs across trading partner
groups. In addition, none of them are statistically significant. In other words, empirical
analysis in Table 3.1 does not provide a supporting evidence that there has been a significant difference in trade flows between regulated sectors and non-regulated sectors since the
EU ETS started. Some other regulation-related variables show consistent and significant
results. The coefficients of ETS dummy are positive with no statistical significances. On
the other hand, post-ETS dummy has positive coefficients with statistical significances. In
other words, since the EU ETS started to implement, net imports of both regulated and
non-regulated sectors have increased. For example, since the EU ETS started to implement,
net imports from outside EU (extra EU) have increased by e142.5 per worker; net imports

82

from low-middle income countries (LM) have increased by e69.53 per worker; and net imports from developing countries (DEVG) have also increased by e72.47 per worker. On
the other hand, permit price has negative coefficients with statistical significance. However,
this does not imply that high regulatory stringency induces low net imports; in order to see
the impact we rather see the coefficients of τ × ET S or τ × θ × ET S, which are mostly
positive with no statistical significance. Industrial sectors with high emissions intensity also
tends to be exporters; the coefficients of emissions intensity are all negative in Table 3.1,
but the coefficients are marginally significant in column (3). Traditional trade variables such
as capital-labor ratio, R&D-labor ratio and tariffs have mostly positive coefficients, which
do not show strong significances. In all, it seems that these trading partner groups, each of
which contains a large number of countries, are too heterogeneous to draw general common
trends.

Trading-partner groups are broken down a little further into six sub-groups in Table
3.2 based on the level of income and region; they are, high-income countries excluding EU
members (nonEU HI), OECD countries excluding EU members (nonEU OECD), low-middle
income East Asian and Pacific countries (LMEAP), low-middle income Latin American countries (LMLAC), low-middle income Middle Eastern and North African countries (LMMNA),
and low-middle income South Asian countries (LMSA). Trading-partner groups in columns
(1) and (2) are relatively richer country groups, high-income countries and OECD countries.
Both of them do not include EU member countries. Again, the intention that trading partner
countries are grouped based on the level of income is to capture the imbalance in regulatory
stringency in emissions regulations via the level of income.
Trading partner groups in columns (3) through (6) are part of low-middle income trading
partner groups. Low-middle income countries are thought to be more directly related to
“carbon leakage” from more developed countries and as potential pollution havens. For ex83

Table 3.1: Net imports from large trading partner groups
(1)
(2)
(3)
(4)
extra EU
LM
LDC
DEVG
τ × θ × ET S
-0.00098
0.00035
0.000032
0.00052
(0.0012)
(0.00089)
(0.00048)
(0.0004)
permit price (τ )
-6.888***
-3.205*
0.0379
-3.359**
(2.517)
(1.856)
(0.110)
(1.321)
θ × ET S
2.909
1.402
0.253*
-0.111
(2.160)
(1.280)
(0.136)
(0.484)
capital-labor ratio
0.0205
0.073***
0.0042
0.0308
(0.0493)
(0.0258)
(0.00451)
(0.0217)
R&D-labor ratio
-0.0343
0.169*
-0.00263
-0.0182
(0.113)
(0.0931)
(0.00330)
(0.0403)
tariffs
0.377
1.767
-0.0387
0.301
(1.145)
(1.326)
(0.0514)
(0.387)
ETS dummy (ETS)
76.12
73.16
4.966
30.00
(64.01)
(50.44)
(3.157)
(41.85)
post-ETS dummy (POST) 142.5**
69.53*
-1.078
72.47**
(55.94)
(40.42)
(2.410)
(28.44)
emissions intensity (θ)
-2.907
-1.403
-0.254*
0.112
(2.160)
(1.280)
(0.136)
(0.484)
ET S × P OST
-1.055
3.921
-0.0263
-4.343
(7.770)
(5.481)
(0.279)
(2.838)
τ × ET S
6.868**
1.150
0.0425
1.351
(3.311)
(2.674)
(0.208)
(1.587)
constant
17.14
-8.748
1.795
-7.194
(25.54)
(21.43)
(1.064)
(18.12)
country FE
yes
yes
yes
yes
obs.
718
716
458
697
R2
0.0233
0.0295
0.1074
0.0210
P rob > χ2
0.0068
0.1313
0.0244
0.0855
Robust standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01
Note: extra EU: the rest of the world outside the EU, LM: low-middle income countries,
LDC: least developed countries, and DEVG: developing countries.

84

Table 3.2: The impact of emissions trading on net imports with smaller trading partners
(1)
(2)
(3)
(4)
(5)
(6)
nonEU HI nonEU OECD LMEAP
LMLAC
LMMNA
LMSA
θ × τ × ET S
0.000199
0.00049
0.00057*
0.000078
-0.00527
0.00089***
(0.00074)
(0.00085)
(0.000298) (0.00029) (0.00703) (0.00033)
permit price (τ )
-4.669*** -3.825***
-1.358*** -1.361**
-0.512*
-0.28***
(1.471)
(1.264)
(0.293)
(0.622)
(0.306)
(0.0961)
θ × ET S
0.43
0.675
0.132
0.161
-0.144
-0.0283
(0.934)
(0.793)
(0.17)
(0.317)
(0.231)
(0.065)
capital-labor ratio -0.00088
0.0012
0.0193*** 0.00559
0.0571*** 0.0404***
(0.0142)
(0.0149)
(0.0019)
(0.0016)
(0.00366) (0.00406)
R&D-labor ratio
-0.257*** -0.228***
-0.0124
-0.0210** -0.014
0.000704
(0.0686)
(0.0527)
(0.0135)
(0.00978) (0.0102)
(0.00259)
tariffs
-1.642**
-0.903**
-0.228**
-0.253
-0.191
0.0979
(0.696)
(0.419)
(0.106)
(0.243)
(0.147)
(0.0914)
ETS dummy
0.706
-5.718
1.458
4.025
-2.551
-1.114
(ETS)
(29.9)
(18.19)
(3.819)
(7.336)
(5.641)
(1.693)
post-ETS dummy 97.56***
80.26***
29.98***
30.74**
10.44
5.946***
(POST)
(32.82)
(27.66)
(6.271)
(14.14)
(6.367)
(2.055)
emissions intensity -0.423
-0.672
-0.132
-0.162
0.137
0.027
(θ)
(0.932)
(0.792)
(0.17)
(0.317)
(0.230)
(0.0651)
ET S × P OST
-8.368*
-4.553
-0.251
-2.439
-1.147
0.948
(4.882)
(3.848)
(0.667)
(1.496)
(1.255)
(0.762)
τ × ET S
4.319**
2.842
-0.446
1.621**
-0.275
0.0583
(1.985)
(2.058)
(0.66)
(0.718)
(0.733)
(0.226)
constant
8.269
15.13*
3.22**
1.442
-0.791
1.317
(9.383)
(9.096)
(1.551)
(2.395)
(1.608)
(1.52)
country FE
yes
yes
yes
yes
yes
yes
obs.
718
717
598
573
559
529
2
R
0.0545
0.0577
0.1613
0.1018
0.2265
0.1863
2
P rob > χ
0.0000
0.0000
0.0000
0.3971
0.0000
0.0000
Robust standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01
Note: nonEU HI: non-EU high income countries, nonEU OECD: non-EU OECD members,
LMEAP: low-middle income East Asian and Pacific countries, LMLAC: low-middle Latin American countries, LMMNA: low-middle income Middle East and North African countries, and LMSA:
low-middle South Asian countries.

85

ample, LMEAP trading partner group includes China, Indonesia, and Malaysia; LMLAC
includes Argentine, Brazil, and Chile; and LMSA includes India.6 Hence, the impact of
emissions trading with low-middle income countries is expected to be more evident.
Our primary interest lies on the relative changes of net imports between regulated and nonregulated industrial sectors after the EU ETS and hence the coefficients of (ET S × θ × τ )
are those that attention should be paid to. With relatively richer trading partner groups
in columns (1) and (2) a significant regulatory impact on ETS sectors relative to non-ETS
sectors is not detected; the coefficients of (ET S × θ × τ ) are positive with no statistical
significance. The emissions regulation through the EU ETS has no noticeable impact on net
imports from relatively richer trading partner groups. On the other hand, net imports from
the low-middle income trading partner groups in columns (3) to (6) exhibit some recognizable differences between regulated and non-regulated industrial sectors; the coefficients of
(ET S × θ × τ ) are mostly positive with some statistical significances. Especially in columns
(3) and (6), net imports of regulated sectors have increased compared to non-regulated sectors since the EU ETS started, and among regulated sectors higher emissions intensity leads
to larger net imports. With low-middle income East Asian and Pacific countries (LMEAP)
an additional euro increment in individual regulatory stringency on regulated sectors leads
to increase net imports more by e0.000569 per worker (relatively to non-regulated sectors)
after the EU ETS. Similarly, with low-middle income South Asian countries, an additional
euro increment of individual regulatory stringency on regulated sectors leads to increase net
imports more by e0.000886 per worker (relatively to non-regulated sectors) after the EU
ETS. Similar interpretations are applied to results of low-middle income Latin American
countries. The overall contrasting outcomes between relatively richer trading partner groups
and low-middle income trading partner groups are interesting, yet reasonable. Among trading partner groups at different income levels the regulatory stringencies on CO2 emissions
6 See

Appendix B.4 for the full list of trading partner groups.
86

may vary much. And it is generally conceded that the environmental quality of a country
exhibits a certain relationship with it income level; in the early stage of economic development, the environmental quality worsens as income grows, but this trend reverses if the
country passes the certain stage of development (See World Bank (1992); Grossman and
Kreuger (1995); Stokey (1998); Brock and Taylor (2005)). Hence, relatively richer countries
tend to have more stringent regulations on emissions than middle-income countries. Therefore, additional regulatory stringency may have a less impact on trade with highly regulated
countries than on trade with loosely regulated low-middle income countries. The contrasting
outcomes depending on income level can be a piece of supporting evidence for the pollution
haven hypothesis. Trade-related variables also have different impacts across trading partner
groups. Traditional Heckscher-Ohlin type input variables work well with relatively richer
trading partner groups, high-income countries (non-EU HI) and OECD countries (non-EU
OECD). Capital-labor ratio known usually as capital intensity plays an important role in net
imports from relatively richer trading partner groups. Since these emission-intensive manufacturing sectors are also known as capital-intensive, relative capital-abundance is expected
to reduce net imports. With relatively richer trading partner groups in column (1) and (2),
the regression outcomes support the argument. In column (1) with non-EU high-income
countries the coefficient of capital-labor ratio is -0.0571 at 1 % significance level. In other
words, one extra euro increment of capital per worker leads the net imports to decrease
by e0.0000571 per worker (capital is measured in billion eand the number of employees in
million). With the non-EU OECD countries in column (2) the coefficient of capital-labor
ratio is also negative, -0.0404, at 1% significance level; one extra euro increment of capital
per worker leads the net imports to decrease by e0.00004 per worker (capital is measured
in billion eand the number of employees in million). However, the results from low-middle
income counterparts are deviant from the expected results. With low-middle income East
Asian and Pacific countries (LMEAP) in column (3) the coefficient of capital-labor ratio is
87

negative with no statistical significance. With low-middle income Latin American countries
(LMLAC) in column (4) and low-middle income South Asian countries (LMSA) in column
(6) the coefficients of capital-labor ratio are positive with no statistical significance. Interestingly, with low-middle income Middle East and North African countries (LMMNA) in
column (5) the coefficients of capital-labor ratio are positive and statistically significant. It
requires further research, but one possible explanation is that many of Middle East and
North African countries are the exporters of crude oil, which is a heavily capital intensive
sector. R&D-labor ratio, a proxy for human capital intensity, exhibit some significant impact
on trade flows. With relatively richer trading partner groups in column (1) and (2), R&D
worker ratio to the total employment has the negative coefficient, indicating that high human
capital intensity helps industrial sectors to increase exports and reduce imports. The coefficients in columns (1) and (2) are statistically significant at 1% level. One additional R&D
personnel out of 1000 workers leads net imports to decrease by e0.257 per worker and by
e0.228 in columns (1) and (2), respectively. In the cases of low-middle income trading partner groups, the coefficients have mostly negative signs too. Tariffs also have different impacts
between relatively richer trading partner groups and low-middle income partner groups. But,
the overall results are consistent with theoretical expectations. With high-income countries
and OECD countries, tariffs have clear significant impacts on trade flows consistent with
the theoretical prediction. In columns (1) and (2), the coefficients of tariffs are -1.642 and
-0.903 at 5% statistical significance. In other words, with high-income countries one percent
increment of tariff rate leads net imports to decrease by e1.642 per worker. Similarly, with
OECD countries one percent increment of tariff rate leads net imports to decrease by e0.903
per worker. Tariffs imposed on relatively richer trading partner groups discourage imports.
However, with low-middle income countries the impact of tariff is weaker. With low-middle
income East Asian Pacific (LMEAP) countries, the coefficient of tariff is negative with statistical significance. With trading partner groups LMLAC and LMMNA tariff has negative
88

coefficients with no statistical significance. In the case of LMSA, the coefficient of tariff is
positive with no statistical significance. Compared to relatively richer trading partner groups
(nonEU HI and nonEU OECD), the magnitude of the coefficient is smaller and statistical
significance is also weaker.

3.5

Conclusion

This study adds a piece of evidence that environmental regulations such as emissions trading may have an unfavorable impact on trade flows of regulated sectors. However, empirical
analyses show fairly modest results about the unfavorable economic impact: trade with relatively richer countries has not affected significantly, and trade with relatively less developed
countries shows some unfavorable impact but the magnitudes are small. Pollution haven effect from implementing the emissions trading seems existing; however, the impact is limited.
More specific contributions can be summarized as follows. This study focuses on a specific
environmental policy, emissions trading system. By doing so, it is avoided to use overall environmental spending data such as pollution abatement cost (PAC) measures as a proxy for
the stringency of environmental regulation. Instead, permit price generated by the permit
market is used because it has clear marginal cost implications on ETS participating sectors.
Since permit price is determined by the market, potential endogeneity problem from the
trade policy intervention can be excluded. Furthermore, by providing a theoretical model
the role of emissions intensity with permit price is emphasized in optimal decision makings
of regulated sectors. Heterogeneity in emissions intensity leads industrial sectors to make
different decisions for the same regulatory stringency (permit price). Considering emissions
intensity also helps to identify each industrial sector in the empirical analysis.
Empirical results are consistent with theoretical predictions provided in the study and traditional theories such as pollution haven hypothesis. With six different trading partner

89

groups, the results exhibit weak yet consistent evidence that the trade flows with middleincome trading partner groups have been influenced by the implementation of the EU ETS
while trade with relatively richer trading partner groups have not been affected significantly.
Net imports of regulated sectors from some sub-trading partner groups in low-middle income
group have increased relative to non-regulated counterparts after the EU ETS. On the other
hand, net imports of regulated sectors from high-income trading partners and OECD member group are not affected by the EU ETS. The empirical outcomes can be interpreted in the
viewpoint of the pollution haven hypothesis; the stringent environmental regulation makes
the regulated sectors to contract or leave to other loosely regulated countries. Therefore,
regulated sectors increase net imports relative to non-regulated sectors.
Even though the theoretical and empirical analysis seems to exhibit some unfavorable impact of emissions trading on trade flows, this is not a warning sign urging to quit emissions
trading. The exhaustive assessment about the emissions trading should include welfare implications with consideration of dynamics.

90

APPENDIX

91

Appendix B

B.1

Summary Statistics
Table B.1: Summary statistics (net imports mil. e(base year=2000))
Variable

Mean

Std.
Min.
Dev.
extra EU
-173.346
1410.218
-12475.62
low-middle income
-7.427
837.082
-9089.531
least-developed
-10.29
72.074
-682.704
developing
-45.526
479.103
-5082.395
high-income (excl. EU)
-40.933
718.800
-8119.06
OECD (excl. EU)
-66.455
695.057
-8086.366
East Asia and Pacific
1.044
123.721
-959.846
Latin America
-8.834
239.033
-1821.235
Middle East and North Africa -27.439
166.304
-1821.877
South Asia
-2.296
45.694
-384.628
Source: World Integrated Trade Solution, World Bank.

92

Max.

obs.

9190.133
6947.321
619.653
4915.045
6935.098
4591.938
2185.661
2783.887
1016.41
523.871

1810
1795
1002
1722
1833
1836
1468
1327
1311
1237

Table B.2: Summary statistics (Explanatory variables)
Variable
capital (mil. e)
R&D expenditure (mil. e)
number of employees (mil.)
emissions intensity†
permit price (e/MtCO2)

Mean
44586.2
104.31
18.464
6.478
20.389

Std.
Dev.
311559.3
462.474
34.287
10.318
1.216

Min.

Max.

obs.

0.8
0
0.01
0.0083
18.179

5493996
4973.6
282.374
90.511
21.877

1347
910
1369
1470
936

16.09
22.51
19.71
20
16.51
17.69
15
20
19.7
18.08

1781
1781
1549
1771
1781
1781
1757
1721
1694
1663

Tariffs (%)
extra EU
2.365
2.216
0
low-middle income
2.428
2.186
0
least-developed
2.683
2.231
0
developing
2.448
2.261
0
high-income
2.308
2.123
0
OECD
2.33
2.16
0
East Asia and Pacific
2.442
2.199
0
Latin America
2.243
1.994
0
Middle East and North Africa 2.56
2.231
0
South Asia
2.652
2.232
0
MtCO2: metric ton of CO2. rGDP: real GDP in 2000.
† :measured in MtCO2/rGDP in emil.
Source: EUROSTAT and World Integrated Trade Solution.

93

B.2

country
AT
BE
CY
CZ
DE
DK
EE
ES
FI
FR
GR
HU
IE
IT
LT
LU
LV
MT
NL
PL
PT
SE
SI
SK
UK
avg
s.e.

The Heterogeneity of Sectoral Emissions Intensity of the EU
ETS sectors
Emissions (1 MtCO2) per
combustion refinery
coke ovens
33.997
10.241
1.871
63.067
14.028
0.623
235.604
417.039
6.576
3.997
127.335
8.456
5.436
95.426
4.193
6.933
741.709
31.780
71.222
11.041
2.246
129.094
16.546
1.096
23.040
6.871
1.762
215.109
16.933
0.366
157.636
13.738
3.054
78.087
1.973
0.908
64.411
16.616
6.074
119.608
64.402
0.601
30.330
98.444
2.552
317.509
91.885
17.031
3.386
507.590
18.086
11.203
102.401
13.281
25.956
7.192
0.873
170.917
0.177
0.059
105.672
29.046
20.845
95.846
9.460
9.641
164.757
14.294
5.491
169.260
13.518
7.755

million GDP (e) in 2000
steel
cement
glass
35.103
8.406
1.380
14.757
12.367
1.150
0.000
40.209
2.291
62.852
15.978
2.277
9.693
7.030
0.416
0.399
3.631
0.152
0.069
19.692
0.628
6.996
12.176
0.782
24.751
4.341
0.053
7.026
5.118
0.348
1.262
20.724
0.057
25.442
12.682
1.766
0.015
9.290
0.003
6.219
9.965
0.348
0.152
10.987
0.560
11.934
9.930
1.631
14.560
10.022
0.183
11.369
0.000
8.968
0.728
0.398
31.726
22.779
1.425
0.932
21.254
0.751
8.446
5.773
0.183
5.213
14.251
0.243
93.326
29.946
4.423
9.761
2.776
0.047
15.639
12.402
0.896
21.654
9.358
1.030

Source: European Environmental Agency. Current GDP in 2000.

94

paper
7.998
1.375
3.924
0.006
0.699
0.279
4.491
18.642
1.821
0.840
0.966
0.075
2.502
1.902
0.423
0.301
1.907
4.030
3.847
4.706
11.536
10.031
3.741
4.615

B.3

The comparison of ETS and non-ETS sectors
Table B.3: Summary statistics of net imports before ETS (ETS sectors, mil e)

Variable
extra-EU
non-EU HI
non-EU OECD
LM
LDC
LMEAP
LMLAC
LMMNA
LMSA
Source: World

Mean
Std. Dev.
-84.825
549.832
-65.612
255.344
-56.776
255.89
55.387
371.726
-11.521
19.14
-10.611
76.095
2.108
60.112
6.396
153.47
-5.932
20.91
Integrated Trade Solution.

Min.
-3128.354
-1636.472
-1387.24
-1350.308
-100.303
-626.108
-275.819
-365.941
-147.18

Max.
1595.355
911.338
1035.159
2397.364
7.351
137.668
255.801
1016.41
84.135

obs.
418
413
414
397
203
298
273
270
237

Table B.4: Summary statistics of net imports before ETS (non-ETS sectors, mil. e)
Variable
extra EU
non-EU HI
non-EU OECD
LM
LDC
LMEAP
LMLAC
LMMNA
LMSA
Source: World

Mean
Std. Dev.
-250.092
1548.283
-31.74
825.634
-50.312
832.516
-105.158
740.751
-8.884
68.113
-11.588
78.379
-33.115
215.652
-41.464
156.043
-6.33
34.216
Integrated Trade Solution.

Min.
-10911.86
-8119.06
-8086.366
-5227.179
-573.464
-560.806
-1350.414
-1468.461
-259.927

95

Max.
5603.282
2829.08
2599.425
2966.341
364.55
221.433
839.431
112.036
75.56

obs.
504
503
503
501
276
407
370
368
357

B.4

Country Classification (GNI per capita)

Category
Country name
High-income
Andorra, Aruba, Australia, The Bahamas, Bahrain, Bermuda, Brunei,
countries (non- Canada, Cayman Islands, Faeroe Islands, French Polynesia, Greenland,
EU HI)
Guam, Hong Kong (China), Iceland, Israel, Japan, Korea Rep., Kuwait,
Liechtenstein, Macao, Monaco, Netherlands Antilles, New Caledonia,
New Zealand, Northern Mariana Islands, Norway, Qatar, San Marino,
Singapore, Switzerland, United Arab Emirates, United States, Virgin
Islands (U.S.)
OECD members
(non-EU OECD)

Australia, Canada, Iceland, Japan, Korea Rep., Mexico, Netherlands,
New Zealand, Norway, Switzerland, Turkey, United States

East Asian & American Samoa, Cambodia, China, East Timor, Fiji, Indonesia, KiriPacific countries bati, Korea Dem. Rep., Lao PDR, Malaysia, Marshall Islands, Microne(LMEAP)
sia Fed. Sts., Mongolia, Myanmar, Northern Mariana Islands, Palau,
Papua New Guinea, Philippines, Samoa, Solomon Islands, Thailand,
Tonga, Vanuatu, Vietnam
Latin
Ameri- Argentina, Belize, Bolivia, Brazil, Chile, Colombia, Costa Rica,
can
countries Cuba, Dominica, Dominican Republic, Ecuador, El Salvador, Grenada,
(LMLAC)
Guatemala, Guyana, Haiti, Honduras, Jamaica, Mexico, Nicaragua,
Panama, Paraguay, Peru, St. Kitts and Nevis, St. Lucia, St. Vincent
and the Grenadines, Suriname, Uruguay, Venezuela
Middle East and Algeria, Djibouti, Egypt Arab Rep., Iran Islamic Rep., Iraq, Jordan,
North
Africa Lebanon, Libya, Morocco, Syria, Tunisia, Yemen
(LMMNA)
South East Asia Afghanistan, Bangladesh, Bhutan, India, Maldives, Nepal, Pakistan, Sri
(LMSA)
Lanka
Source: World Bank (Atlas Method)

96

Category
Country name
Low-middle in- Afghanistan, Albania, Algeria, American Samoa, Angola, Antigua and
come countries Barbuda, Argentina, Armenia, Azerbaijan, Bangladesh, Belarus, Belize,
(LM)
Benin, Bhutan, Bolivia, Bosnia and Herzegovina, Botswana, Brazil, Bulgaria, Burkina Faso, Burundi, Cambodia, Cameroon, Cape Verde, Central African Republic, Chad, Chile, China, Colombia, Comoros, Congo
Dem. Rep., Congo Rep., Costa Rica, Cote d’Ivoire, Cuba, Djibouti, Dominica, Dominican Republic, East Timor, Ecuador, Egypt Arab Rep., El
Salvador, Eritrea, Ethiopia(excludes Eritrea), Fiji, Gabon, The Gambia,
Georgia, Ghana, Grenada, Guatemala, Guinea, Guinea-Bissau, Guyana,
Haiti, Honduras, India, Indonesia, Iran Islamic Rep., Iraq, Jamaica, Jordan, Kazakhstan, Kenya, Kiribati, Korea Dem. Rep., Kosovo, Kyrgyz Republic, Lao PDR, Lebanon, Lesotho, Liberia, Libya, Lithuania, Macedonia
FYR, Madagascar, Malawi, Malaysia, Maldives, Mali, Marshall Islands,
Mauritania, Mauritius, Mayotte, Mexico, Micronesia Fed. Sts., Moldova,
Mongolia, Morocco, Mozambique, Myanmar, Namibia, Nepal, Nicaragua,
Niger, Nigeria, Pakistan, Palau, Panama, Papua New Guinea, Paraguay,
Peru, Philippines, Romania, Russian Federation, Rwanda, Samoa, Sao
Tome and Principe, Senegal, Seychelles, Sierra Leone, Solomon Islands,
Somalia, South Africa, Sri Lanka, St. Kitts and Nevis, St. Lucia, St.
Vincent and the Grenadines, Sudan, Suriname, Swaziland, Syrian Arab
Republic, Tajikistan, Tanzania, Thailand, Togo, Tonga, Tunisia, Turkey,
Turkmenistan, Tuvalu, Uganda, Ukraine, Uruguay, Uzbekistan, Vanuatu,
Venezuela, Vietnam, Yemen, Zambia, Zimbabwe
LeastAfghanistan, Angola, Bangladesh, Benin, Bhutan, Burkina Faso, Budeveloped
rundi, Cambodia, Central African Republic, Chad, Comoros, Congo, Dem.
countries
Rep., Djibouti, East Timor, Equatorial Guinea, Eritrea, Ethiopia(excludes
(LDC)
Eritrea), Gambia, Guinea, Guinea-Bissau, Haiti, Kiribati, Lao PDR,
Lesotho, Liberia, Madagascar, Malawi, Mali, Mauritania, Mozambique,
Myanmar, Nepal, Niger, Rwanda, Samoa, Sao Tome and Principe, Senegal, Sierra Leone, Solomon Islands, Somalia, Sudan, Tanzania, Togo, Tuvalu, Uganda, Vanuatu, Yemen, Zambia.
Developing
Albania, Antigua and Barbuda, Argentina, Armenia, Barbados, Belize,
members
Bolivia, Botswana, Brazil, Bulgaria, Chile, China, Colombia, Costa Rica,
(DEVG)
Croatia, Dominica, Dominican Republic, Ecuador, Egypt Arab Rep., El
Salvador, Fiji, Gabon, Georgia, Grenada, Guatemala, Guyana, Honduras,
India, Indonesia, Jamaica, Jordan, Kyrgyz Republic, Macedonia FYR,
Malaysia, Mexico, Moldova, Mongolia, Morocco, Namibia, Nicaragua,
Oman, Pakistan, Panama, Papua New Guinea, Paraguay, Peru, Philippines, Romania, South Africa, St. Kitts and Nevis, St. Lucia, St. Vincent
and the Grenadines, Swaziland, Thailand, Trinidad and Tobago, Tunisia,
Turkey, Uruguay, Venezuela
Source: World Bank (Atlas Method)
97

B.5

Robustness Checks

The following empirical results come from a bit different identification strategy, which uses
less interaction terms. Then, year dummies fore pre-policy periods and sector dummies for
non-regulated sectors are included. The results are mostly consistent with those in 3.4. The
most significant difference is shown in the effect of individual regulatory stringency in Table
B.6. The coefficients of ET S × θ × τ are positive with high statistical significances when
the trading partner groups are low-middle income countries. In other words, the pollution
haven effect is more significantly shown.

98

Table B.5: Net imports from large trading partner groups II
(1)
(2)
(3)
(4)
extra EU
LM
LDC
DEVG
capital-labor ratio
0.0196
0.0663**
0.00461
0.0331
(0.0438)
(0.0260)
(0.00437)
(0.0224)
R&D-labor ratio
-0.0300
0.169*
0.00308
-0.0189
(0.118)
(0.0933)
(0.00297)
(0.0411)
tariffs
0.512
1.701
-0.0667
0.481
(1.245)
(1.332)
(0.0565)
(0.422)
ETS dummy (ETS)
-82.30***
-62.73**
-5.259**
-19.15
(30.81)
(28.06)
(2.187)
(23.75)
emissions intensity (θ) -3.231
-1.368
-0.260*
-0.0953
(2.013)
(1.293)
(0.141)
(0.470)
ET S × θ
3.246
1.364
0.259*
0.0987
(2.011)
(1.294)
(0.140)
(0.472)
permit price (τ )
0.0174
-1.722
-0.0756*
1.404*
(2.639)
(2.448)
(0.123)
(0.849)
ET S × θ × τ
-0.000709
0.000151
-0.000112
-0.000148
(0.00195)
(0.00111)
(0.000174)
(0.000469)
constant
35.33
27.88
5.081***
6.719
(30.27)
(22.89)
(1.971)
(18.16)
country dummy
yes
yes
yes
yes
sector dummy†
yes
yes
yes
yes
††
year dummy
yes
yes
yes
yes
obs.
718
716
458
697
2
R
0.1654
0.2044
0.2737
0.1966
Note: Robust standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01
†: sector dummies for non-ETS sectors. ††: year dummies before ETS years. extra EU: the
rest of the world outside the EU, LM: low-middle income countries, LDC: least developed
countries, and DEVG: developing countries.

99

Table B.6: Net imports from sub-group trading partners II
(1)
HI
capital-labor ratio

0.0388***
(0.00826)
R&D-labor ratio
-0.264***
(0.0644)
tariffs
-1.654**
(0.710)
ETS dummy
-18.96
(11.95)
emissions intensity -0.457
(θ)
(0.884)
ET S × θ
0.709
(0.892)
permit price
4.445**
(τ )
(2.098)
ET S × θ × τ
-0.013
(0.0087)
constant
23.36
(15.48)
country dummy
yes
sector dummy†
yes
year dummy††
yes
obs.
718
2
R
0.3151

(2)
OECD
0.0267***
(0.00933)
-0.235***
(0.0500)
-1.002**
(0.473)
-17.97*
(9.870)
-0.347
(0.717)
0.382
(0.717)
2.027
(1.699)
-0.0017
(0.0012)
20.64
(14.21)
yes
yes
yes
717
0.1823

(3)
LMEAP
0.000353

(4)
LMLAC
0.00288*

(5)
(6)
LMMNA
LMSA
0.0163*** 0.00599

(0.00185)
-0.00686
(0.0124)
0.119
(0.142)
-1.615
(1.471)
-0.115
(0.182)
0.110
(0.182)
0.118
(0.172)
0.00023ψ
(0.00009)
-1.816
(2.047)
yes
yes
yes
598
0.3174

(0.00150)
-0.00979
(0.00645)
0.254
(0.229)
-10.77***
(3.857)
-0.0130
(0.293)
0.00372
(0.293)
1.209*
(0.647)
0.00048ψ
(0.00012)
8.729**
(3.542)
yes
yes
yes
573
0.3115

(0.00528)
-0.00745
(0.0102)
-0.241*
(0.137)
-0.639
(1.083)
0.0378
(0.238)
-0.0573
(0.238)
0.611
(0.568)
0.0011ψ
(0.00005)
-0.251
(1.072)
yes
yes
yes
559
0.2804

(0.00442)
0.000489
(0.00296)
0.00252
(0.110)
-0.274
(0.426)
0.0186
(0.0644)
-0.0196
(0.0643)
-0.562
(0.428)
0.000038ψ
(0.00001)
-0.840
(0.548)
yes
yes
yes
529
0.3753

Note: Robust standard errors in parentheses. *: p < 0.1, **: p < 0.05, ψ : p < 0.01.
†: sector dummies for non-ETS sectors. ††: year dummies before ETS years. HI: non-EU
high income countries, OECD: non-EU OECD members, LMEAP: low-middle income East
Asian and Pacific countries, LMLAC: low-middle Latin American countries, LMMNA:
low-middle income Middle East and North African countries, and LMSA: low-middle South
Asian countries.

100

Chapter 4
MARKET EFFICIENCY IN THE EU EMISSION TRADING SCHEME

4.1

Introduction

Emissions trading is designed as a regulatory system that can efficiently accomplish the
emission reduction to a designated target at a minimum cost (Croker (1966); Dales (1968);
Montgomery (1972)). One reason why emissions trading is a preferable emissions reduction
tool is that the overall emission target is always met at the minimum cost as permit market
clears by flexible price. Furthermore, most emissions trading systems around the globe start
the system with grandfathering permit allocation method for some reason. Theoretically, the
permit allocations have no implications on the equilibrium decisions of output or emissions
as far as the permit market is efficient. Of course, the efficiency of emissions trading depends
heavily on the efficiency of the permit market. Therefore, the potential permit market inefficiency has been constantly discussed and researched.
This paper investigates the permit market efficiency with the data from the EU emissions
trading system (EU ETS), which is by far the largest multi-national, multi-sectoral emissions trading system in the world. One of ways testing the permit market efficiency is to
examine the irrelevance of the initial permit allocation with the equilibrium decisions of
individual firms. Economic theory tells that predetermined permit allocations should not
affect the equilibrium decisions as far as market is efficient. In the terminology of Hahn and
Stavins (2010), the independence property holds when the equilibrium decisions of firms are
not affected by the initial permit allocations. They also mention the cases that can hurt
the market efficiency: transaction costs, market power, uncertainty, conditional permit allocations (output-based updating allocations), non-cost-minimizing behavior by firms, and

101

differential regulatory treatment of industrial sectors. However, the most relevant factor that
can hurt the permit market efficiency in the EU ETS eventually narrows down to transaction costs, because other factors do not fit into the characteristics of the EU ETS or based
on non-rational behavior of firms. Stavins (1995) points transaction costs as a culprit that
can undermine market efficiency. He shows that if transaction cost exists and the marginal
transaction cost changes depending on the volume of permit trading, then the equilibrium
results even in a competitive setting will be affected by the permit allocations. Eventually,
the post-trading equilibrium even in the competitive permit market may not coincide with
the desired cost-effective equilibrium. Hahn (1984) addresses a case with market power and
shows that if permit markets are not perfectly competitive, then the initial allocation may
matter so that the ideal efficient market equilibrium may not be accomplished. However,
the EU ETS is composed of nine sectors in twenty seven member states. By design, it is
highly unlikely that one of industrial sectors in a country can wield a dominant market
power. Montero (1997) goes one step further and considers uncertainty based on the Stavins
(1995)’s model. He shows that in the presence of transaction cost and trading uncertainty
the initial allocation may not be neutral in terms of efficiency and essentially uncertainty
increases the degree to which transaction impede mutually-beneficial trades. Output-based
updating allocation is thought of as a production subsidy that can affect the post-trading
allocation and hence can affect the efficiency (Fischer (2001); Fischer and Fox (2007)). This
type of permit allocation is justified by a desire to protect certain industrial sectors from
adverse competitiveness impacts from the regulation. Such modified allocation hurts the
market efficiency (Houser, Bradley, Childs, Werksman and Heilmayr (2008); Fowlie (2010)).
In the case of the EU ETS, output-based updating allocation is not relevant because there
is no annual update of permit allocation in first two phases of the EU ETS and then from
the third phase permits are allocated by auctioning. Different regulatory treatments of industrial sectors seem possible in the EU ETS. Every single participant in the EU ETS faces
102

the same permit price, which is determined solely at the permit market. Regarding permit
price, there is no way for one industrial to have different treatment from another. In terms
of permit allocation in the EU ETS, however, the national allocation plans are designed by
individual countries and they can be approved by the European Commissions as far as the
overall national target is not altered. Therefore, there is chance for a local government to
manipulate permit allocation in favor of certain industrial sectors. This factor is related
to the permit allocation method (i.e. grandfathering). The endogeneity problem from the
initial permit allocation is explained how to control for in detail below. Finally, the noncost-minimizing behavior of firms might be possible, but is hard to be justified. Hence,
until there is substantial evidence for the non-cost-minimizing behavior, this paper assumes
that firms under the regulation behave based on the traditional optimal decision-makings.
Then, the immediately following question is whether the permit allocations actually affect
the marginal decision of production or emission level through transaction cost in the EU
emissions trading system.
The empirical study for independence property goes beyond just checking the correlation
between initial permit allocations and equilibrium decisions of outputs or emissions. In the
regression of equilibrium decisions of production and emissions on the initial permit allocation, the null hypothesis to be tested is whether the initial permit allocation has zero
coefficient. If the EU ETS is the cost-effective efficient market, then the initial allocation is
irrelevant with post-trading equilibrium decisions and hence the zero coefficient is expected.
However, there is a reason that the empirical results may show false casual relationships
between the initial permit allocation and equilibrium decisions. That is the grandfathering
(free-of-charge) permit allocation method mostly based on historical emissions and output.
Note that grandfathering permit allocation itself does not hurt the market efficiency unlike
six factors mentioned above. Then, we need to separate the non-efficiency-distorting factor that attributes to historical emissions and output from efficiency-distorting factors and
103

control for non-efficiency-distorting factor so that we can see whether the market is really
efficient and cost-effective. The following explains why the initial allocation can show a false
causal relationship with post-trading equilibrium. The grandfathering permit allocation in
the EU ETS has not been free from the political interest groups, which usually try to protect
heavy emitters from worsening in comparative (dis)advantage in the global market. Thus,
these political pressure leads the permit allocation decisions to be highly proportionate to
the historical emissions or output levels. Then, there is high chance that the initial permit
allocation is positively correlated with the (post-trading) equilibrium decisions of output or
emissions. However, this paper emphasizes that such positive correlations do not necessarily
imply the causal relationship. Appendix C.3 depicts the seeming correlations between the
initial allocation and (post-trading) equilibrium decisions of output and emissions, which
show the positive correlations. In order to resolve such endogeneity problem of permit allocation in empirical analysis, two stage least squares method is adopted. The first chapter
of the dissertation tries to explain how the initial permit allocation can be determined theoretically and empirically. The empirical specification in the first chapter is used to obtain
the fitted value of the initial permit allocation. Then, the fitted values are substituted for
the actual initial permit allocation of the EU ETS.
This type of endogeneity problem has been pointed out in the literature. One recent paper
that studies the permit allocation independence is written by Fowlie and Perloff (2008), who
empirically test for a causal relationship between facility-level emissions and initial permit
allocations in South California’s RECLAIM program. Using the random assignment of firms
to different permit allocation cycles they find that initial allocation does not have a statistically significant impact on the equilibrium outcome.
First, a straightforward conceptual model is introduced based on Stavins (1995). The distinctive feature is that this paper investigates not only abatement decision but the production
decision as well. If the shape of marginal transaction cost function is not linear, then the
104

post-trading equilibrium deviates from the cost-effective equilibrium. The direction and size
of the deviation depend on the shape of marginal transaction cost function. If the marginal
transaction cost is an increasing (decreasing) function in the volume of permit trading, the
extra unit increment of permit allocation leads the post-trading equilibrium output to increase (decrease) and the post-trading equilibrium abatement to decrease (increase). The
empirical results tell that the performance in phase I of the EU ETS is partially under the influence of the initial permit allocation. The productions at sector level do not depend on the
initial permit allocation after controlling the endogeneity. However, the levels of emissions
at sector level exhibit a persistent correlation to the initial permit allocations regardless of
the endogeneity control. At installation level, empirical results indicate the significant influence of the initial allocations on final emissions in the entire phase I. However, the result
tested with data including first two years of phase II shows that the significant influence
of the initial allocations on final emissions vanishes, this could be interpreted, with caveat,
that the permit market had been settling down over time as a more efficient market. This
paper proceeds as follows. In section 4.2, a conceptual model is introduced based on Stavins
(1995) and shows how the initial permit allocation can affect the equilibrium decisions in the
presence of transaction cost. Section 4.3 explains data and variables used in this study. In
section 4.4, empirical models are built based on the suggestion from the conceptual model.
Section 4.5 provides results. Finally, section 4.6 concludes.

4.2

Theoretical Framework

A representative firm of sector i (i = 1, ..., N ) produces a homogeneous good at market price
in the competitive market. When the emissions trading as an environmental regulation is
implemented, firms under the regulations have to deal with new costs. The profit function

105

for a representative firm participating in the emissions trading is given by:

π = px − C(x) − D(a) + τ (A + a − θx),

(4.1)

where x denotes the amount goods produced at market price p; C(x) is the total production
cost of x (Cx > 0 and Cxx > 0); a denotes the emissions abatement and D(a) is the
corresponding abatement cost (Da > 0 and Daa > 0); permit price determined in the
permit market is denoted by τ ; A is predetermined (tradable) permits allocated from the
government (free of charge); finally, θ indicates the emissions intensity, or how much carbon
dioxide is emitted with the production of x. In this profit function, we can easily see that
firm’s predetermined initial allocation A cannot affect its equilibrium production x. Rather,
output is a function of output price, permit price, and emissions intensity i.e. x = x(p, τ, θ).
However, once transaction cost of permit trading is considered in the analysis, they may
lead to different results for the effect of initial allocation on equilibrium decisions. Let t
denote the quantity of permits traded by a firm and it is defined by the absolute difference
between actual emissions level and the number of permits allocated:

t = |v − A| ,

(4.2)

where v = θx − a is the level of emissions after abatement. Based on the volume of permits
traded, a common transaction cost function is defined as T (t), for which is continuous and
Tt > 0 and Ttt can be positive, negative or zero. All participants of the emissions trading
pay the same fixed cost for the trading, if any, regardless of whether a participant trades
permits or not. For instance, each regulated installation has to register, report and verify the
level of its emissions, and submit permits as many as emitted. These costs are not variable
depending on the level of emissions; however, they happens as far as installations are under

106

the emissions trading system regardless of whether they trade permits or not.1 By assuming
this, the discontinuous point at zero trading can be excluded. Keep in mind that the trading
volume t is not a choice variable so that there is no optimal trading point. Only choice
variables are output and abatement and hence even though the permit trading volume of
a certain firm appears zero, that is not the point that a firm pick as the optimal point; it
is rather a coincidental combination of output, abatement and permit allocation.2 Now the
firm’s profit function becomes:

π = px − C(x) − D(a) + τ (A + a − θx) − T (t).

(4.3)

Without loss of generality, only the case of permit buyers is considered ( v > A ). Then,
the first order conditions follow. First, with respect to output the first order condition is:

πx = p − Cx − τ θ − θTt = 0.

(4.4)

First order condition tells that the level of output is decided where the marginal revenue
(p) is equal to the marginal cost (Cx + τ θ + Tt ). In another way, producer price (p − τ θ − θTt )
should be equal to the marginal production cost (Cx ). Since it is implicitly assumed that
the regulation does not put firms in a position that they have to shut down their businesses,
only positive producer price and thereby positive outputs are considered (p − τ θ − θTt > 0).
However, it is possible that firms’s optimal choices are zero abatement (a ≥ 0). This problem
then yields the following solution:
1 It may then be possible that the registration fee is so high that no installation is willing to

register and trade permits. In this case, the equilibrium decisions on emissions may strictly
depend on the initial allocation. However, in this paper low enough fixed costs in permit
trading is assumed so that the decision making process of each installation depend only on
variable costs.
2 A different modeling technique with the explicit trading price, which is different from
transaction cost, can be applied if the optimal trading point is pursued. However, this is not
pursued in the paper.
107

− Da + τ + Tt ≤ 0

(4.5)

a (−Da + τ + Tt ) = 0

(4.6)

a ≥ 0.

(4.7)

If the optimal abatement is positive, then both output and abatement are a function of
output price, permit price, emissions intensity, and the number of permits allocated i.e.
x = x(p, τ, θ, A) and a = a(p, τ, θ, A). If the optimal choice of abatement is zero, then
our focus lies only on output decision influenced by the permit allocation. The following
discussion is based on the positive abatement decision. To see the effect of permit trading on
output and abatement, the first order conditions are total-differentiated and then the partial
derivatives of the emission function are obtained:

θDaa Ttt
dx
=
dA
|H|
da
Cxx Ttt
=−
.
dA
|H|

(4.8)
(4.9)

The second order conditions for profit maximization imply the positive Hessian, |H| =
Cxx Daa + Ttt (Cxx + θ2 Daa ) > 0. As we can see, the change of output and abatement
dx
da
depends on the sign of Ttt . If Ttt = 0, then dA = 0 and dA = 0; there will be no impact

of permit trading on output or abatement and so the usuals in the absence of transaction
dx
da
cost still hold. If Ttt > 0, then dA > 0 and dA < 0; if transaction cost is an increasing

function of volume of permit trade, then as the government increases the initial allocation
of permits the level of emission control will be reduced (output increases and abatement
decreases). Thus, the post-trading equilibrium outcome deviate from the cost-effective equidx
da
librium. Finally, if Ttt < 0, then dA < 0 and dA > 0; if transaction cost is an decreasing

function of volume of permit trade, then output will decrease but abatement will increase as
108

permit allocation increases. This also makes the post-trading equilibrium outcome depart
from the desired cost-effective equilibrium. Note that these properties described above hold
regardless of whether a firm is a seller or buyer of permits.

4.3

Data

This empirical study covers seven out of nine heavy-polluting EU ETS sectors from 2005 to
2007 (phase I) including combustion installations, mineral oil refinery, coke oven, production
of iron and steel, production of cement and lime, manufacture of glass, paper pulp and board.
Metal ore and manufacture of ceramic are under the regulation but excluded because of the
data unavailability. Twenty four out of EU27 member states participated in the first phase
of the EU ETS. For emissions trading, one EU Allowance Unit (EUA) for one ton of CO2 is
the official permit of the EU ETS.

Variables used in this study consist of two groups: production-related and regulationrelated. Production-related data imply input variables or variables that may affect the
production process, whereas regulated-related variables imply emissions and emissions market related variables. Sector-level data related to production are obtained from EUROSTAT,
which provides detailed statistics and statistical information on the EU and EU-candidate
countries. Capital, the number of R&D workers and the number of total employees are included as input variables. Other historical sectoral data (2001 to 2003) are also obtained from
EUROSTAT including variables measuring profits (i.e. gross operating surplus), production
level, and fixed cost variables (i.e. gross investment in tangible goods).3
3 There

is a concern about potential discrepancy between industrial classifications.
Production-related data are established based on standard ISIC or NACE industrial classifications, emissions-related reported in the UNFCCC are based on the Common Reporting
109

Regulation-related data are emissions, permit allocations and permit prices. The EU’s
Community Independent Transaction Log (CITL) keeps data for individual installation allocations and actual (verified) emissions after 2005. Before 2005, historical emissions data
are rare at the installation and firm level. The only country that keeps installation level
emissions data before 2005 is the UK At the sector level the United Nations Framework
Convention on Climate Change (UNFCCC) has kept emissions data by country, sector and
pollutant since 1990. Energy prices such as natural gas prices and electricity prices are also
included as controls in the analysis. Permit prices of EUA can be obtained from the permit
exchange markets such as European Climate Exchange (ECX) or Bluenext. Permit prices
include spot prices, futures prices and option prices. All the prices are reported daily.
Another available data layer of this paper is firm-level. Most of installations belong to firms.
Carbon Market Data database provides the link between installations and firms (i.e. ownership information of each installation).4 The database also indicates how many installations
each firm operates, how many sectoral activities a firm is involved in, and whether a firm is
Framework (CRF) as the Intergovernmental Panel on Climate Change (IPCC) directs, and
EU ETS uses its own classification. The EU ETS classification is close to the recommendation of the Intergovernmental Panel on Climate Change(IPCC) (See IPCC (1996) and
Intergovernmental Panel on Climate Change (2006) for detailed explanations). However,
the discrepancy between the EU ETS and the CRF of the IPCC exists. According to European Environmental Agency (2007) the sum of emissions in the GHG inventory from the
relevant CRF categories is always higher than the verified emission from the EU ETS because inventory includes all plants and does not use any threshold criteria for the inclusion
of installations as in the EU ETS. The calculated share of the ETS total in the CRF ranges
from 75% to 96% and the average share of the EU ETS total in the CRF total emissions
is 85.4% for EU23. Furthermore, even though the Common Reporting Framework (CRF)
of the IPCC directs that emissions should be reported according to the ISIC classification
whenever possible, matching with the ISIC is somewhat limited. European Environmental
Agency (2007) suggests classification matches between the EU ETS sectors and the Common Reporting Framework. This paper follows the suggestion of European Environmental
Agency (2007).
4 It is a private data provider. Visit http://www.carbonmarketdata.com/ for more information.

110

multinational. This study encodes these indications as dummy variables.

4.4
4.4.1

Identification Strategy
Models

A theoretical model in section 4.2 shows how the equilibrium output and abatement can vary
depending on the changes of permit allocation. As explained in the introduction section,
other efficiency-hurting factors are irrelevant in the EU ETS; transaction cost is the most
relevant culprit that can hurt the cost-effectiveness of the EU emissions trading system.
The equilibrium output and abatement are a function of output price, permit price and
possibly initial permit allocation in principle. Then two reduced form regression models are
suggested:

P RODict = α0 + α1 ALLOCict + α2 τt + α3 θic + α3 τt θic + γΦict + ηict

EM ITict = β0 + β1 ALLOCict + β2 τt + δΦict + ict ,

(4.10)

(4.11)

where subscripts i, c, and t denote sector, country, and time, respectively; P RODict and
EM ITict as dependent variables denote output and emissions of sector i in country c at
time t; ALLOCict is the number of permits allocated; τt is the permit price faced by all
participants at time t and θic is emissions intensity of sector i in country c; Φict denotes
a vector of other input-related variables such as capital, labor or R&D expenditures and
energy price variables such as natural gas prices and electricity prices; γ and δ are vectors
of coefficients to be estimated.
Since the interest of this paper lies on the independence of equilibrium decisions from the
111

initial permit allocation, the null hypothesis to be tested is that the allocation coefficients
are zero (α1 = 0 and β1 = 0).
In terms of the levels of data usage, the empirical analysis can be categorized into two
groups: sector and installation. Sector level data are mostly available in all twenty four EU
ETS participating countries. On the other hand, installation level data are available only in
the UK
Equations (4.10) and (4.11) are almost identical because emissions are understood as one
of by-products of production processes. However, there is a distinction between them. In
equation (4.11), emissions intensity (θ) is not included because verified emissions data of the
EU ETS (v = θx − a) readily have emissions intensity in it. Hence, equation (4.11) does not
include the interaction term between permit price and emissions intensity. In addition, to
control for the unobserved specific factors of each industrial sector or installation in a country,
historical emission data are included. Furthermore, each regression also includes fixed effect
dummies of countries, industrial sectors and firms. However, since permit price plays a role
as year dummy, year-specific dummies after the EU ETS are not included. For installation
level analysis, other firm-character dummies are used; multi-installation, multi-nation, and
multi-sector dummies.

4.4.2

The Control for Endogeneity of Permit Allocation

The first chapter discusses the reasons the initial permit allocation may matter when the
grandfathering permit allocation method is adopted. Simply put, the potential welfare loss
from the relocation of firms, including lost profits of firms, leads governments to allocate
permit strategically so as to minimize the overall welfare loss. In other words, governments
intend to compensate the potential profit loss of firms caused by emissions regulation and
keep those firms under the regulation by allocating more permits if those firms have played

112

important roles in terms of the welfare of the economy. The first chapter argues that the
three main components affecting the permit allocation decision are the level of emissions,
the profitability and mobility of firms. The level of emissions is the principal basis of the
initial permit allocation. We may understand the process of permit allocation in the following way. The allocation decision authorities (i.e. governments) first set an allocation
rule based on the level of emissions. Then, they add or subtract permits based on other
considerations. The first chapter argues that the profitability and mobility are the two most
important considerations. Since the aggregate profits of firms are part of social welfare, the
relocation of high profit firms to other countries can hurt the economy. However, not all
firms are apt to move. The mobility of firms vary across industries and it depends on various
factors, including transport cost of products, market accessibility, and the share of fixed
costs. Thus, governments are less likely to distribute permits to immobile firms and will
rather give permits to mobile firms in order to minimize the social costs from the emissions
regulation. After all, the initial permit allocation is more of a strategic act of a government
than a random process.
Even though predetermined permit allocation cannot be a culprit that hurts the permit
market efficiency, the historical-emission-based permit allocation can lead the empirical results to show a false causal relationship between the initial permit allocation and post-trading
equilibrium decisions. Such endogeneity problem rises because the initial permit allocation
reflecting historical emissions before the emissions regulation is potentially correlated with
post-trading equilibrium emissions or output. Without controlling for the endogeneity problem, the empirical analysis may show a strong causal relationship between the initial permit
allocation and the post-trading equilibrium decisions, which is probably a biased result.
In order to overcome this problem, two stage least squares estimation is adopted. The
empirical specification in the first chapter is included in the first stage regression. The historical emissions are used in both first and second stages. The instrumental variables are
113

the historical profits, historical fixed cost, net imports and number of employees; these are
not included in the second stage regression. The reasons that the instruments are considered
valid are following. Basically, all the historical data are the average value from 1999 to 2002
and the EU ETS has started 2005. Then, the question is how exogenous those historical
variables can be in the future decisions. Fixed cost is argued to be the most independent
of the optimal decisions in the future. The economic theory tells us that the decisions of
production and emissions are based on the marginal revenue and the marginal cost. Fixed
cost is not part of the optimal decision process. Hence, it is not convincing that historical
fixed cost has a lagged impact on future equilibrium decisions. The correlation between the
historical profits and the optimal decisions in the future is not very convincing either. The
level of profits depends not only on the decisions of firms or industries but on the economic
environment. So, can we answer for sure to the question whether the profits in a certain year
will a reliable indicator for the next year’s profits or level of production? The same reasoning
can apply to two other instruments, net imports and the number of employees. The following
reduced-form regression model is the empirical specification for the initial permit allocation
at sector level introduced in Chapter 1.

ALLOCic = λ0 + λ1 HISTic + λ2 N IMic + λ3 P ROF ITic + λ4 F IXEDic + λ5 EM Pic + ξic ,
(4.12)
where the subscripts i and c denote industrial sector and country; HISTic denotes average
historical emissions; N IMic denotes the level of net imports and proxies for the competitiveness; P ROF ITic denotes the level of profits and represents the contribution of sector i
to country c; F IXEDic denotes fixed costs and represents the mobility of an sector; and
EM Pic denotes the number of employees and is included as an proxy for political pressure
from sector i; and finally ξic is the error term.

114

4.5

Results

Table 4.1 provides the results of baseline estimations that check the simple correlations between permit allocations and equilibrium decisions of production and emissions in the first
phase (2005 to 2007). Dependent variables in column (1) and (2) are production and emissions at sector level, whereas dependent variable in column (3) is emissions at installation
level (UK only). Country, sector and year dummies are included in the sector level. In
the installation level, since there is no other country but the United Kingdom, the countryspecific dummy is not included. On the other hand, firm level information is only available
in the installation level regression using the UK data.
All outcomes exhibit clear positive correlations, which are also consistent with the graphs
provided in Figures C.1a through C.3a. In column (1), an additional increment of permit
allocation leads production to increase by 396 in value. In the emission-allocation relations
in column (2) one additional permit allocation implies 1.015 metric tons increase in emissions
(MtCO2). Similarly at the installation level in column (3) an additional permit leads to increase emissions by 1.291 metric tons more of emissions (MtCO2). However, these outcomes
showing high correlations between permit allocations and equilibrium decisions may be spurious. Thus, later tables show comparable results with more covariates and endogeneity
control.

Table 4.2 provides the results on the independence property between permit allocation
and production at the sector level with proper controls. Column (1) is with no endogeneity control, while column (2) uses two stage least squares to control for the endogeneity of
permit allocations. The dependent variable is the level of production, measured in billions

115

Table 4.1: Baseline regressions in phase I
(1)
production-sector
permit allocation 0.396***
(0.0551)
constant
6.779***
(1.605)
country dummy
yes
sector dummy
yes
year dummy
yes
firm dummy
obs.
292
R2
0.839
Robust standard errors in parentheses.
* p < 0.05, ** p < 0.01, *** p < 0.001

(2)
emission-sector
1.015***
(0.0193)
-0.154
(16.16)
yes
yes
yes
382
0.986

(3)
emission-installation
1.291***
(0.045)
-0.056
(0.191)
yes
yes
yes
2817
0.945

of euros. The overall outcomes show that once the endogeneity is controlled for, there is
no statistically significant evidence that initial permit allocations affect the impact on the
output decisions.
The first explanatory variable in Table 4.2 is permit allocation at the sector level (in
million MtCO2), whose impact on production is of our primary interest. Column (1) includes relevant production-related and regulation-related variables. The first and primary
variable, permit allocation, has a positive coefficient with 1% statistically significance. The
magnitude of the coefficient is a little larger than that in the baseline regression. According
to the result, one additional permit allocation leads production to increase by e452 in value.
In other words, it is hard to say that there is no influence of permit allocation on output
decision. However, once the endogeneity of permit allocation is controlled for, the coefficient
is still positive, but the statistical significance of the coefficient vanishes. This result implies
that the independence property cannot be rejected in the production decisions in the first
phase of the EU ETS and hence the post-trading equilibrium does not deviate significantly
far from the desired cost-effective equilibrium regardless of the initial permit allocations.
116

Other explanatory variables other than permit allocations show consistent results with
each other in both columns. First, capital plays an important role in the productions in the
EU ETS sectors. In columns (1) and (2), the coefficients are 0.00401 and 0.00406, respectively. Both of them are statistically significant at 1% level. An additional euro in capital
leads production to rise by about e4. Unlike capital, other input variables such as labor
and R&D employment do not exhibit significant impacts on emissions. The coefficients of
emission-related variables, natural gas price, electricity price, emissions intensity or permit
price do not show statistical significances either. To capture the individual idiosyncrasy of
each industrial sector in each country historical emissions from 1990 to 2000 are included.
But most of historical emissions do not show significant influence on productions and the
signs of these coefficients do not have in common.

Table 4.3 provides the estimation results about the impact of permit allocation on emissions in the sector level. The verified emissions of the EU ETS, measured in metric tons of
CO2 (MtCO2), are the officially reported emissions level to the European Commission. The
potential endogeneity is not controlled for in column (1), whereas column (2) uses two stage
least squares for the endogeneity control. The first stage regression is from 4.12.
The regression outcomes statistically reject that there is no influence of the initial permit
allocation on the level of emissions regardless of the control for the potential endogeneity.
In column (1) with no endogeneity control, the coefficient of permit allocation is 0.861; one
additional permit leads to increase emissions by 0.861 MtCO2 in the first phase. In column
(2) with the endogeneity control through two stage least squares method, the result still
exhibits a significant impact of initial permit allocation on the level of emissions. The coefficient of permit allocation is around 0.794 at 1% statistical significance; one extra permit
leads to increase emissions by 0.794 MtCO2. Based on the theoretical model built in this
paper, the result tells that the marginal transaction cost is an increasing function of the
117

Table 4.2: Independence property with respect to production at sector level
(1)

(2)

OLS

s.e.

2SLS

s.e.

permit allocation
0.452***
(0.171)
0.288
(0.253)
capital
0.00401***
(0.000949)
0.00406***
(0.00094)
labor (mil. workers)
-0.00433
(0.0731)
-0.00576
(0.0759)
R&D workers
0.00403
(0.00301)
0.00393
(0.00293)
natural gas price
0.178
(0.819)
0.242
(0.846)
electricity price
9.188
(93.78)
2.557
(92.85)
emissions intensity (θ) -0.000267
(0.00290)
-0.000664
(0.00309)
permit price (τ )
-0.0619
(0.531)
-0.0269
(0.544)
θ×τ
0.0000459
(0.000151)
0.0000591
(0.00016)
hist. emissions 1990
-2.046
(1.321)
-2.199
(1.445)
hist. emissions 1991
1.088
(1.001)
1.423
(1.223)
hist. emissions 1992
0.574
(1.278)
0.778
(1.377)
hist. emissions 1993
1.981
(1.331)
1.726
(1.179)
hist. emissions 1994
-2.032
(1.989)
-2.358
(2.221)
hist. emissions 1995
-2.521
(1.598)
-1.601
(1.941)
hist. emissions 1996
-2.060
(1.262)
-2.130
(1.307)
hist. emissions 1997
5.222**
(2.311)
4.676**
(1.919)
hist. emissions 1998
2.910*
(1.688)
2.809*
(1.642)
hist. emissions 1999
-5.813**
(2.427)
-4.701***
(1.693)
hist. emissions 2000
2.227
(1.584)
1.279
(1.610)
constant
-1.661
(16.16)
-2.045
(16.47)
country dummy
yes
yes
yes
yes
sector dummy
yes
yes
yes
yes
obs.
153
153
R2
0.897
Robust standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001

118

Table 4.3: Independence property with respect to emissions
(1)
OLS

(2)
s.e.

2SLS

s.e.

permit allocation
0.861***
(0.0232)
0.794***
(0.0445)
capital
0.000583***
(0.000168)
0.000588***
(0.000170)
labor (mil. workers) -0.0134
(0.00815)
-0.0173**
(0.00735)
R&D workers
-0.000603
(0.000478)
-0.000658
(0.000471)
natural gas price
0.0204
(0.212)
0.0632
(0.221)
electricity price
36.41
(36.25)
31.99
(36.40)
permit price
0.0550
(0.111)
0.0740
(0.114)
hist. emission 1990 -1.116***
(0.147)
-1.128***
(0.153)
hist. emission 1991 0.800***
(0.168)
0.850***
(0.177)
hist. emission 1992 1.638***
(0.194)
1.635***
(0.213)
hist. emission 1993 -1.500***
(0.183)
-1.503***
(0.188)
hist. emission 1994 -1.143***
(0.269)
-0.955***
(0.292)
hist. emission 1995 2.588***
(0.332)
2.441***
(0.310)
hist. emission 1996 0.530***
(0.119)
0.539***
(0.126)
hist. emission 1997 -1.259***
(0.225)
-1.373***
(0.248)
hist. emission 1998 -1.303***
(0.227)
-1.387***
(0.236)
hist. emission 1999 0.769***
(0.218)
0.899***
(0.217)
hist. emission 2000 0.145
(0.194)
0.186
(0.185)
constant
-3.220
(4.057)
-3.438
(4.076)
country dummy
yes
yes
yes
yes
sector dummy
yes
yes
yes
yes
obs.
161
159
R2
0.999
Robust standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001
volume of permit trading. Furthermore, the results indicate that the post-trading equilibrium departs further from the desired cost-effective equilibrium as the the volume of permit
trading increases. In other words, the market efficiency that makes emissions trading more
attractive than emissions tax is not reached in the first phase of the EU ETS. This outcome
in emissions contradicts the outcome in production. This contradiction will be discussed
more later with installation-level estimation results.
The coefficients of capital in columns (1) and (2) are positive around 0.00058 with 1% sta-

119

tistical significance. An additional euro in capital leads emission to rise by 0.00058 MtCO2.
The number of employees has negative coefficients in columns (1) and (2). The sectors with
large employment tend to emit less, all else equal. R&D workers indicating human capital
also has negative coefficients; however, they are not statistically significant.
Price variables such as permit price, natural gas price and electricity price do not pose any
statistical significances. Historical emissions from 1990 to 2000 to capture sectoral unobserved factors in different countries show strong statistical significances. However, it is hard
to find general rules to draw because the signs of coefficients are not consistent across years.
The impact of permit price on emissions is not statistically significant in the first phase of the
EU ETS. Each specification includes country and sector dummies. However, since annual
price data play a role of year dummies, year dummies are not included.
To investigate the contradictory empirical results between production and emissions in
the sector level, estimations in the installation level are also performed. The benefits of
installation level analysis are the larger sample size and the use of firm-character dummies.
On the other hand, the defect is that production-related variables are not available at installation level. Table 4.4 provides the results for the impact of permit allocations at installation
level in the UK. The dependent variable is reported verified emissions measure in million
metric tons of CO2 (MtCO2). Columns (1) and (2) use phase I (2005 to 2007) data only.
On the other hand, data in columns (3) and (4) are extended to the first two years (2008
and 2009) of phase II.

First two columns in Table 4.4 show the estimation results within the phase I. Column
(1) do not control for the endogeneity, whereas column (2) controls for endogeneity with the
two stage least squares (2SLS) method. The coefficients of permit allocations are positive
with statistical significances; in the first phase, the initial allocation has a significant impact

120

Table 4.4: Independence property emissions at installation UK
(1)
(2)
(3)
(4)
OLS - phase I 2SLS - phase I OLS - ext. years 2SLS - ext. years
permit allocations 0.331***
0.393*
0.432***
0.343
(0.117)
(0.230)
(0.115)
(0.565)
permit price
-0.00203
-0.00239
0.00532**
0.00632***
(0.00571)
(0.00576)
(0.00210)
(0.00219)
natural gas price
0.00327
0.00286
0.00618*
0.00700
(0.00482)
(0.00508)
(0.00329)
(0.00434)
emission 1998
0.320***
0.316***
0.237***
0.241***
(0.0873)
(0.0901)
(0.0876)
(0.0903)
emission 1999
-0.316**
-0.333**
-0.090
-0.0701
(0.136)
(0.146)
(0.116)
(0.177)
emission 2000
-0.0967
-0.0928
-0.158
-0.163
(0.0946)
(0.0932)
(0.0993)
(0.0992)
emission 2001
0.471***
0.467***
0.292**
0.306**
(0.133)
(0.132)
(0.133)
(0.140)
emission 2002
-0.248
-0.258
-0.171
-0.161
(0.170)
(0.173)
(0.166)
(0.189)
emission 2003
0.595***
0.577***
0.512***
0.532***
(0.158)
(0.171)
(0.141)
(0.177)
multi-installation 0.0866
-1.099
-0.0258
-0.601
(0.119)
(1.109)
(0.0465)
(1.684)
multi-nation
-0.554**
-0.0706
-0.541***
-0.531**
(0.219)
(0.0522)
(0.199)
(0.217)
multi-sector
0.520**
-0.703
1.345***
1.384***
(0.225)
(0.440)
(0.418)
(0.465)
constant
-0.248
1.516
-0.190***
0.349
(0.173)
(1.191)
(0.0650)
(1.624)
sector dummy
yes
yes
yes
yes
firm dummy
yes
yes
yes
yes
obs.
1594
1594
2621
2621
2
R
0.982
0.968
Robust standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01

121

on emissions decision. The magnitudes of the coefficients are around 0.39 with 2SLS; in other
words, an additional permit allocation leads emissions in phase I to increase by 0.39 metric
ton of CO2. This result changes if the empirical analysis includes the years of phase II as
shown in columns (3) and (4). Results provided in column (3) have no endogeneity control
and show a significant influence of permit allocation on emissions decisions as in column (1)
with phase I data only. However, with endogeneity control results provided in column (4) do
not statistically reject that there is no influence of permit allocation on emissions decisions
in extended years (2005 to 2009).
Besides permit allocations, historical emissions as explanatory variables exhibit strong correlations with emissions decisions. Furthermore, multi-nation dummy have negative coefficients across specifications with relatively high statistical significances. Some may argue that
multi-national firms share more efficient technologies than local firms so that they tend to
emit less. Columns (1), (2), and (4) tell that if installations belong to multi-national firms,
then they tend to emit less by 0.5 million MtCO2 compared to local firms.

Table 4.4 seems to show that the permit market of the EU ETS is in the process of
settling down over time, which seems understandable because the implementation of such
a huge multinational market has just begun. To see the annual changes of independence
property Table 4.5 provides the annual estimation results at the installation level (UK only)
with no endogeneity control. Since these are annual cross-sectional estimations, annual price
data such as permit price, natural gas price or electricity price are not included. Despite
insufficient controls, the outcomes exhibit an interesting evidence that the stability of its
own market or other linked financial systems play an important role in the efficiency of the
permit market. Columns (1) and (2) show the results in 2005 and 2006 that initial permit
allocations have a statistically significant influence on the levels of emissions. However, the
122

Table 4.5: Annual independence property at installation level (UK)
(1)
(2)
emit2005
emit2006
permit allocation 0.287**
0.405*
(0.128)
(0.244)
emissions 1998
0.129
0.465***
(0.126)
(0.153)
emissions 1999
-0.217
-0.716***
(0.196)
(0.177)
emissions 2000
-0.0886
0.108
(0.133)
(0.109)
emissions 2001
0.542***
0.640***
(0.190)
(0.187)
emissions 2002
-0.158
-0.240
(0.240)
(0.207)
emissions 2003
0.543***
0.448**
(0.179)
(0.223)
multi-installation 0.218
-1.269
(0.133)
(1.502)
multi-nation
-1.044
-0.101
(1.222)
(0.0718)
multi-sector
0.442***
1.190
(0.120)
(1.449)
constant
-0.285**
-0.374**
(0.140)
(0.179)
sector dummy
yes
yes
firm dummy
yes
yes
obs.
539
533
2
R
0.990
0.991
Robust standard errors in parentheses
* p < 0.1, ** p < 0.05, *** p < 0.01

123

(3)
emit2007
0.181
(0.242)
0.373*
(0.213)
0.0233
(0.284)
-0.318
(0.208)
0.235
(0.222)
-0.326
(0.332)
0.829***
(0.312)
0.264
(0.182)
-0.925
(0.807)
0.765***
(0.228)
-0.270
(0.195)
yes
yes
522
0.980

(4)
emit2008
0.530
(0.337)
0.308
(0.284)
0.0885
(0.387)
-0.381
(0.251)
-0.0273
(0.359)
-0.0534
(0.412)
0.604**
(0.272)
-4.530**
(1.901)
-0.310
(0.190)
1.049
(0.664)
4.401**
(1.888)
yes
yes
517
0.964

(5)
emit2009
0.762***
(0.273)
-0.104
(0.143)
0.437**
(0.179)
-0.136
(0.129)
-0.0315
(0.233)
-0.0160
(0.225)
0.202
(0.215)
0.0477
(0.185)
-0.670***
(0.158)
1.871***
(0.312)
-0.170
(0.221)
yes
yes
510
0.981

Figure 4.1: The volatility of permit price (futures contracts)

124

result provided in column (3) shows that the influence of permit allocation is not significant
in 2007. In the first year of the second phase the similar insignificant result shows up again
in column (4). It seems that the permit market is in the process of settling down and
getting more efficient over time in terms of ‘independence property’ of the permit allocation.
However, in column (5) the influence of permit allocation reappears. The reverse trend
requires further research. It may, however, reflect the abrupt market instability and increased
uncertainty caused the financial crisis in late 2008. Figure 4.1 depicts the daily time-series
data of permit price (futures) from May 2005 to May 2009. Dashed line lasting from mid2005 to 2007 depicts the price of EUA futures contract delivering in December 2007; solid
line depicts the price of EUA futures contract delivering in December 2009; and dotted line
starting from April 2008 depicts the price of EUA futures contract delivering in December
2013. There are two major price plunges through the entire time-series; right after April
2006 and after September 2008. First price price plunge happened when the actual overall
permit holding position of the entire EU ETS was announced for the first time and turned
out to be over-allocations. Insufficient information about the permit market in 2005 and
early 2006 may have caused the transaction costs and uncertainty to rise and hence lead
regulated sectors to stay rather close to the neutral permit holding position. The second
price plunge happened after the U.S. financial crisis hit the global financial system hard
in 2008. In most of 2009, the financial systems were extremely unstable, which may lead
industrial sectors under the EU ETS to stay neutral in permit holding position.
Before the financial crisis, the permit market seemed to have been stabilized and gotten more
efficient over time. Since the EU ETS is a newly introduced multi-national, multi-sectoral
artificial market in the intention of curbing CO2 emissions into the air, it is natural to think
that it will take a while for the market to adjust itself as a well-functioning parallel market
with other existing economic systems.

125

4.6

Conclusion

This paper studies the independence property of the EU emissions trading market as a test
for market efficiency; the initial permit allocation should not affect the equilibrium decisions
of output or emissions. The market efficiency is one of the primary factors to assure that
the targeted emissions reductions of heterogeneous sectors are accomplished at the minimum
cost through the flexible price system. Since the EU ETS (Emissions Trading System) was
first implemented in 2005, there have been speculations about imperfect market operations,
at least, in phase I (2005 to 2007). One of them is whether permit market performance is
under an influence of initial permit allocations. The purpose of this study is to test whether
the initial permit allocation influences equilibrium outcomes. First, in a theoretical model
transaction cost is introduced in the profit function of firms following Stavins (1995). The
model predicts that unless marginal transaction cost is constant the initial permit allocation
has an impact on equilibrium decisions of output and abatement. Then, using data from
the EU ETS phase I, the prediction of independence property is tested empirically at the
sector level and the installation level. Since permits are allocated strategically in order to
minimize the welfare loss from the emissions regulation, the potential endogeneity problem
of initial permit allocation is considered to exist and so two stage least squares method
is used to control for endogeneity. The empirical specification derived from a theoretical
model in the first chapter is used in the first stage regression. Among five factors which can
influence the initial permit allocation, historical emissions; historical profits; historical fixed
cost; historical net import; and historical employment. Four factors other than historical
emissions are argued to be independent of the optimal decisions during the first phase of
the EU ETS and hence used as instrumental variables. In the case of production at the
sector level, without the endogeneity control the impact of the initial permit allocation
on the level of production is statistically significant in the first phase. However, once the

126

endogeneity of the initial permit allocation is controlled for, the impact of permit allocation
on productions vanishes. On the other hand, in the case of emission at sector level, which
readily contains abatement information in it, the impact of permit allocation on equilibrium
decisions in the first phase seems evident regardless of controlling for endogeneity. Empirical
results of emissions at installation level in the UK are consistent with those at sector level
in the first phase; regardless of endogeneity control the initial permit allocations have a
significant influence on emissions in the first phase. However, the statistical significance of
the coefficient vanishes after the endogeneity control as two years (2008 and 2009) of the
second phase (2008 to 2012) are included in the analysis. This paper interprets the results
that the efficiency of the permit market for the EU ETS has gained across years. The EU
ETS is by far the world’s largest multi-national multi-sectoral market-based regulation and
so the adjusting period seems unavoidable. Therefore, the influence of permit allocation on
equilibrium decisions may not be so unreasonable especially for the first phase.
There remains a further research question; what the relationship between permit market
and existing financial system is. This research also provides a suggestive evidence that the
financial crisis in late 2008 may affect the permit market efficiency of the EU ETS. Since
the market efficiency is one of the main factors that can lead to the success of the emissions
trading system and the inter-market influence has not been studied sufficiently yet, it is
important to understand the permit market in the context of integrated financial market.

127

APPENDIX

128

Appendix C

C.1

Summary Statistics 1
Table C.1: Summary statistics at sector level
Variable

Mean

Std.
Dev.
production (ebil.)
10.895
23.62
verified emission (mil. MtCO2)
15.228
43.702
permit allocation (mil. MtCO2)
15.407
42.305
capital (emil.)
15850.694 176013.96
labor (mil. workers)
21.153
32.59
R&D workers (1000 workers)
174
331.642
permit price (e/MtCO2)
19.823
1.456
natural gas price (e/giga Joules)
6.903
2.013
electricity price (e/kWh)
0.071
0.017
emissions intensity (MtCO2/rGDP 6.478
10.318
in emil.)
emissions 1990 (mil. MtCO2)
13.196
38.916
emissions 1991 (mil. MtCO2)
13.112
38.469
emissions 1992 (mil. MtCO2)
12.393
36.659
emissions 1993 (mil. MtCO2)
11.88
34.869
emissions 1994 (mil. MtCO2)
12.078
34.611
emissions 1995 (mil. MtCO2)
11.935
33.731
emissions 1996 (mil. MtCO2)
12.092
34.565
emissions 1997 (mil. MtCO2)
11.827
33.098
emissions 1998 (mil. MtCO2)
11.938
33.662
emissions 1999 (mil. MtCO2)
11.726
32.87
emissions 2000 (mil. MtCO2)
12.002
34.175

129

Min.

Max.

obs.

0.013
0
0.006
0.1
0
0
18.179
2.752
0.041
0.0083

226.171
378.428
378.901
2298293.5
207.654
2600
21.716
12.15
0.112
90.511

309
385
387
309
328
244
483
399
483
291

0.003
0.002
0
0
0.001
0
0.001
0
0
0.001
0.001

335.781
328.394
315.838
307.335
307.221
302.103
314.099
299.883
305.592
296.235
309.536

435
432
438
438
438
444
441
444
444
441
441

C.2

Summary Statistics 2
Table C.2: Summary statistics at installation level (UK)
Variable

Mean

verified emissions (mil. MtCO2)
permit allocation (mil. MtCO2)
permit price (e/MtCO2)
natural gas price (e/giga Joules)
electricity price (e/kWh)
emissions 1998 (e/MtCO2)
emissions 1999 (e/MtCO2)
emissions 2000 (e/MtCO2)
emissions 2001 (e/MtCO2)
emissions 2002 (e/MtCO2)
emissions 2003 (e/MtCO2)
multi-installation
multi-nation
multi-sector

0.278
0.249
19.102
8.051
0.087
0.22
0.22
0.232
0.241
0.242
0.261
0.955
0.422
0.361

Std.
Dev.
1.22
0.901
3.52
1.596
0.017
1.055
0.989
1.035
1.056
1.007
1.172
0.208
0.494
0.48

130

Min.

Max.

obs.

0
0
12.936
5.811
0.057
0
0
0
0
0
0
0
0
0

22.765
14.554
23.104
10.552
0.108
21.105
19.58
19.569
18.875
16.417
21.826
1
1
1

4494
4259
6690
6690
6690
5310
5310
5310
5310
5310
5310
3740
3740
3740

C.3

Seeming Correlation between the Initial Allocation and Equilibrium Decisions

Figure C.1: Independent of or dependent on allocation?: Initial allocation vs. Production
Source: the UNFCCC and EUROSTAT
Note: Horizontal axis and vertical axis indicates the initial permit allocation (million
MtCO2) and production (billion e) at sector level in the first phase of the EU ETS, respectively. The straight line indicates the fitted values.

131

Figure C.2: Independent of or dependent on allocation?: Initial allocation vs. Emissions
(sector)
Source: the UNFCCC and EUROSTAT
Note: Horizontal axis and vertical axis indicates the initial permit allocation (million
MtCO2) and emissions (million MtCO2) at sector level in the first phase of the EU ETS,
respectively. The straight line indicates the fitted values.

132

Figure C.3: Independent of or dependent on allocation?: Initial allocation vs. Emissions
(inst.)
Source: the UNFCCC and EUROSTAT
Note: Horizontal axis and vertical axis indicates the initial permit allocation (million
MtCO2) and emissions (million MtCO2) at installation level in the first phase of the EU
ETS, respectively. The straight line indicates the fitted values.

133

Chapter 5
EPILOGUE

In the beginning of the dissertation, the theme question asked is the economic impact of the
emissions trading system as a climate change protection policy in ever competitive global
market. In addition, we would like to learn lessons from the implementation of the EU
emissions trading system. I believe that since conditions suggested in theoretical models
are not guaranteed to be satisfied, the success of a policy depends upon how effectively the
problems and obstacles from reality are recognized and overcome. In such a sense, overall
the EU emissions trading system is believed to have managed to be on the right track so far.
In the first chapter, the initial permit allocation when keeping firms under the regulation
is a potential challenge to the success of the EU emissions trading system. Grandfathering
permit allocation is theoretically justified as a tool to minimize the welfare loss from the relocation of industries to pollution havens. Another important question answered in the paper
is whether the European governments participating in the EU ETS follow economic theory
and overcome the political voices that urge to allocate more permits to less cost-competitive
industries. Protecting relatively less cost-competitive industries by distributing more permits
is politically popular, but theoretically unsound because permit allocation is irrelevant with
cost-competitiveness. Empirical analysis with data from the EU ETS provides supporting
evidence that the governments of European countries follow economic theory rather than
political rhetoric in general. It turns out that in the EU ETS permit allocation is focused
more on the welfare loss from the relocation of firms rather than the cost-competitiveness.
Since the relocation of firms causes not only the reduction in tax revenues and employment
but can crowd out all the efforts made through the emissions trading, weighing more on the
problems of the relocation of firms is crucial to make the emissions trading system viable

134

and sustainable.
Second chapter tries to answer a long-lasting question whether there is the possible unfavorable impact of the emissions trading on trade flows. The overall outcome of the chapter
is in line with recent research that adds weak yet existing evidence on the pollution haven
effect. The contribution of the second chapter is that a Heckscher-Ohlin type general equilibrium model incorporated with an emissions trading system suggests theoretically sound
variables to use in empirical analysis, permit price and emissions intensity, and that the
empirical analysis uses suggested variables with data from the EU ETS and shows some
evidence on pollution haven effect. Net imports of regulated industrial sectors from lowmiddle income trading partner groups have increased since the EU ETS started compared
to non-regulated sectors. Detecting significant economic impact of the emissions trading
on regulated industrial sectors does not imply or suggest the halt of the emissions trading:
the general equilibrium analysis and the corresponding empirical investigation provides a
relative impact between regulated and non-regulated sectors; and the unfavorable impact is
not happening with major trading partners and the magnitudes of the unfavorable impact
are rather small. Actually, not covering all dirty sectors at the same time may not be a
great idea in a sense that regulating relatively dirtier sectors may give a room for less dirty
sectors to expand. Therefore, the entire economy transits to a cleaner state at a manageable
speed without hurting the economy too much at once. I believe that if the government leads
industrial sectors to a steady transition, then it will face less oppositions and can make the
emissions trading system much more successful.
The core reason why emissions trading is preferable is the minimum cost property of emissions trading. The flexible price system guarantees to reach the emission reduction target at
the minimum cost under certain conditions. The cost-effectiveness is fairly important in a
sense that the emissions trading in Europe has been operated at the cost of industrial competitiveness. Therefore, if the cost-effectiveness of the emissions trading cannot be reached,
135

then the reasons to implement the emissions trading need to be reconsidered. Transaction
cost is the main factor that can hinder the cost-effectiveness of the emissions trading. If the
initial permit allocation affects the equilibrium decisions through transaction cost, then it is
a symptom that shows the permit market is not efficient enough and so the emissions trading
may not be cost-effective as desired. Chapter 3 empirically tests whether the initial permit
allocation has a significant influence on equilibrium decisions of output and emissions. In
the pilot phase (2005-2007) of the EU ETS, empirical results show that output decision is
already independent of the initial allocation but emissions decision seems under the influence of the initial permit allocation. However, the extended years including the part of the
second phase change the results to insignificant influence of the initial allocation. Chapter
3 interprets the overall outcome as an indication that the permit market is in the process
of settling down. Especially, insufficient information about the overall position status of
the entire permit allocation in the early years of the first phase seems to cause industries
to make emissions decisions around the number of permits allocated. In other words, the
market uncertainty generates the dependence of equilibrium decisions on the initial permit
allocation. However, since the information about allocations and emissions became public,
the fluctuation of permit price has become stabilized and the impact of the initial permit
allocation on emission decision has vanished.
Even though the urgency of preventing further climate change is agreed around the world,
taking serious actions together with countries at different developmental stages seems hard to
achieve. European countries made a meaningful first step with the emissions trading system.
It has not been easy to establish a stable multi-national multi-sectoral emissions reduction
system. It may be too early to say whether the emissions trading system in EU is successful
or not. However, most experiences the EU ETS has been through are not so departed from
what has been theoretically predicted and hence the EU ETS is believed to be on the right
track. Based on the firm ground the EU ETS has built during phase I and II, from the
136

third phase of the EU ETS permits will be allocated with an auctioning method instead of
grandfathering and the national emissions cap will be replaced by the EU-wide cap. These
significant revisions of the EU ETS will raise much more challenges than before. However, I
hope that successful transition to better and more sophisticated system will encourage other
countries to participate in the climate change problems.

137

BIBLIOGRAPHY

138

BIBLIOGRAPHY

Antweiler, Werner, Brian R. Copeland, and M. Scott Taylor, “Is free trade good for the
environment?,” American Economic Review, 2001, 91, 877–908.
Barrett, Scott, “Strategic Environmental Policy and International Trade,” Journal of Public
Economics, 1994, 54, 325–338.
Becker, R. A. and J. V. Henderson, “Effects of Air Quality Regulations on Polluting Industries,” Journal of Political Economy, 2000, 108, 379–421.
Brock, W. and M. S. Taylor, “Economic Growth and the Environment: a Review of Theory
and Empirics,” in S. Durlauf and P. Aghion, eds., The Handbook of Economic Growth,
North Holand, 2005.
Brunnermeier, Smita B. and Arik Levinson, “Examining the Evidence on Environmental
Regulations and Industry Location,” Journal of Environment & Development, March
2004, 13 (1), 6 – 41.
Coase, R, “The Problem of Social Cost,” Journal of Law and Economics, 1960, 3, 1–44.
Cole, Matthew A. and Robert J. R. Elliott, “Determining the tradeenvironment composition
effect: the role of capital, labor and environmental regulations,” Journal of Environmental Economics and Management, 2003, 46, 363 – 383.
Copeland, Brian R. and M. Scott Taylor, Trade and the Environment, Princeton University
Press, 2003.
Croker, T. D., The Structuring of Atmospheric Pollution Control Systems, Norton, 1966.
Dales, J. H., Pollution, Property, and Prices, University of Toronto Press, 1968.
Dixit, Avinash, “Comparative Statics for Oligopoly,” International Economic Review, 1986,
27, 107–122.
Douglas, Straford and Shuichiro Nishioka, “International Differences in Emissions Intensity
and Emissions Content of Global Trade,” 2010.
Ederington, J. and Jenny Minier, “Is environmental policy a secondary trade barrier? An
empirical analysis,” Canadian J, 2003, 36, 137 – 154.
Ederington, Josh, Arik Levinson, and Jenny Minier, “Footloose and Pollution-Free,” The
Review of Economics and Statistics, February 2005, 87 (1), 92 – 99.
European Environmental Agency, “Comparison Verified Emissions and GHG Inventories for
2005,” Technical Report, European Environmental Agency 2007.
139

Fischer, Carolyn, “Rebating Environmental Policy Revenues: Output-based Allocations and
Tradable Performance Standards,” Resources for the Future, 2001.
and Alan K. Fox, “Output-based Allocation of Emissions Permits for Mitigating Tax
and Trade Interactions,” Land Economics, 2007, 83, 575–599.
Fowlie, M. and Jeffrey M. Perloff, “Distributing Pollution Rights in Cap-and-Trade Programs: Are Outcomes Independent of Allocation?,” September 2008.
Fowlie, Meredith, “Updating the Allocation of Greenhouse Gas Emissions Permits in a Federal Cap-and-Trade Program,” June 2010.
Goulder, L. H., I. W. H. Parry, R. C. Williams III, and D. Burtraw, “The cost-effectiveness of
alternative instruments for environmental protection in a second-best setting,” Journal
of Pu, 1999, 72, 329–360.
Greenstone, Michael, “The impacts of Environmental Regulations on Industrial Activity:
Evidence from the 1970 and 1977 Clean Air Acts and the Census of Manufactures,”
Journal of Political Economy, 2002, 110, 1175 – 1219.
Grossman, G. and A. Kreuger, “Economic growth and the environment,” Quarterly Journal
of Economics, 1995, 110, 353 – 377.
Hahn, Robert W., “Market Power and Transferable Property Rights,” Quarterly Journal of
Economics, 1984, 99, 753–765.
and Robert N. Stavins, “The effect of allowance allocations on cap-and-trade system
performance,” FEEM Working Paper No. 80, 2010.
Hanemann, Michael, “Cap-and-Trade: a sufficient or necessary condition for emission reduction?,” Oxford Review of Eocnomic Policy, 2010, 26, 225–252.
Houser, T., R. Bradley, B. Childs, J. Werksman, and R. Heilmayr, “Leveling the Playing
Field: International Competition and U.S. Climate Policy Design,” Technical Report,
Petersib Institute for International Economics and World Resources Institute 2008.
Intergovernmental Panel on Climate Change, “IPCC Guidelines for National Greenhouse
Gas Inv,” Technical Report, Intergovernmental Panel on Climate Change 2006.
IPCC, “IPCC Guidelines for National Greenhouse Gas Inventories,” Technical Report, Intergovernmental Panel on Climate Change (IPCC) 1996.
Jaffe, Adam B., Steven R. Peterson, and Robert N. Stavins, “Environmental Regulation
and the Competitiveness of U.S. Manufacturing: What DOes the Evidence Tell Us?,”
Journal of Economics Literature, 1995, 33, 132 – 163.

140

Kamien, Morton I. and Nancy L. Schwartz, “Conjectural variations,” Canadian Journal of
Economics, 1983, 16, 191–211.
Levinson, Arik and Scott Taylor, “Unmasking the pollution haven effect,” International
Economic Review, 2008, 49, 223 – 254.
List, J. A., W. W. McHone, D. L. Millimet, and P. G. Fredriksson, “Effects of Environmental
Regulations on Manufacturing Plant Birth: Evidence from a Propensity Score Matching
Estimator,” Review of Economics and Statistics, 2003, 85, 944 – 952.
Markusen, James R., Edward R. Morey, and Nancy Olewiler, “Competition in regional environmental policies when plant locations are endogenous,” Journal of Public Economics,
1995, 56, 55–77.
Montero, Juan Pablo, “Marketable Pollution Permits with Uncertainty and Transaction
Costs,” Resource and Energy Economics, 1997, 20, 27–50.
Montgomery, David W., “Markets in Licenses and Efficient Pollution Control Programs,”
Journal of Economic Theory, 1972, 5, 395 – 418.
OECD, “Summary report of the workshop on environmental policies and industrial competitiveness,” Technical Report, Organization of Economic Cooperation and Development
1993.
European Union, “DIRECTIVE 2003 87 EC OF The European Parliament and of the Council,” Technical Report, European Parilament and the Council of the European Union
2003. establishing a scheme for greenhouse gas emission allowance trading within the
Community and amending Council Directive 96/61/EC.
Quiroga, Miguel, Thomas Sterner, and Martin Persson, “Have Countries with Lax Environmental Regulations a Comparative Advantage in Polluting Industries?,” RFF Discussion
Paper, 2007.
Romalis, John, “Factor proportions and the structure of commodity trade,” American Economic Review, 2004, 94, 67 – 97.
Stavins, Robert N., “Transaction Costs and Tradable Pemrits,” Journal of Envrionmental
Economics and Management, 1995, 29, 133–148.
Stokey, N., “Are there limits to growth?,” International Economic Review, 1998, 39, 1 – 31.
Ulph, A., “The choice of environmental policy instruments and strategic international trade,”
in R. Pethig, ed., Conflicts and Cooperation in Managing Environmental Resources,
Springer-Verlag, Berlin, 1992.
World Bank, “World Development Report 1992,” Technical Report, Oxford University Press
1992.

141