DISS. ETH NO. 19475
Voluntary Climate Change Initiatives in the U.S.: Analyzing Participation
in Space and Time
Dissertation submitted to
ETH ZÜRICH
for the degree of
DOCTOR OF SCIENCES
presented by
Lena Maria Schaffer
Dipl.-Verw.Wiss., Universität Konstanz
born
04.12.1979
citizen of Germany
accepted on the recommendation of
Prof. Thomas Bernauer, examiner
Prof. Jude Hays, co-examiner
Prof. Vally Koubi, co-examiner
2011
iii
Acknowledgements
This dissertation is the outcome of my work at the Center for Comparative and International
Studies (CIS) at ETH Zurich. Writing about a continuously evolving topic has proven to be a
challenging, but also rewarding experience. My thesis would not be as it is without the help
and support of advisors, colleagues and friends.
First and foremost, I would like to thank my main advisor Thomas Bernauer for his encouragement and support throughout the years. He has always given me great freedom to pursue this
project and has provided valuable input for the success of this dissertation. I am also thankful
to Vally Koubi, who supported me throughout the years and always pushed me to think harder.
Her professional and personal advice has been and continues to be invaluable and greatly appreciated. I also thank my advisors for the possibility to attend various summer schools and
conferences and the inspiring academic environment they provided.
Jude Hays contributed important methodological and substantive advice for this dissertation.
I further want to thank Jude for his exceptional hospitality during my three month stay with
the University of Illinois in Urbana-Champaign. A lot of the ideas that came out of our lunch
appointments during this time have found their way into the dissertation and have improved it
in various ways. Besides his academic input for this dissertation, he also helped me to finally
understand American Football.
I also want to thank Dana Fisher, who has hosted me as a visiting scholar at Columbia
University. I profited a lot from her substantive knowledge in subnational environmental politics
and the great experience she has with respect to doing qualitative research.
Participants of the EITM summer institute in Michigan and various other conferences gave
useful comments on previous versions of the dissertation papers, and I want to thank the discussants and participants for their comments and feedback.
Gabriele Spilker and Julian Wucherpfennig gave valuable input for my work and supported
me during different stages of this project. Tobias Hofmann and Thomas Malang have gone out
of their way in commenting and proof-reading this dissertation and I am indebted to them for
their help.
iv
The immense data collection effort for this dissertation would not have been possible without
the knowledgable assistance of Stefan Schtz, Gwen Tiernan and Krzysztof Wojtaniec, whom I
thank for their hard work.
A crucial point for my research project has been the funding by the Competence Center
Environment and Sustainability (CCES). The work for this dissertation was done within the
funded project ClimPol, whose financial and academic support is greatly acknowledged.
Many thanks go to my friends, Hanja Blendin, Christian Canstein, Benjamin Kreibich, Daniel
Rittlinger, Gabriele Spilker and Susanna Steinbach, who have endured my complaints and selfdoubts and provided me with a healthy work-life balance during the dissertation years.
Finally, I am more than thankful for the support and constant motivation of my husband
Thomas Malang. For continuous support and so much more I am also indebted to my parents,
Ute and Andreas Schaffer, and my brother Philipp.
v
Abstract
Global climate change is one of the biggest challenges mankind is facing in the 21st century.
Prominent attempts to deal with global climate change and to mitigate its consequences have
focused on multilateral cooperation and international institutions. The success of these international efforts has been limited. A prime example is the Kyoto protocol. Especially large emitters
of greenhouse gases such as the United States, which refused to ratify the Kyoto protocol, are
generally reluctant to contribute to this global environmental effort. Given that effective cooperation between nation-states in climate politics is very difficult, the question arises whether there
are other possibilities for coping with global climate change at lower political levels. In fact, such
climate policy efforts already exist. The largest effort of this kind is the U.S. Mayors Climate
Protection Agreement (MCPA). It was initiated by U.S. cities in 2005. Its aim is to advance the
goals of the Kyoto Protocol on the local level even if the U.S. federal government lags behind.
As of November 2010, 1044 cities have signed this agreement, representing 87 million people,
i.e., more than one quarter of the U.S. population.
Since the subnational formation of climate policy institutions is a new phenomenon, we know
very little about the conditions that motivate subnational jurisdictions to join such efforts. Such
knowledge is important for the evaluation of whether cooperation at this level can offer a feasible
substitute or at least a useful complement for global efforts. This dissertation seeks to fill this research gap. It examines the determinants of cities’ willingness to commit to local climate change
policies. I develop arguments on how various factors influence local governments’ decisions to
voluntarily contribute to climate change mitigation efforts. These factors include communityspecific (e.g., income, partisanship, education) and interdependency factors. I argue that local
governments’ decisions concerning climate change policies are dependent upon the choices of
other local governments, and my aim is to explain these interdependencies in participation in
voluntary agreements by determining the role of external influences (e.g., geography or social
networks) that act as channels for the diffusion of policies.
Overall, the results from this inquiry show the importance of community-specific characteristics for participation in the Mayors Climate Protection Agreement. Conditions that are
vi
conducive to voluntary climate change mitigation efforts are a well-educated and liberal population. Factors pertaining to the natural environment play a role especially for those communities
situated along coastal areas. A substantive contribution of this thesis comes from the consideration of external factors that are assumed to have an impact on the locality above and beyond
its community-specific characteristics. I find some evidence for interdependent decision-making
in local climate change policymaking. The importance of social networks of mayors for the
diffusion of innovation is backed by evidence in both the quantitative as well as the qualitative
parts of this study. A further notable contribution of this thesis lies in the collection of a unique
data set on city-level MCPA adoption dates in seven Midwestern states. A particular merit of
the research design is the possibility to compare results obtained from the large-N contexts with
qualitative evidence from the local decision-makers.
Thus, this dissertation provides the first comprehensive and systematic account of the temporal and spatial diffusion of voluntary climate change policies within a large country, using
GIS and advanced statistical methods to that end. By studying in-depth the largest economy
and second largest emitter in the global system, I gain insights that are also relevant to other
countries, especially countries with federal political systems.
vii
Zusammenfassung
Die Folgen des globalen Klimawandels zu bekämpfen ist eine der grössten Herausforderungen
der Menschheit im 21. Jahrhundert. Zu den Versuchen mit Klimawandel umzugehen und
daraus entstehende Konsequenzen abzuschwächen gehörten bislang vor allem multilaterale Verhandlungen und internationale Institutionen. Der Erfolg solcher internationaler Anstrengungen
war jedoch begrenzt. Ein Paradebeispiel hierfür ist das Kyoto Protokoll. Vor allem Staaten,
die einen grossen Treibhausgasausstoss haben – wie z.B. die USA, welche auch das KyotoProtokoll nicht ratifiziert haben – zeigen sich generell zurückhaltend bis unwillig zu solchen
globalen Bemühungen beizutragen. Da eine effektive Zusammenarbeit zwischen Nationalstaaten
im Bereich der Klimapolitik auf globaler Ebene sehr schwierig ist, stellt sich die Frage, ob es weitere Möglichkeiten gibt, Klimawandel auf anderen politischen Ebenen nachhaltig zu bekämpfen.
In der Tat kann man feststellen, dass es bereits einige Anstrengungen sowohl auf verschiedenen Ebenen als auch zwischen Staaten gibt, bei denen es sich um Klimapolitik handelt. In
diesem Projekt soll nun vor allem ein detaillierter Blick auf die subnationalen Ebenen geworfen werden. Die grösste nationale Initiative dieser Art ist das U.S. Mayors Climate Protection
Agreement (MCPA). Diese Initiative wurde 2005 von U.S. Städten die von der Passivität der
eigenen Regierung innerhalb der Klimapolitik frustiert waren ins Leben gerufen. Das erklärte
Ziel dieser Initiative ist es, die für die USA im Kyoto-Protokoll anvisierten Treibhausgassenkungen in den jeweiligen Städten zu erreichen und hiermit ihren Teil zum Klimaschutz zu leisten.
2005 mit 141 Städten gestartet, umfasst das MCPA mittlerweile 1044 Städte in denen insgesamt 87 Millionen Menschen leben; das entspricht einem Viertel der US-Bevölkerung. Da die
Entstehung und Entwicklung solcher klimapolitischer Institutionen auf subnationaler Ebene ein
neues Phänomen darstellt, wissen wir vergleichsweise wenig über die Faktoren, die lokale Einheiten dazu veranlassen solchen freiwilligen Abkommen beizutreten. Genau dieses Wissen ist
jedoch erforderlich, um bewerten zu können, ob Kooperation auf Ebenen jenseits der globalen
die Möglichkeit bietet, ein Ersatz oder zumindest ein brauchbarer Zusatz zu sein. Diese Dissertation will zu dieser Forschung beitragen. Ich untersuche welche Bedingungen einen Beitritt einer
Stadt zu einem freiwilligen Klimaschutzprogramm fördern bzw. hindern. Ich erörtere hierbei
viii
zuerst theoretisch, welche Faktoren bei der Entscheidung der lokalen Entscheidungträger relevant sind. Diese Faktoren werden in interne (d.h. stadt-spezifische Faktoren wie z.B. Einkommen, Parteizugehörigkeit, Bildung) und externe Interdependenzfaktoren aufgeteilt. Ich argumentiere dass Entscheidungen lokaler Entscheidungsträger bezüglich Klimaschutzpolitiken von
den Entscheidungen anderer Entscheidungsträger abhängen. Mein Ziel ist es diese Interdependenzen in der Teilnahmen an freiwilligen Abkommen über die Rolle externer Einflüsse (z.B.
Geographie oder soziale Netzwerke) zu erklären.
Die Resultate meiner Untersuchung zeigen die Bedeutung von stadt-spezifischen Merkmalen
für die Teilnahme an freiwilligen Abkommen. Förderliche Bedingungen für eine Teilnahme sind
hier vor allem eine liberale und gut ausgebildete Bevölkerung. Ein substantiver Beitrag dieser
Dissertation ist die explizite Einbeziehung externer Faktoren, von denen eine Auswirkung auf
die Wahrscheinlichkeit einer Teilnahme an freiwilligen Klimaschutzabkommen erwartet wird.
Die Analyse zeigt einige Hinweise dass die Entscheidungen verschiedener Entscheidungsträger
voneinander abhängig sind. Zudem kann ein grosser Einfluss von sozialen Netzwerken auf die
Diffusion innovativer Politiken sowohl in der quantitativen als auch in der qualitativen Studie
festgestellt werden. Eine weiterer nennenswerter Beitrag dieser Studie liegt in der Erstellung
eines neuen Datensatzes zu MCPA Beitrittsdaten von Städten in sieben Staaten des Mittleren
Westens.
Diese Dissertation liefert folglich die erste umfassende und systematische Untersuchung
der zeitlichen und räumlichen Diffusion freiwilliger Klimaschutzpolitiken auf lokaler Ebene.
Dadurch, dass die grösste Volkswirtschaft und der zweitgrösste CO2 Verursacher im Detail
untersucht werden, können wichtige Einblicke gewonnen werden, die auch für andere Länder,
insbesondere jene mit föderalen politischen Systemen, Relevanz besitzen.
ix
Contents
1 Introduction
1
1.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
1.2
Motivation
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2
1.3
Structure of Dissertation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4
1.4
Conclusion and Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6
2 Climate Change Governance
9
2.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9
2.2
The Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10
2.2.1
Geophysical Aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10
2.2.2
Social Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
14
Evolution of the Global Governance System . . . . . . . . . . . . . . . . . . . . .
16
2.3.1
Goals of the Global Governance Effort . . . . . . . . . . . . . . . . . . . .
17
2.3.2
IPCC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18
2.3.3
FCCC and Kyoto Protocol . . . . . . . . . . . . . . . . . . . . . . . . . .
19
Why is International Cooperation Difficult? . . . . . . . . . . . . . . . . . . . . .
21
2.4.1
Global Public Goods and the Free-Rider Problem . . . . . . . . . . . . . .
21
2.4.2
The Contested Economics of Climate Change Mitigation . . . . . . . . . .
24
2.3
2.4
2.5
2.6
2.7
Measuring and Explaining Variation in National Contributions to the Global Public Good . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
25
2.5.1
Measuring Variation in Contributions to the Public Good . . . . . . . . .
25
2.5.2
Explaining Variation in Mitigation Efforts . . . . . . . . . . . . . . . . . .
26
Alternative Forms of Climate Change Governance: Local Dynamics in Federal
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
30
Normative Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31
x
3 Nature, Nurture or Neighbors? Testing Participation Patterns in Voluntary Initiatives
in U.S. Counties
35
3.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
35
3.2
Determinants of Participation in Voluntary Climate Change Initiatives . . . . . .
38
3.2.1
Natural System Factors (Nature) . . . . . . . . . . . . . . . . . . . . . . .
39
3.2.2
Socio-Economic and Political Factors (Nurture) . . . . . . . . . . . . . . .
41
3.2.3
External Influences (Neighbors) . . . . . . . . . . . . . . . . . . . . . . . .
45
Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
47
3.3.1
Dependent Variable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
48
3.3.2
Independent Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
50
Analysis and Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
55
3.4.1
Exploratory Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
56
3.4.2
Statistical Analysis and Results . . . . . . . . . . . . . . . . . . . . . . . .
59
3.4.3
Robustness Checks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
68
3.3
3.4
3.5
Conclusion
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
79
3.6
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
81
4 Voluntary Climate Change Initiatives in the U.S.: Testing Participation in Space and
Time
87
4.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
87
4.2
Background on Voluntary Climate Change Policies in the U.S. . . . . . . . . . .
90
4.3
Theoretical Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
95
4.3.1
Puzzle and Theoretical Argument
. . . . . . . . . . . . . . . . . . . . . .
95
4.3.2
Determinants of Adoption of Voluntary Climate Change Initiatives . . . .
99
4.4
4.5
Research Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
4.4.1
Case Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
4.4.2
Data Collection: Date of MCPA Adoption . . . . . . . . . . . . . . . . . . 114
4.4.3
Dependent Variable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
4.4.4
Independent Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.4.5
Statistical Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Analysis and Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
4.5.1
Statistical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
4.5.2
Qualitative Evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
Kurztitel
xi
4.6
Conclusion
4.7
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
4.7.1
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Questionnaire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
5 ’Where the Rubber Meets the Road’
Understanding the Mechanisms Leading to Cities’ Participation and Non-Participation
in the MCPA
157
5.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
5.2
Research Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
5.3
5.2.1
Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
5.2.2
Conduction of the Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
5.3.1
Reasons for Non-Participation
. . . . . . . . . . . . . . . . . . . . . . . . 163
5.3.2
Reasons for Participation . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
5.3.3
Further Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
5.3.4
Success of the Agreement as a Whole
5.3.5
Impact on the Federal Government . . . . . . . . . . . . . . . . . . . . . . 171
. . . . . . . . . . . . . . . . . . . . 169
5.4
Conclusion
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
5.5
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
5.5.1
Email . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
5.5.2
Study Information Sheet . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
5.5.3
Interview Template . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
xiii
List of Figures
2.1
Comparison of climate model predictions with empirical climate records (Source:
IPCC 2007) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12
2.2
GHG Emissions Scenarios (Source: IPCC 2007) . . . . . . . . . . . . . . . . . . .
13
2.3
Estimates of damage resulting from unmitigated climate change. Source: Tol and
Yohe 2006 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1
Counties in the U.S. according to their participation in the MCPA; darker shades
indicate a higher percentage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2
36
Climate zones by county for the U.S.: from 1(very hot, moist) to 7(very cold)
(Source Department of Energy) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3
15
40
Model projections of summer average temperature and precipitation changes
in Illinois and Michigan for midcentury (2040-2059), and end-of-century (20802099)(Source (Hayhoe et al. 2009) . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4
Detailed map of participation and cities in the northeast; darker shades indicate
a higher percentage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5
41
50
Counties with at least one city over 10.000 inhabitants; depicted in red are leaders
of innovation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
55
3.6
Interaction effect between income and % Democrats in the county . . . . . . . .
69
3.7
Interaction effect between % Democrats and unemployment in the county . . . .
71
3.8
Actual count distribution against negative binomial and poisson distribution . .
75
3.9
Comparison of the pattern of % MCPA and fatalities from natural hazards in the
1109 counties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
83
3.10 Comparison of the pattern of % Democrats and unemployment rate in the 1109
counties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
84
3.11 Analysis of spatial clusters with Local Moran’s I statistic (W: county contiguity)
in the 1109 counties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
85
xiv
3.12 (Cont’d)Analysis of spatial clusters with Local Moran’s I statistic (W: county
contiguity) in the 1109 counties . . . . . . . . . . . . . . . . . . . . . . . . . . . .
86
4.1
Cities that have signed the Mayors Climate Protection Agreement . . . . . . . .
88
4.2
Sample of cities and signatories in the seven Midwestern states . . . . . . . . . . 113
4.3
Adoption of the Mayors climate protection agreement (2005-2010) . . . . . . . . 118
4.4
% of cities who signed the MCPA by state (2005-2010) . . . . . . . . . . . . . . . 120
4.5
% of cities who signed the MCPA (2005-2010) . . . . . . . . . . . . . . . . . . . . 121
4.6
Neighbor relations within 100km . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
4.7
Neighbor relations within 100km, reduced sample >=30k . . . . . . . . . . . . . 125
4.8
Baseline hazard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.9
Predicted Probabilities (Baseline model); other variables are kept at their mean;
binary variables at their median . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
5.1
Random sample of 60 cities from 2700 cities according to their participatory status159
5.2
Interviewed cities from sample according to their participatory status
. . . . . . 161
xv
List of Tables
2.1
Stratospheric Ozone and Climate Cooperation . . . . . . . . . . . . . . . . . . . .
23
3.1
City-level distribution of MCPA participants . . . . . . . . . . . . . . . . . . . .
49
3.2
Distribution of counties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
49
3.3
Regression table (Baseline) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
60
3.4
Spatial ML; W based on county contiguity (0 / 1 ) . . . . . . . . . . . . . . . . .
64
3.5
Spatial ML; W based on Inverse Distance . . . . . . . . . . . . . . . . . . . . . .
65
3.6
Spatial ML; W based on Inverse Distance up to 100 km . . . . . . . . . . . . . .
67
3.7
Basic Models with Proximity to Leaders of Innovation . . . . . . . . . . . . . . .
67
3.8
Regression table with Interaction terms . . . . . . . . . . . . . . . . . . . . . . .
70
3.9
Logistic Regression whether at least one city in the county has signed MCPA . .
72
3.10 OLS (Mean) Regression vs. Quantile (Median) Regression with Koenker-Basset
(kb) (1982) standard errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
74
3.11 Negative Binomial Regression on the count of MCPA signatories in the county .
77
3.12 Zero Inflated Negative Binomial Regression on the count of MCPA signatories in
the county . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.13 Summary statistics
78
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
81
3.14 Pattern of signature of MCPA in all states . . . . . . . . . . . . . . . . . . . . . .
82
4.1
Comparison of the 7 Midwestern states and the U.S. average . . . . . . . . . . . 111
4.2
Climate change related policies on the state-level (Source: Pew Center on Global
Climate Change) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.3
Signatories (cities > 10k) per state by July 2010 . . . . . . . . . . . . . . . . . . 118
4.4
Main Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
4.5
Robustness of main results I; exclucding t and reduced model
4.6
Robustness of main results II; external factors
4.7
Robustness of main results III: networks and impute . . . . . . . . . . . . . . . . 137
. . . . . . . . . . 133
. . . . . . . . . . . . . . . . . . . 136
xvi
4.8
Robustness of main results IV: reduced sample (cities >= 30.000)
. . . . . . . . 138
4.9
Simulated predicted probabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
4.10 Simulated Probabilities, City Examples (for May 2005) . . . . . . . . . . . . . . . 142
4.11 Qualitative evidence is based on returned questionnaires from the following cities 143
5.1
Distribution of states in the sample . . . . . . . . . . . . . . . . . . . . . . . . . . 160
5.2
List of Interviewees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
1
1 Introduction
Lena Maria Schaffer
1.1 Introduction
Global climate change is one of the biggest challenges mankind is facing in the 21st century.
Prominent attempts to deal with global climate change and to mitigate its consequences have
focused on multilateral cooperation and international institutions. The success of these international efforts has been limited. A prime example is the Kyoto protocol. Especially large emitters
of greenhouse gases such as the United States, which refused to ratify the Kyoto protocol, are
generally reluctant to contribute to this global environmental effort. Given that effective cooperation between nation-states in climate politics is very difficult, the question arises whether there
are other possibilities for coping with global climate change at lower political levels. In fact, such
climate policy efforts already exist. The largest effort of this kind is the U.S. Mayors Climate
Protection Agreement (MCPA). It was initiated by U.S. cities in 2005. Its aim is to advance the
goals of the Kyoto Protocol on the local level even if the U.S. federal government lags behind.
As of November 2010, 1044 cities have signed this agreement, representing 87 million people,
i.e., more than one quarter of the U.S. population.
Since the subnational formation of climate policy institutions is a new phenomenon, we know
very little about the conditions that motivate subnational jurisdictions to join such efforts. Such
knowledge is important for the evaluation of whether cooperation at this level can offer a feasible
substitute or at least a useful complement for global efforts. This dissertation seeks to fill this research gap. It examines the determinants of cities’ willingness to commit to local climate change
policies. I develop arguments on how various factors influence local governments’ decisions to
voluntarily contribute to climate change mitigation efforts. These factors include communityspecific (e.g. income, partisanship, education) and interdependency factors. I argue that local
governments’ decisions concerning climate change policies are dependent upon the choices of
other local governments, and my aim is to explain these interdependencies in participation in
2
voluntary agreements by determining the role of external influences (e.g. geography or social
networks) that act as channels for the diffusion of policies. In the context of the Mayors Climate
Protection Agreement, the two main research question of this dissertation are:
1. What are the internal characteristics that lead communities to voluntarily commit to local
climate change policies, when there are powerful incentives to do otherwise?
2. What influence do interdependence and network effects play?
These two research questions are addressed in different scope throughout the four papers of
this dissertation. The first paper, which is co-authored with Thomas Bernauer, gives an overview
of global climate change governance and provides the framework within which the following papers on local climate change governance operate. The second paper, then, looks at participation
patterns in the Mayors Climate Protection Agreement throughout the United States. I argue
that to explain county-level patterns, natural characteristics (Nature), socio-economic and political features (Nurture), as well as the geographical surroundings (Neighbors) have to be taken
into account. Interdependence is conceptualized through geographical proximity to neighboring
counties and to leaders of innovation. In the third paper, my focus then zooms in on the individual city. I argue that three factors are crucial to the explanation of the temporal and spatial
patterns in participation in voluntary climate change agreements: city-internal characteristics,
external factors, and time-specific events. External factors that matter for the interdependence
in cities’ decision-making concerning the MCPA are geographical neighbors, peers, and social
networks. The fourth and final paper then sets out to validate the previous findings concerning
internal and external factors using evidence from interviews with mayors and representatives.
This paper further addresses the reasons of non-participation with the initiative.
1.2 Motivation
The general motivation for looking at these questions is threefold. First, the substantive topic of
voluntary political efforts to circumvent the collective action problem in the context of a global
public good is inherently interesting and has not been comprehensively analyzed in the context
of local climate change policies. Furthermore, arguably absent of an overarching authority that
would regulate GHG emissions, implications drawn from this setting represent a unique insight
also for models operating from an international politics and comparative politics perspective,
where no central authority does exist and collective action problems are omnipresent.
Introduction
3
Second, at the theoretical level, the fact that one observes political tensions arising from
different approaches (federal and subnational) to the same problem (climate change) within
one nation is in itself an interesting topic. Given that research on comparative environmental
governance tends to focus on national or supranational policies, but does largely neglect the
subnational level, makes it an innovative endeavour to study the subnational interdependence
of policy initiatives. Recent developments in the U.S. show that researchers should pay more
attention to federalism in general and in particular to the reaction of subnational units in
the absence of federal regulation. Furthermore, results from research on these processes can
be important for the understanding of environmental policy-making in other federal states or
systems of multi-level governance (e.g. the European Union) or, alternatively, in other issue
areas.
Third, at the methodological level, for many phenomena of interest to social scientists, it is
very difficult to argue that we are dealing with independent observations. Far more often, the
policy choices of governments depend on previous decisions by other governments. This is mostly
referred to as Galton’s Problem (Braun & Gilardi 2006; Jahn 2006). From a methodological
perspective, incorporating interdependencies and taking Galton’s Problem seriously is recognized
as a necessary step towards improving explanatory models (Simmons & Elkins 2004; Franzese
& Hays 2007, 2008b). This task is especially difficult in the comparative and international
politics field, where one normally has to control for strong in-sample heterogeneity on top of the
difficulties of distinguishing spurious diffusion (Braun & Gilardi 2006) from truly interdependent
behavior. Therefore, one of the goals of my dissertation is to contribute to existing research by
studying local government behavior towards participation in voluntary initiatives, taking into
account city-specific socio-economic, political, as well as environmental factors and distinguishing
these factors from interdependent decision-making that emanates from external factors.
Thus, this dissertation provides the first comprehensive and systematic account of the temporal and spatial diffusion of voluntary climate change policies within a large country, using
GIS and advanced statistical methods to that end. By studying in-depth the largest economy
and second largest emitter in the global system, I gain insights that are also relevant to other
countries, especially countries with federal political systems. The findings of this dissertation
contribute to a better understanding of whether and how subnational climate change mitigation
efforts can complement policies at the international level.
4
1.3 Structure of Dissertation
In the first paper, Thomas Bernauer and I provide an overview over global climate change
governance. We first introduce the climate change problem and provide an historical account on
how international institutions have thus far coped with the problem. Having elaborated that the
standard governance level for mitigating GHG emissions and, thus, combating global warming,
would be a concerted global effort combined with implementation of internationally agreed
measures at the level of nation-states, we then turn to the reasons that make the implementation
of global governance systems so hard to achieve. Even though cooperation at the global level is
difficult, there is strong variation in countries’ level of effort in this respect. We examine how
levels of effort can be measured and how variation in effort can be explained. After having moved
from the global (systemic) to the national level of analysis, we also explore climate policy-making
at the subnational level. We establish that local policy-making is, from an analytical viewpoint,
particularly interesting in the case of federal political systems.
The three subsequent papers concentrate on the Mayors’ Climate Protection Agreement. In
the second paper, I look at the participation pattern of U.S. counties and ask which incentives and
disincentives for participation exist. The paper explains participation patterns in the MCPA at
the county-level throughout the U.S. by factors regarding the natural characteristics (Nature),
socio-economic and political features (Nurture), and geographical surroundings (Neighbors).
It is hypothesized that voluntary participation in climate change mitigation efforts are highest
when the county is at risk of being adversely affected by climate change, when there are favorable
socio-economic and political conditions present, and when the surrounding counties are also
strong participators.
As far as the natural system characteristics are concerned, being a coastal county seems to
drive signature behavior. For socio-economic and political characteristics, drivers of participation are a well-educated population as well as a high percentage of Democratic voters within
the county. Furthermore, the presence of universities significantly increases the relative number
of signatories. Disincentives for participation include especially bad economic conditions in a
county. Depending on model specification, proxies for potential abatement costs are also significantly related to participation. Findings from the spatial analysis indicate some evidence of
spatial interdependence in participation, meaning that some of the variation in a county’s participation level can be attributed to participation levels in connected counties. However, of the
four different specifications of geographic influence tested, only two support the conjecture of a
Introduction
5
significant effect of spatial interdependence. In this cross-sectional large-N perspective, some of
the issues that might play a large role for participation namely the temporal variation in signing
on to the agreement cannot be studied. Furthermore, while explaining the participation rate in
counties serves as a good starting point to examine which factors are important to voluntary
action, it cannot provide a complete picture of the phenomenon, as it is on the city-level where
the decision-makers within this agreement operate.
This is where the third paper ties in. It examines the determinants of cities’ willingness
to join the MCPA in cities over 10,000 inhabitants in seven Midwestern states. I develop
arguments on how various factors influence local government decisions to voluntarily contribute
to climate change mitigation efforts by joining the Mayors Climate Protection Agreement. These
factors include community-specific internal factors and external factors accounting for cities’
interdependence in decision-making. As far as internal factors are concerned, I argue that there
are specific determinants leading to a higher demand for voluntary climate change policies within
the city, e.g. a higher percentage of voters that support the Democratic Party, which then make
governments more likely to participate in voluntary efforts. I further assume that an initiative to
join the MCPA can also emanate from the supply side of local government. Here, the existence
of a policy entrepreneur and a mayor-council form of government are considered to increase the
propensity to participate. Interdependencies between cities are conceptualized as emanating
from geographic linkages to other jurisdictions, peer group effects, as well as from participation
in social networks. My theoretical framework is complemented by the consideration of certain
events that are assumed to be associated with a city’s decision to sign on to the MCPA.
To test my claims, I use an original data set on monthly signing behavior of 749 Midwestern
cities from February 2005 to June 2010. I find that a city’s participation in social networks as
well as the MCPA signing behavior of cities within the same population group increases a city’s
likelihood to participate. Although internal characteristics, such as the % of Democratic votes
in the city or human capital endowment, are also significantly related to a higher propensity to
sign, their substantive effects are much smaller. As far as pivotal events are concerned, I find
that the likelihood of participation significantly increases in the months leading up to a mayoral
election in the city. Insights gained from this large-N study are further supported by qualitative
evidence from an analysis of questionnaires sent to a subsample of the 749 Midwestern cities.
Although this analysis provides me with a more complete picture of the temporal dynamics, the
6
mayors as the actors behind the a city’s participation decision have only partially been taken
into account.
The previous three papers of this dissertation on local climate change policies continuously
reduce their focus. Starting from a general overview of global climate change governance, I then
concentrate on subnational units within the U.S. I explain county-level participation patterns in
the Mayors Climate Protection Agreement as well as city-level diffusion effects of the initiative
in a temporal setting. In the two papers on the MCPA, mayors have implicitly been considered
to be the ones that have to make the decision to participate in MCPA, but have not been
explicitly asked about the reasons why they decided to (not) participate in the MCPA. The last
paper finally opens the black box of the decision-maker in providing qualitative evidence from
interviews with mayors. As regards the motivational reasons of mayors to join the MCPA, a
majority of the interviewed mayors that have not signed, argue that the whole MCPA initiative
is only about symbolic politics. In their opinion it serves to potentially advance a signing
mayor’s political career, but produces no significant climate change policy output. Conversely,
mayors who have signed the agreement claim that their reasons for participation had to do
either with the federal inaction on the climate change topic or an already existing interest for
environmental politics. With respect to the reasons for not signing, mayors state that there
either was no demand for such policies or they have other local policy priorities, e.g. economic
development policies.
1.4 Conclusion and Outlook
Recent developments in the U.S., but also on a global scale, show that researchers should pay
more attention to the reaction of subnational units to climate change in the absence of federal
or global regulation. On November 21, 2010, Mayors from around the world met at the World
Mayors Summit on Climate in Mexico City and signed a voluntary pact committing themselves
to the reduction of urban greenhouse gas emissions.1 They agreed to install a ’carbon Cities
Climate Registry’ (cCCR) as a first step towards harmonizing the reporting of urban GHGs.
Through this mechanism, city residents will be able to track how their city is performing vis-a-vis
other cities around the globe.
1
For more information, visit http://edition.cnn.com/2010/WORLD/americas/11/22/world.mayor.summit.
review/index.html?hpt=Mid
Introduction
7
With the increasing commitment of cities in the U.S. and worldwide to the reduction of CO2
emissions in order to mitigate the consequences of anthropogenic climate change, it becomes ever
more important to understand the factors that make localities more likely to engage in local
climate change policies. This dissertation has provided the first comprehensive and systematic
account of the main factors that lead to participation as well as concerning the temporal and
spatial diffusion of voluntary climate change policies in the U.S.
Overall, the results of this inquiry highlight the importance of community-specific characteristics for the participation in the Mayors Climate Protection Agreement. Conditions that are
conducive to voluntary climate change mitigation efforts are a well-educated and liberal population. Factors pertaining to the natural environment play a role, especially for those communities
situated along coastal areas, linking participation to the vulnerability to impacts from global
climate change.
A substantive contribution of this thesis arises from the consideration of external factors
that are assumed to have an impact on the locality above and beyond its community-specific
characteristics. I find some evidence for interdependent decision-making in local climate change
policymaking. A notable finding of my thesis is that geographical proximity does not turn
out to be a robust factor to explain interdependencies in local climate change policy adoption.
This contradicts most of the scholarly literature on policy diffusion, where the influence of
geographical neighbors on policy adoption is one of the most consistent research findings (Berry
& Berry 1999). A comparatively stronger external effect on a local community’s decision to
sign the agreement emanates from social networks and peer groups. The importance of social
networks of mayors for the diffusion of innovation is backed by evidence in both my quantitative
and qualitative studies. While the influence of other cities’ choices on local climate change
policies is consistently confirmed, it becomes clear that ’space is more than geography’ (Beck
et al. 2006) when it comes to MCPA participation .
An additional notable contribution of this thesis lies in the collection of a unique data set
on city-level MCPA adoption dates in seven Midwestern states. The particular merit of this
large-scale data collection effort lies in the possibilities that it creates for the comparison of
results from large-N studies and qualitative evidence from local decision-makers.
Future research should build on and extend the findings of my research by looking into the
patterns of actual implementation and the substantive policy output that emerges from subnational climate change efforts. Such an extension would allow us to shed light on the different
8
motivations of decision-makers to join voluntary efforts. Being able to distinguish symbolic participation from substantive participation would be a major improvement over existing research.
Nevertheless, even if the actual impact from these local climate change initiatives on measurable indicators of mitigation would turn out to be negligible, my research has already found
that even symbolic, voluntary agreements can act as good complements to actual policies in
that they raise awareness for the issue and break common ground. The mayors’ engagement in
climate change issues has definitely brought cities on the forefront of this issue. This can be seen
as a fairly new development in American politics, where states are traditionally considered to
be the ’laboratories of democracy’ that test novel approaches that eventually become national
policy. Together with states, cities have emerged as the de facto leaders on climate change issues. Although a comprehensive national policy on climate change mitigation has still not been
enacted and now, after the November 2010 election, looks even less likely to be enacted, actions
in individual states and cities put pressure on the federal government by creating a patchwork
of different rules in different states and cities that goes at the expense of and causes frustration
among corporate actors. Furthermore, due to the 2009 decision by the Environmental Protection Agency (EPA) that greenhouse gases can be regulated under the Clean Air Act if they pose
a danger to public health, the EPA has gained additional powers to regulate, e.g. minimum
standards of energy efficiency. However, what happens within the United States on the issue
of climate change mitigation policy as well as on a global scale is far from being settled and
remains an important and interesting question for academic inquiry.
9
2 Climate Change Governance
Thomas Bernauer and Lena Maria Schaffer
2.1 Introduction
Within less than three decades, climate change has developed from a rather obscure scientific
topic into a key item on the global political agenda. It has also attracted strong attention in
many areas of scientific research, including the social sciences. Social scientists, and notably
governance specialists focusing on climate change have addressed a wide range of important
questions, including the following:
• What are the key political challenges in establishing and implementing governance systems
to cope with climatic changes?
• Why are some countries in the international system more cooperative than others in this
respect?
• To what extent can local efforts in climate policy support national and global efforts?
• What are the main normative issues associated with climate change policy, notably, how
should the costs and benefits associated with solving the problem be distributed across
countries and time?
• Which policy instruments are likely to be more effective and/or efficient in dealing with
climate change?
In this contribution we focus mainly on the first four of these questions. The existing literature on particular climate policy instruments (e.g. carbon taxes, tradable permits, joint
implementation) is very large and has been summarized elsewhere (e.g. Stavins 2003).1
1
These include regulations and standards, taxes and charges, tradable emissions permits, subsidies and tax
credits, voluntary agreements between industry and government, awareness campaigns, government sponsored
and/or subsidized R&D, and integration of climate policy objectives in development, trade, and investment
policies.
10
We start with an overview of the climate change problem (section 2), followed by a discussion
of international institutions that have thus far been established to cope with the challenge
(section 3). We then look at the reasons why global cooperation for climate change mitigation
is difficult to achieve (section 4). Section 5 shows that, even though cooperation at the global
level is difficult, there is strong variation in countries’ level of effort in this respect. We examine
how levels of effort can be measured and how variation in effort can be explained. After having
moved from the global (systemic) to the national level of analysis, we also explore climate
policy-making at the sub national level (section 6). Local policy-making is, from an analytical
viewpoint, particularly interesting in the case of federal political systems. The chapter ends in
section 7 with a brief discussion of normative issues.
2.2 The Problem
2.2.1 Geophysical Aspects
In contrast to the weather, which is highly variable both spatially and temporally, the global
climate is much more stable. It can be regarded as the Earth’s average weather and/or its
variability over longer periods of time (typically at least decades). Whereas the weather can be
experienced directly by humans, the climate is a scientific (essentially statistical) construct. For
instance, while temperatures can easily vary by 20˚C in a particular location within a single
day, the average global temperature does not vary by more than 1-5˚C within time-spans of
thousands of years.
Changes in the Earth’s climate took place also in pre-modern times (before the industrial
revolution). Such changes occurred due to non-human factors, e.g. changes in heat output
of the sun and volcanic activity. However, starting in 1896 with the Swedish chemist Svante
Arrhenius, scientists have produced a mounting stream of evidence demonstrating that so called
greenhouse gases (GHG) emitted by human activities are influencing the Earth’s climate as well.
Several gases in the atmosphere, most notably water vapor and carbon dioxide (CO2 ), are
instrumental in trapping some of the sun’s energy to which the Earth is exposed. This greenhouse
effect is essential for life on Earth. Without this heat trapping the Earth would be more than
30˚C colder. Yet, human activity, in particular the combustion of fossil fuels (coal, oil, gas)
and land-use changes, have led to a large increase in concentrations of GHGs in the atmosphere.
Atmospheric concentrations of the two most important GHGs emitted by human activity, CO2
Climate Change Governance
11
(carbon dioxide) and CH4 (methane)2 , were far higher in the year 2005 than the natural range
of these gases in the past 650.000 years. CO2 has increased from a pre-industrial (i.e. prior to
about 1750) level of 280ppm to 379ppm in 2005, and CH4 from 715 to 1774ppm (IPCC 2007b).
Climate scientists have over the past several decades invested an enormous amount of effort
in trying to understand the causal pathways leading from
1. vastly increased anthropogenic GHG emissions since the industrial revolution (with a 70%
growth in 1970-2004 alone) to
2. growing atmospheric concentrations of GHGs to
3. changes in radiative forcing to
4. changes in temperature (about 0.75˚C over the past 100 years, with more warming in
northern latitudes, and greater warming over land than over the oceans) and precipitation
to
5. various effects of changes in temperature and precipitation on plants, animals, and humans.
The Intergovernmental Panel on Climate Change (c.f. section 3.2 IPCC 2007a), a large
global network including thousands of scientists and also policy-makers, has so far issued four
comprehensive reports. These reports summarize and assess the available scientific evidence
on the causes and implications of climate change, as well as policy options for coping with
the problem. The IPCC stands out, by orders of magnitude, as the largest and most tightly
organized science-policy nexus in the history of governance efforts in any policy-area we can
think of.
This unprecedented scientific effort has over the past two decades resulted in increasingly
firm international agreement that anthropogenic GHG emissions are responsible for a large part
of the observed global warming trend. Interestingly, the main conclusion from this scientific
effort is rather close to what Arrhenius argued more than 100 years ago; that a doubling of
CO2 concentrations in the atmosphere would increase global average temperature by around
5˚C (current estimates are around 1.5 to 4.5 degrees). More generally, the IPCC (2007b, 10)
notes that: ’Most of the observed increase in globally averaged temperatures since the mid-20th
century is very likely due to the observed increase in anthropogenic (human) greenhouse gas
concentrations.’
2
Other important GHGs include nitrous oxide (N2O) and halocarbons.
12
Figure 2.1: Comparison of climate model predictions with empirical climate records (Source:
IPCC 2007)
Note: Blue = models including only natural causes. Red = models including natural and
anthropogenic causes. Black line = decadal averages of observations 1906-2005 plotted against
the centre of the decade and relative to the corresponding average for 1901-1950. Dashed line
= spatial coverage less than 50%.Blue shaded bands = 5 to 95% range for 19 simulations from
five climate models using only the natural forcings (solar activity, volcanoes). Red shaded
Bands: 5 to 95% range for 58 simulations from 14 climate models using both natural and
anthropogenic forcings.
Figure 2.1 illustrates that computer models trying to reconstruct the empirical climate record
tend to perform better once anthropogenic emissions are included alongside non-human drivers
of climatic changes.
While ex post explanation of climatic changes in terms of anthropogenic GHG emissions
is very complex, prediction of future temperature and precipitation is even more challenging.
The main reason is that, besides incomplete understanding of geophysical mechanisms, there is
great uncertainty concerning future GHG emissions. For instance, depending on assumptions
about technological innovations, economic growth (which in itself is hard to predict over several
decades) may be associated with very different levels of emissions.
As illustrated by Figure 2.2, one IPCC scenario (A1FI) assumes rapid economic growth, rapid
introduction of new and more efficient but fossil fuel intensive technologies, a mid century peak
of global population, and a substantial reduction in regional differences in per capita income. In
Climate Change Governance
13
!
Figure 2.2: GHG Emissions Scenarios (Source: IPCC 2007)
Note: Global GHG emissions (in GtCO2 -eq per year) in the absence of additional climate
policies: six illustrative SRES marker scenarios (coloured lines) and 80th percentile range of
scenarios published since SRES (post-SRES) (gray shaded area). Dashed lines show the full
range of post- SRES scenarios. The emissions include CO2 , CH4 , N2O and F-gases
this scenario, global GHG emissions are predicted to increase from around 40 Gt CO2 -eq/yr in
2000 to about 130 Gt in 2100. Another IPCC scenario (B1) assumes a convergent world with
the same population, but rapid changes in economic structures toward a service and information
economy, the introduction of clean and resource efficient technologies, and an emphasis on global
solutions to problems of environmental sustainability. In this scenario, emissions are predicted to
increase to about 30 Gt CO2 -eq/yr by 2100 (IPCC 2007, Special Report on Emissions scenarios).
Because GHGs are quite long-lived, even very optimistic emissions scenarios are likely to result
in considerable global warming. The IPCC (46ff. 2007a) notes that:
’Anthropogenic warming and sea level rise would continue for centuries due to
the timescales associated with climate processes and feedbacks, even if greenhouse
14
gas concentrations were to be stabilized, although the likely amount of temperature
and sea level rise varies greatly depending on the fossil intensity of human activity
during the next century [...]. The probability that this is caused by natural climatic
processes alone is less than 5% ...World temperatures could rise by between 1.1 and
6.4˚C during the 21st century. Sea levels will probably rise by 18 to 59 cm [...].
There is a confidence level >90% that there will be more frequent warm spells, heat
waves and heavy rainfall [...]. There is a confidence level >66% that there will be
an increase in droughts, tropical cyclones and extreme high tides [...]. Both past
and future anthropogenic carbon dioxide emissions will continue to contribute to
warming and sea level rise for more than a millennium.’
2.2.2 Social Implications
The projected social implications of climatic changes depend very much on projected emissions
and their radiative forcing. The IPCC (2007a)’s best estimates range from +0.6˚C by 20902099, compared to 1980-1999, in the case of continuing year 2000 concentrations (which is next
to impossible in view of still growing global emissions), to +2.8˚C in a moderately optimistic
scenario (A1B), to 4.0˚C and more in pessimistic scenarios. Projected sea level rise by the end
of the 21st century is up to 0.59 meters in the standard scenarios. In extreme scenarios, such
as those involving a complete loss of the Greenland and West Antarctica ice sheets, sea levels
could rise by 7 meters or more.
Besides the large natural sciences literature on the implications of climate change for weather
patterns, water availability, natural disasters, plants, animals, and ecosystems, a considerable
social sciences literature on climate change implications has developed as well. This literature
seeks to clarify the social, economic, political and security implications of climate change.
The largest part of existing social sciences research examines climate change implications in
terms of economic losses and other forms of social damage (e.g. changing livelihoods, public
health problems, migration), as well as adaptive capacity (e.g. Adger 2010; Fuessel 2010). By
and large, this research arrives at the conclusion that poorer countries are at greatest risk,
both in terms of exposure to climatic changes and sensitivity to such changes, and in terms of
their capacity to adapt. Exposure, sensitivity, and capacity to adapt determine how vulnerable
particular countries or social groups are to climatic changes (Fuessel 2010).
Social scientists have also sought to quantify overall effects of climatic changes on economic
growth in the past and project economic losses under different emissions and mitigation scenarios
into the future (e.g. Stern et al. 2006; Stern 2008). Ex post statistical analysis has thus far
produced some evidence that higher temperature and lower precipitation are associated with
Climate Change Governance
15
Figure 2.3: Estimates of damage resulting from unmitigated climate change. Source: Tol and
Yohe 2006
lower economic growth, particularly in Africa, though these findings are not very robust (e.g.
Miguel et al. 2004; Dell et al. 2008; Bernauer, Koubi, Kalbhenn & Ruoff 2010).
Estimates of future effects on economic growth under different climate scenarios are based on so
called integrated assessment models that explore national, regional, and global cost implications.
The findings from these models vary enormously. Having reviewed many such studies, the IPCC
(2007b) for instance concludes: ’Global mean losses could be 1-5% of GDP for 4 degrees of
warming, but regional losses could be substantially higher.’ Yet, as illustrated by Figure 2.3, cost
implications reported by the influential Stern Review (Stern et al. 2006) are substantially larger
than estimates provided by other scientific reports.
One of the principal sources of vast differences in estimated costs of unmitigated climate
change (i.e. costs in the absence of international cooperation and GHG emissions cuts) is the
discount rate. The discount rate captures the extent to which future losses are less important
economically than present losses. The reasons for discounting future losses are that people
generally prefer the present to the future, that consumption will be higher in the future due to
increased wealth (with decreasing marginal utility), that future consumption levels are uncertain,
and that future technology may make it cheaper to cut emissions then. The higher the discount
rate used to deflate the stream of future losses to a present value, the lower is the presently
valued damage from future climate change. For instance, if we assume an annual discount rate
16
of 3%, a climate damage of $100 occurring in 25 years is worth only $50 today. While some
economists (e.g. Nordhaus 2010) use standard discount rates from the investment world (around
2-3%), others (e.g. Stern 2008; Cline 1999) argue that such discount rates are too high. They
use discount rates in the order of 1-2%, which are close to real interest rates for government
bonds. The choice of discount rate has important implications for the assessment of governance
options and involves strong normative components, to which we return in the final section of
the chapter (see final section of the chapter).
Repeated statements by high-ranking politicians about climate change-related wars have triggered yet another intense research effort in which social scientists are examining the validity of
this claim. US President Obama, for instance, claimed in 20093 that
’The threat of climate change is serious, it is urgent and it is growing [...] The
security and stability of each nation and all peoples – our prosperity, our health, our
safety – are in jeopardy. And the time we have to reverse this tide is running out.’
The most likely scenario for an interstate war involves competition over scarce international
water resources, food and energy, or mass migration (for extreme scenarios, see Schwartz &
Randall 2003). Interestingly, existing research offers virtually no historical evidence for climate
related international wars. Whether climate change could increase the probability of intrastate
(i.e. civil) war is more strongly debated. A few studies (e.g. Burke et al. 2009) identify such
an effect for Africa in the 1980s and 1990s and make rather worrying projections for the future.
Yet, these findings remain very much contested and other authors, using similar data, do not
find a significant effect of climatic changes on the probability of intrastate war (e.g. Buhaug
et al. 2008; Theisen et al. 2010; Bernauer, Koubi, Kalbhenn & Ruoff 2010).
2.3 Evolution of the Global Governance System
As noted above, science plays a major role in climate policy. Hence we start by discussing
what are, from the viewpoint of many scientists and policy-makers, the basic goals of the global
governance effort. We then describe the IPCC, the principal global institution for knowledgegeneration in this policy area. Finally we discuss the UN Framework Convention on Climate
Change (FCCC) and the Kyoto Protocol (KP). The latter two are, from a legal viewpoint, the
backbone of the existing global governance system.
3
Delivering a speech at the climate change summit of the United Nations on 22nd of September 2009.
Climate Change Governance
17
2.3.1 Goals of the Global Governance Effort
A strong global consensus has emerged over the past few years that climatic changes must be
addressed through mitigation of GHG emissions and, because some major climatic changes are
unavoidable even with extremely ambitious mitigation efforts, adaptation. The key questions in
this respect are:
1. by how much should GHG emissions be reduced, and in what time frame?
2. how much would this cost, and how should the burden be distributed among countries and
over time?
3. how much should be invested in adaptation and who should pay for it?
(1) The policy positions of many countries have, over the past few years, converged on the
goal of limiting the global average temperature increase to 2˚C, relative to the mid-18th century
level. From the perspective of most scientists, a temperature target makes more sense than an
emissions or concentrations target because it is ultimately temperature that affects ecosystems
and humanity. The 2˚C target emerged from discussions among scientists and policy-makers
in Germany in the mid-1990s. The 2˚C temperature increase was initially used as a rather
arbitrarily chosen parameter to examine climate change impacts, e.g. impacts on the Earth’s
major ice sheets. When many models indicated major damages or uncertainties beyond that
level (e.g. with respect to the long-term stability of the Greenland ice sheet), the two degrees
developed into a political target, even though there is no clear-cut scientific reason for this
particular choice.
Recent calculations by Allen et al. (2009) show that it would be necessary to limit total CO2
emissions in the 2000 – 2050 period to 1000 billion tons in order to meet the 2˚C target. One
third of this CO2 budget has already been used in 2000 – 2009. Consequently, emissions would
have to be cut by 50% by 2050, which would implicate reductions of 25-40% by industrialized
countries until 2020, and 80-95% until 2050. These targets are, by and large, in line with IPCC
2007 statements and the Stern Review.
(2) Various studies have tried to estimate by how much global carbon prices (the total
cost an emitter of a unit of GHG would have to pay for) would have to increase in order
to reach specific reduction targets. The IPCC (2007b) for instance notes a figure of $20-80
18
per ton of CO2 equivalent by 2030 to stabilize GHG concentrations at 550ppm (roughly a
doubling of pre-industrial concentrations, which were 280ppm then and 379ppm in 2005) by
2100. Optimistic studies indicate $5-$65 (IPCC 2007b).
The IPCC’s best estimates of the costs of stabilizing GHG concentrations at 535-590ppm,
which would probably meet the 2˚C target, are in the order of a 0.1% reduction of average
annual GDP growth rates. The Stern Report arrives at a similar estimate.
On the more pessimistic side, Nordhaus (2010) estimates that reaching the 2˚C target would
require a carbon price of $64 in 2010 (at 2005 prices), whereas the global average price today is
around $5, and rapid growth of this price over the next few years.
How to share the burden of GHG reductions remains disputed. At the most general level,
there is agreement that industrialized countries must shoulder most of the mitigation costs over
the coming decades. The Kyoto Protocol (see below) in fact assigns that responsibility to this
group of countries in the 2008 – 2012 period. But there is no consensus on how to deal with
very large, and rapidly growing developing countries, notably Brazil, China, and India. We
return to this point in the final section of the chapter.
(3) As noted by the IPCC (2007b): ’Much less information is available about the costs and
effectiveness of adaptation measures than about mitigation measures.’ In any event, the costs
are likely to be high and can most probably not be met by poor countries, which tend to be
most vulnerable to climatic changes. Estimates of adaptation costs range from lower two digit
billion figures to $200 billion and more per year. At the Copenhagen Conference in late 2009,
industrialized countries promised adaptation support in the order of $100 billion per year in the
future. But it remains unclear how firm these promises really are, how much each industrialized
country would contribute, and how the funding mechanism should be designed.
2.3.2 IPCC
The IPCC is an intergovernmental institution. Its task is to summarize and assess existing
scientific knowledge on human-induced climate change and its impacts, as well as options for
mitigation and adaptation. It was set up in 1988 by the UN World Meteorological Organization
(WMO) and the UN Environment Programme (UNEP). Its secretariat is located in Geneva,
Switzerland. Its activities are funded by WMO, UNEP, and by direct contributions from governments.
Climate Change Governance
19
The IPCC does not carry out ’in-house’ research, nor does it act as a monitoring agency in
implementing global climate agreements (see below). It acts primarily as manager of a large
network of scientists worldwide. Its activity centers around so called Assessment Reports. Such
reports have thus far been published in 1990/92, 1995, 2001 and 2007. The next report is scheduled for 2014. The scientists involved, usually several thousand from more than one hundred
countries, review the relevant scientific literature and, with the help of lead editors, summarize
and assess the existing knowledge. This process is organized in three working groups: Working
Group I examines geophysical aspects of the climate system and climate change; Working Group
II examines vulnerability of socio-economic and natural systems to climate change, consequences,
and adaptation options; and Working Group III examines options for limiting greenhouse gas
emissions and mitigating climate change in other ways.
The IPCC also includes a ’Task Force on National Greenhouse Gas Inventories’. In the
judgment of most observers, the work on the Assessment Reports proceeds largely according to
scientific criteria of due diligence. However, the synthesis work and summaries for policy-makers
are also exposed to political influence because the Panel, which is composed of government
delegates from all member countries, ultimately decides on their adoption. Hence the wording
in the summary for policy-makers (but not the content of the detailed reports by the working
groups) is subject to some political negotiation. However, governments have thus far hesitated
to modify, for political purposes, the main conclusions drawn from scientific assessments.
2.3.3 FCCC and Kyoto Protocol
The United Nations Framework Convention on Climate Change (FCCC) was formally adopted at
the Rio, or Earth Summit in 1992 (UN Conference on Environment and Development, UNCED).
Its aim is the
’stabilization of greenhouse gas concentrations in the atmosphere at a level that
would prevent dangerous anthropogenic interference with the climate system. Such a
level should be achieved within a time-frame sufficient to allow ecosystems to adapt
naturally to climate change, to ensure that food production is not threatened and to
enable economic development to proceed in a sustainable manner.’ (Art. 2, FCCC)
This global treaty does not set forth mandatory emission constraints, overall or for specific
countries. Yet it has established the basic legal structure for future agreements and has defined,
at a very general level, the goals to be achieved in climate policy. The FCCC entered into force
in March 1994 and, as of late 2009, has attracted 192 member countries. Supported by the
20
IPCC Task Force on National Greenhouse Gas Inventories and the FCCC secretariat in Bonn,
Germany, the FCCC members have established national inventories of greenhouse gas (GHG)
emissions and removals. These inventories served to identify the 1990 emission levels that are
the benchmarks for emission reduction obligations under the Kyoto Protocol. The so-called
Annex I countries (OECD countries and transition economies) are committed to periodically
update these inventories.
Since 1995 the member countries of the FCCC have met each year in Conferences of the Parties
(COP). These meetings serve to review the implementation of the agreement and negotiate
follow-up agreements. The most important outcome thus far is the Kyoto Protocol (KP). This
Protocol was adopted in December 1997 and entered into force in February 2005 (after 55
countries representing 55% of global CO2 emissions in 1990 had ratified). The Protocol has (as
of late 2009) 187 countries that have ratified it. The most important holdouts are the United
States, Afghanistan, Somalia, and Taiwan.
Under the KP, industrialized countries (Annex I countries) have undertaken to reduce six
GHGs (carbon dioxide, methane, nitrous oxide, sulphur hexafluoride, hydrofluorocarbons, and
perfluorocarbons4 ), of which carbon dioxide and methane are the most important in terms of
the size of their greenhouse effect. 39 of 40 potential Annex I countries (except the USA) have
ratified, and 34 countries have committed to emission reductions – 5 of the KP Annex I members
are allowed to maintain or increase their 1990 emission levels (e.g. Russia, Australia, Iceland).
The European Union is treated as a ’bubble’: it received a single target and then allocated
emission rights to its member countries. Total reductions are supposed to be in the order of
5.2% by 2012, from the 1990 level (each GHG is weighed by its global warming potential). The
KP also provides for ’flexible mechanisms’, such as emissions trading, the clean development
mechanism, and joint implementation. The purpose of these economic instruments is to make
GHG emissions cuts more cost-efficient, with the assumption that countries are willing to curb
their emissions more if doing so is cheaper. Monitoring of compliance relies primarily on annual
reports of GHG emissions by Annex I countries and (on a voluntary basis) by other countries.
Most observers of the KP agree that many Annex I countries are currently experiencing
difficulties in meeting their emissions targets domestically and are likely to make use of flexible
mechanisms in order to be able to meet their legal obligations. Also, the USA, which has not
ratified the KP but could still implement its Kyoto targets voluntarily, has increased its emissions
4
CO2 , CH4 , N2O, HFC, PFC, SF6.
Climate Change Governance
21
quite dramatically (its Kyoto target was -7% relative to the 1990 level). Moreover, negotiations
on a follow-up agreement to the KP, which ends in 2012, have thus far failed, most recently in
Copenhagen. A recent study (Rogelj et al. 2010) suggest that, even if all unilateral reduction
pledges made at Copenhagen were implemented, the probability of limiting global warming to
3˚C by 2100 would only be 50%, while global emissions would increase by 20% over 2010 levels.
If emissions were cut by 50% by 2050 the probability of exceeding 2˚C would still be 50%.
2.4 Why is International Cooperation Difficult?
GHG reduction targets set forth in existing governance arrangements are still far from what
would be required to limit temperature increases to 2˚C. It remains unclear whether those
rather non-ambitious targets will be reached by 2012, and even greater uncertainty exists with
respect to unilateral pledges for the post-2012 period and the prospects for formal, follow-up
international agreements.
At the most general level, namely the global political and economic system, climate change
mitigation is difficult because it has the character of a global public good. Moreover, there is
considerable disagreement over the costs and benefits of GHG mitigation. We discuss these two
problems in this section. The following section (section 5) sheds light on additional challenges to
effective governance that emanate from country characteristics, such as differences in economic
conditions and political institutions, which make some countries more reluctant to cooperate
than others.
2.4.1 Global Public Goods and the Free-Rider Problem
Climate change mitigation is one of the most typical public goods problems imaginable. Efforts
to reduce GHGs correspond by and large to an N-actor prisoner’s dilemma, which is similar to
the tragedy of the commons (Hardin 1968). The Earth has one indivisible atmosphere that can
be used as a sink for GHG emissions worldwide; i.e. it is a common pool resource characterized
by open access and rivalry in consumption (Ostrom et al. 1994). By implication, GHG reductions
by any country generate costs and benefits (in terms of avoided damages) for that country, but
also benefits for other countries (positive externalities). Because in the climate case positive
externalities from emission cuts are quite large in relation to national benefits, international
cooperation is necessary, but countries are reluctant to do so.
22
For example, if Italy or Ireland were to cut its GHG emissions by 20%, but no other country
did the same, this reduction would probably create some local benefits of a non-climatic nature
(e.g. less local air pollution, more technological innovation) and some, albeit minuscule climatic
benefits. But the overall net benefit for the respective country would probably be very small and
could even be negative. Assuming that countries follow a rationalist, interest-based logic when
deciding on their climate policy, they will not implement any major unilateral GHG emission
cuts unless other countries credibly commit to a similar policy (e.g. Sandler 1997; Barrett 2003;
Mitchell 2006). The Kyoto Protocol reflects this problem very clearly: it requires ratification by
55 countries representing 55% of global emissions before entry into force. This clause protects
countries from getting ’caught up’ in legal obligations to reduce emissions if they ratify early
but other countries (and major emitters in particular) end up not joining.
In essence, global governance in climate change policy uses mechanisms of reciprocity to prevent free riding on positive externalities. Reciprocity implies that each country exchanges its
commitment to reduce emissions against similar commitments by other countries. The international climate change regime described above is quite typical in this regard. It offers an arena for
step-by-step cooperation and exchanges of information (monitoring). As is the case with most
global governance systems, the climate regime has no centralized enforcement mechanisms but
relies on monitoring instruments inside and outside the regime to identify non-complying countries, and on decentralized enforcement in the form of political and economic pressure imposed
by governments and other actors on non-complying countries.
Several cases of successful international cooperation for the provision of global public goods,
such as cooperation to protect the stratospheric ozone layer, demonstrate that problems of this
type can be solved. Hence the public goods character of climate change mitigation alone cannot
explain why climate policy is progressing much slower than say cooperation in the ozone case
(Barrett 2003; Oye & Maxwell 1995). The similarities and differences between the two cases,
some of which are summarized in Table 2.1, suggest that we need to account for costs and
benefits as well.
Table 2.1 suggests that the costs of climate change mitigation are much higher than the costs
of dealing with the stratospheric ozone problem. However, as discussed in the next subsection,
mitigation costs and benefits remain contested. This circumstance, together with the global
public goods character of the climate change issue, makes global cooperation difficult.
Climate Change Governance
23
Table 2.1: Stratospheric Ozone and Climate Cooperation
Stratospheric ozone regime
Climate change regime
Type of problem
Framework convention, followed
by protocols
Framework convention, followed
by protocols
Costs (damage) if
mitigation effort
fails
Short to long term, clearly identifiable damages (e.g. higher skin
cancer rates, crop damage); developing and developed countries
are approx. equally vulnerable to
damages
Medium to long term damage
that is difficult to quantify; developing countries are more vulnerable to damage
Mitigation costs
Less than $10 billion globally;
substitutes for ozone depleting
chemicals (ODS) are available;
phase-out costs are spread over
around 2-3 decades; costs are
spread across a vast number of
consumers, but per capita costs
of more expensive substitutes for
ODS are very small
Several hundred billion $; substitutes are partly available; phaseout costs are spread over many
decades; costs are spread across a
vast number of consumers, with
rather high per capita costs
24
2.4.2 The Contested Economics of Climate Change Mitigation
As noted above, uncertainty concerning the costs of failing to reduce GHG emissions (and,
conversely, the benefits of GHG reductions) remains rather high. The same holds for the costs
of reducing GHG emissions. Uncertainty with respect to benefits and costs combines to create
serious difficulties in estimating the net benefits (benefits minus costs) of reducing emissions. The
IPCC and the Stern Review arrive at a favorable net benefit assessment because they use rather
pessimistic assumptions about climate change related damages, rather optimistic assumptions
about mitigation costs, and a low discount rate. For instance, Watson (2009) argues that
• ’do nothing’ would result in an average annual loss of 5-20% of global GDP now and
forever due to a 50% chance of exceeding a 5˚C temperature increase by 2100 (relative to
pre-industrial levels);
• moving to a 550ppm trajectory would result in costs of 1% of global GDP in 2050, with a
50% change of exceeding a 3˚C temperature rise;
• moving to a 450ppm trajectory would cost about 3% of GDP in 2050 and would offer a
50% change of remaining below a 2˚C temperature increase.
Net benefit estimates by other social scientists, e.g. Nordhaus (2010) and Tol & Yohe (2006)
are more pessimistic. The main reason is that they use higher discount rates, which leads to
lower estimates of the present value of (discounted) future climate change-related damages and
lower costs of mitigation the more mitigation is postponed. Based on their respective assessment,
the IPCC and the Stern Review arrive at very different conclusions compared to Nordhaus, Tol,
and some other economists. While the former point to large net benefits of starting early with
major GHGs reductions, the latter advocate starting slowly and implementing deep cuts only
in the long run. Note, however, that none of these studies denies that human-induced climate
change exists and poses very serious problems, and that major emission cuts are necessary. But
they disagree on when emissions should be cut by how much in order to generate net benefits
to present-day decision-makers.
It is easy to see why many policy-makers are more attracted to the Nordhaus-type estimates
than the Stern-type estimates. Policies that incur rather high costs in the short-term and
uncertain, even though potentially high benefits in the medium to long run are inherently less
attractive than policies that generate a ’return on investment’ within the near future.
Climate Change Governance
25
2.5 Measuring and Explaining Variation in National Contributions to the Global
Public Good
Most research on climate change governance concentrates on describing and explaining the
climate change policies of individual countries or regions. Rather few studies focus explicitly
on explaining observed variation across a large number of countries in national contributions to
climate change mitigation. The following section discusses how national contributions to global
climate change mitigation have been measured. The subsequent section deals with the main
explanations in the existing literature.
2.5.1 Measuring Variation in Contributions to the Public Good
To explain differences in climate change mitigation efforts across countries, we need, first of
all, indicators that provide accurate und useful information on various dimensions of national
mitigation efforts. These indicators must cover two principal dimensions: policy outputs and
policy outcomes. Policy outputs include laws, regulations and various other types of policy
measures that can tell us how serious or ambitious a government is about climate change mitigation. Policy outcomes are phenomena located either at the interface of human behavior and
the environment, such as emissions, or environmental conditions mitigation policies are aiming
at, such as GHG concentrations in the atmosphere.
Existing research focuses mainly either on policy output or policy outcome, though some
composite indicators have recently been developed to bundle information on different facets
of climate change mitigation efforts. Economists have concentrated mainly on environmental
outcomes, such as emissions, and usually explain those with economic factors (Holtz-Eakin &
Selden 1995). The political science and international relations literature, in contrast, pays more
attention to environmental policy-making and thus also policy output, for instance international
political commitments (e.g. Von Stein 2008; Bernauer, Kalbhenn, Koubi & Spilker 2010; Sprinz
& Vaahtoranta 1994). Very few studies offer a direct comparison of differences in climate change
policy output and outcomes (see Congleton 1992; Baettig & Bernauer 2009; Ward 2008).
Environmental performance indicators have become quite popular in research on sustainable
development in recent years (e.g. Singh et al. 2009). Yet, the construction of such composite
indicators is methodologically challenging and their validity is usually contested (e.g. Boehringer
& Jochem 2007; Singh et al. 2009). Freudenberg (2003, 29), for instance, advises researchers to
26
bear the conceptual limits of composite indicators they use in mind and accompany them ’by an
account of their methodological limits and include detailed explanations of the underlying data
set, choice of standardization technique and selection of weighting method ’.
To draw robust inferences about the determinants of cross-national and longitudinal variation
in national climate change mitigation efforts we need data on many countries over longer periods
of time. Such data is readily available for GHG emissions, and existing scientific debates focus
primarily on whether explanatory models should focus on emission levels or on trends, and which
GHGs and sources should be included in policy outcome variables. The largest data gap exists
with respect to policy outputs. Existing large-N data sets for climate change policy outputs are
thus far rather simple in terms of the types of policy output they capture, and they are mostly
cross-sectional (e.g. Baettig & Bernauer 2009; Germanwatch 2010). Finally, large-N data on
climate adaptation efforts does, unfortunately, not yet exist.
2.5.2 Explaining Variation in Mitigation Efforts
The most prominent explanations of variation in mitigation efforts focus on economic factors,
political factors, and risk-related factors.
The Environmental Kuznets Curve and related determinants
Among economists the most popular explanation for differences in environmental behavior across
countries and over time is the Environmental Kuznets Curve (EKC).5 The latter holds that an
inverted u-shaped relationship exists between income and pollution. Grossman & Krueger (1995)
are usually credited for the first empirical test of the EKC (c.f. Dasgupta et al. 2002; Dinda 2004)
in their study of the relationship between pollutants (SO2 and smoke) and income per capita,
where they identify such a relationship.
Recent research argues, however, that economic growth has somewhat more complex effects on
pollution, including GHG emissions. Three types of effects are usually considered: a scale effect,
a composition effect, and a technological effect. Since more economic output due to economic
growth tends to increase pollution and waste, economic growth is assumed to have a negative
scale effect on the environment. The composition effect is argued to have a positive impact
on environmental performance because economies usually develop from (dirtier) manufacturing
towards (cleaner) services industries. As long as the composition effect does not simply lead to
5
See also de Bruyn & Heintz (1998); Dinda (2004).
Climate Change Governance
27
a re-location of dirty production to poorer and less regulated countries, the composition effect
can also reduce global (rather than only local) pollution levels. Economic growth is usually
associated with technological innovations that help replace old technologies with newer and
cleaner ones (technology effect). In addition, growing income is presumably associated with
increasing public demand for environmental protection once a country’s population has satisfied
its basic needs and becomes willing to invest in ’postmaterial’ goods.
While the basic tenets of the EKC may well be plausible, critics argue that it conveys the
message to developing countries that they should ’grow first, then clean up’ (Dasgupta et al. 2002;
Hill & Magnani 2002; Huang et al. 2008). This has obvious implications for the discount rate
(see section 2). There is a lively academic (and also policy) debate on the empirical relevance
of the EKC in general and CO2 and other GHG emissions in partiuclar (e.g. Millimet 2003;
Galeotti et al. 2006). While many studies identify a statistically significant relationship between
income and different local pollutants (notably, SO2, NOx, CO; e.g. Lempert et al. (2009), global
pollutants such as CO2 tend to either increase monotonically with income or have high turning
points (e.g. Holtz-Eakin & Selden 1995; Dinda 2004; Neumayer 2002a). Some studies also point
to a less favourable N-shaped curve to describe the relationship between income and CO2 (e.g.
Galeotti et al. 2006; Dinda 2004). Moreover, Galeotti et al. (2006) find that the inverted ushaped relation for some pollutants exists for OECD countries, but not for other countries.6
One conclusion from this research (c.f. Holtz-Eakin & Selden 1995; Huang et al. 2008) is that
gambling on an automatic reduction of GHG emissions as income grows would be risky and
probably be a mistake.7
Yet another problem with empirical results for the EKC is that they do not take into account
regulatory policies. This implies that it remains hard to tell whether observed decreases in GHG
emissions are due to income, technology, or composition effects, or whether they are caused also
by effects of regulatory policies or other factors (Hill & Magnani 2002, see below). Moreover, it
remains contested to what extent GHG reductions observed in some country are due to ’bad’
composition effects, meaning relocation of GHG-intensive production to pollution havens. The
main long-term problem with ’bad’ composition effects is that they may allow richer countries
to reap the ’low-hanging’ fruits and could eventually leave poor countries at the bottom of the
risk-shifting cascade where beneficial composition effects must be achieved within the respective
country.
6
7
One explanation is that poorer countries are still on the upward slope of the EKC (Lempert et al. 2009).
Huang et al. (2008, 246) argue that an expansion of the Annex I group under the KP is necessary.
28
Effects of the political system
The political system of a country is likely to have implications for climate change mitigation
policy (Fredriksson & Millimet 2004a). Many studies show that democracies tend to be better
providers of environmental quality (e.g. Bernauer & Koubi 2009). Even though democracy offers greater political access also for non-green interests and the median voter may not always
prefer more environmental protection, existing theories expect, on balance, a positive net effect
of democracy on environmental protection. The gist of the argument is that, in democracies,
freedom of information and political rights enable citizens to acquire more information on environmental risks and express their demands more easily vis-a-vis policy-makers. The latter, in
turn, have greater incentives than autocrats to meet citizen demands because they are more dependent on broad public support, for instance in elections (Baettig & Bernauer 2009; Gleditsch
2002; Li & Reuveny 2006). Note that this argument is relative, not absolute. While the environmental performance of democracies may well be bad, the performance of non-democracies is
likely to be even worse.
Whether democracies outperform non-democracies with respect to climate change policies is
largely an empirical question. Existing studies on policy outcomes (usually defined as GHG
emissions) arrive at mixed results. For instance, Gleditsch (2002) find that democracy is associated with lower CO2 emissions. Congleton (1992) finds that democracies emit less methane.
Midlarsky (1998) observes that democarcies emit more CO2 . Li & Reuveny (2006) find that
democracy is associated with less per capita CO2 emissions.
The relationship between democracy and climate policy output appears to be more robust
than the relationship between democracy and climate policy outcomes (e.g. Neumayer 2002b).
Von Stein (2008) observes a positive effect of democracy on climate change treaty participation.
Baettig & Bernauer (2009) compare climate policy output and outcomes side-by-side. They
find that democracies contribute more to the global public good in terms of policy output, i.e.
political commitments, but that the effect on policy outcomes is ambiguous. They describe
this result in terms of a ’word-deeds’ gap, which appears to be larger in democracies than in
autocracies. Reasons include the fact that mitigation efforts have started only a few years ago,
and that, relative to local public goods, such as air pollution, there is a stronger free-rider
problem. One major research gap in this research area is whether the positive democracy effect
is driven more by the demand or the supply side (e.g. Ward 2008; Baettig & Bernauer 2009).
Climate Change Governance
29
Researchers have recently started to disaggregate democracy and examine the implications of
different types of democracy, such as presidential vs. parliamentary systems, consensus democracies vs. other types, etc.. Ward (2008), for instance, finds that presidential democracies
perform worse than parliamentary democracies in environmental terms. Fredriksson & Millimet
(2004b) observe that governments set stricter environmental policies under proportional than
under majoritarian systems.
Effects of the natural system
The natural (i.e. geophysical climate) system may, on the one hand, influence the emissions
behavior of countries. On the other hand, it may also affect their vulnerability to climatic
changes and thus their willingness to contribute to the global public good.
Neumayer (2002a), for instance, examines natural factors such as climatic conditions, the
availability of renewable and fossil fuel resources, and transportation requirements. He finds
that these factors have significant effects on cross-country differences in CO2 emissions, though
the income level remains the most important determinant. Aldy (2005) examines U.S. states
and observes that climatic conditions and coal endowments are positively related to per capita
CO2 emissions.
Natural system characteristics may also contribute to variation in climate risk exposure, which
in turn could affect countries’ willingness to commit to climate change mitigation. Sprinz &
Vaahtoranta (1994), for example, argue that countries facing greater vulnerability and lower
costs of cooperation are more likely to commit to stronger international environmental policies.
However, empirical research on climate policy has thus far not been able to identify such a
vulnerability effect. For instance, Baettig & Bernauer (2009) do not find any evidence that
climate risk exposure has a positive effect on policy output or policy outcomes. Their analysis
uses a climate change risk exposure index (Baettig et al. 2007) and several other indicators for
risk exposure. One potential explanation for the absence of a positive vulnerability-cooperation
effect is that the available scientific information has not yet spurred sufficient public demand
for risk mitigation. Another explanation is that the most vulnerable countries may have greater
incentives to invest in climate adaptation, which is a national, ’private’ good, rather than mitigation, which is a global public good associated with positive externalities; or, they may be poor
countries that are unable to invest in either adaptation or mitigation.
30
2.6 Alternative Forms of Climate Change Governance: Local Dynamics in Federal
Systems
Mitigation of climate change through effective global treaties to which all countries adhere has
proven very difficult. In this section we look at other forms of governance that have emerged
out of this conundrum. The focus is on sub-national climate change governance.8
Local climate policy-making is particularly interesting in the case of federal political systems.
One noteworthy example is the United States, a typical federal state that is also important
because it accounts for around 25% of global GHG emissions, but has thus far refused to ratify
the Kyoto Protocol. The absence of federal laws and regulations on GHG emissions has led to
a plethora of state- and city-level initiatives over the past few years.
As of June 2010, 32 U.S. states have adopted a climate action plan, 21 states have adopted
GHG emission targets, and 12 states have adaptation plans (Pew Center 2010). Furthermore, the
first mandatory cap-and-trade program for CO2 in the U.S. started in 2009 for the ten member
states of the Regional Greenhouse Gas Initiative (RGGI). Similar regional initiatives have been
initiated in the Western U.S. states as well as in the Midwest (Pew Center 2010). At the level
of cities, the U.S. Conference of Mayors Climate Protection Agreement (MCPA) is the largest
agreement. It involves more than one thousand U.S. cities. It was initiated in February 2005 by
the then Seattle mayor Greg Nickels. Yet another local initiative is the International Council for
Local Environmental Initiatives (ICLEI)’s Cities for Climate Protection (CCP) program. The
ICLEI is an international initiative that involves around 600 cities worldwide. It started in 1991
and has also helped generate political support for reducing local GHG emission in U.S. cities
(Betsill 2001).
Such local and regional activities are interesting from an academic viewpoint. But they also
beg the question of whether bottom-up activity can substitute for absent national level climate
policy. Lutsey & Sperling (2008) examine the effects of decentralized climate change policies in
U.S. states and cities on GHG emissions, exploring the development of emissions based on current
inventories and chosen sub-national policies. They argue that ’efforts of states and cities are so
pervasive at this point that future federal policy will benefit by adopting the most popular and best
functioning GHG mitigation programs [...]’ (Lutsey & Sperling 2008, 683). Selin & VanDeveer
(2007) are less optimistic about the emission-reducing effects of local climate initiatives. But
8
For reasons of space, we cannot discuss yet other forms of governance in climate policy, such as public-private
partnerships and civil society involvement.
Climate Change Governance
31
they also stress the importance of such programs because they allow policy-makers to ’[...] see
which of the many available policy options are gaining support in the public and private spheres’
(Selin & VanDeveer 2007, 22) and thereby are most likely to influence future federal policy
development. Tang et al. (2010) study 40 local climate change action plans in U.S. cities. They
find that, although these plans reflect a high level of environmental awareness, they have only
limited effects on emissions.
While existing research has not yet been able to demonstrate the effectiveness of local and regional initiatives in terms of reducing GHG emissions, recent research offers interesting insights
into the factors that affect the dynamics of such initiatives. Employing event history analysis,
Vasi (2006) finds that adoption of the Cities for Climate Protection (CCP) program is driven
by spatial or cultural proximity to earlier adopters, and that organizational embeddedness in
transnational frameworks also fosters participation. With respect to CCP county-level participation patterns, Brody et al. (2008) observe that counties with landscape characteristics of
high risk, low stress, and high opportunity are more likely to join the CCP campaign. Schaffer
(2010) examines county-level participation patterns for the Mayor’s Climate Protection Agreement. She highlights the importance of natural system characteristics, such as whether the
county is a coastal county, as well as political preferences of the inhabitants to determine where
participation rates in this initiative are highest. Lee (2009b) uses a multilevel setting to analyze
what cities participate internationally in the CCP and other networks, such as C40. Controlling
for city and country-level variables, he finds that cities’ position in the global economy and
transportation hub characteristics significantly influence participation in these networks.
By and large, existing studies suggest that local and regional climate policy initiatives have
gained ground in recent years, particularly in federal political systems. There is little evidence
that such initiatives can substitute for slow progress in adopting and implementing ambitious
mitigation policy at national and international levels. Nonetheless, evidence from the U.S. and
other countries indicates that such activities can serve as policy-experiments in trying to find
efficient mitigation options and changing the ’mind-set’ of business actors and citizens in a
climate-friendly direction.
2.7 Normative Issues
We end this chapter with a discussion of two normative issues that have received particular
attention in research on climate change governance (e.g. Paterson 2001; Allen 2003; Klinsky &
32
Dowlatabadi 2009; Vanderheiden 2008; Posner & Sunstein 2007; Johnson 2009). One concerns
intergenerational fairness, the other concerns the fair division of responsibilities for mitigation
and adaptation.
As discussed above, there is strong disagreement in the scientific literature on whether and by
how much future costs and benefits of climate change and its mitigation should be discounted
when deciding today how much to invest in solving the problem.
Some scientists (e.g. Nordhaus 2010; Lempert et al. 2009) view the climate problem as one of
many problems policy-makers need to deal with in parallel; so they need to weigh the costs and
benefits and decide in which policy to invest more when and where. In their view, there is, from
an economic perspective, nothing that makes e.g. investment in combating infectious diseases
or maintaining law and order inherently different from investment in solving an environmental
problem like climate change. They also assume that future generations will be wealthier and
have more technological means to mitigate climatic changes at lower cost. Hence they apply a
higher discount rate to mitigation costs and also assume that damage from unmitigated climate
change is less costly as it occurs in the future.
Other scientists regard this position as unethical because it burdens future generations with
an environmental problem. Many of them consider that only a small or even no discount rate is
the appropriate choice, and the issue should be viewed from a ’rights’ perspective (e.g. Collier
2010).
The policy implications of these two contrasting views are quite obvious: the former position
is associated with proposals to start mitigation very slowly and invest more over the mediumto long-term; the latter position generates proposals to ’front-load’ mitigation efforts, i.e., start
early and invest a lot in the short- to medium-term. However, most analysts agree that there is
no scientific solution that could identify the ’correct’ discount rate.
Yet another issue that has attracted considerable attention in social sciences research is the
question of fair burden sharing in mitigation and adaptation. Since past and current emissions
have a greenhouse effect over many years to come and predictions of future emissions vary greatly,
researchers have used complex models to calculate how much particular countries and regions
contribute to global warming. A paper by Den Elzen et al. (2005), for instance, shows that
responsibilities of specific regions can be calculated, though these responsibilities differ somewhat
depending on the time period of emissions, the mix of GHG, climate impact indicators, and
climate models. Such calculations indicate what share in temperature increases can be attributed
Climate Change Governance
33
to specific countries or regions. Den Elzen et al. (2005) find that the average contributions
to the global mean surface temperature increase in the year 2000 amount to around 40% for
OECD countries, 14% for Eastern Europe and the Former Soviet Union, 24% for Asia, and 22%
for Africa and Latin America. The OECD share decreases when later attribution periods are
selected and increases for industrial latecomers, such as Asia. Including land-use related GHG
emissions tends to reduce the OECD share at the expense of other regions. Other authors, e.g.
Boehringer & Helm (2008) have sought to come up with specific modes of fair division based
on compensation schemes. Again, such calculations show that normative assumptions (e.g. how
industrial latecomers should be treated, which GHG should be considered) play an important role
in establishing responsibilities for and, consequently, also burden-sharing formulas for climate
change mitigation.
Finally, another line of research assumes that climate-related damage is unavoidable even
with the most ambitious mitigation efforts. Accordingly, it asks who should pay for adaptation
measures. Dellink et al. (2009) use two principles, historical responsibility for radiative forcing
and capacity to pay, to estimate the shares of individual countries in the financial burden. The
results turn out to be more sensitive with respect to variation in capacity to pay assumptions
than model input concerning historical responsibility. The authors assume adaptation costs of
USD 100 billion per year and conclude that Annex I countries should contribute around USD
65-70 billion, which amounts to around USD 43-82 per capita and year in Annex I countries and
USD 1-21 in non-Annex I countries.
35
3 Nature, Nurture or Neighbors? Testing Participation Patterns in
Voluntary Initiatives in U.S. Counties
Lena Maria Schaffer
3.1 Introduction
The global scientific community, as organized through the Intergovernmental Panel on Climate
Change (IPCC), is largely in agreement that anthropogenic causes, emissions from the combustion of fossil fuels in particular, are to blame for increasing temperatures in many areas of the
world and on global average. It also agrees that the economic and social effects of human-induced
climate change are grave and potentially disastrous, notably for societies with insufficient means
to adapt (Stern et al. 2006; Stern 2008; IPCC 2007a). While international treaties to reduce CO2
and other greenhouse gases (GHG) – such as the Framework Convention on Climate Change
(FCCC) and the Kyoto Protocol – have been agreed on by a majority of countries around the
world (IPCC 2007), several large emitters of GHGs such as the U.S. are still reluctant to support
global efforts in climate change policy.
Regardless of its reasons why, the 2005 decision of the U.S. federal government not to participate in the effort to cut GHG emissions can be regarded as a landmark decision. It impacted
on the international community, but more importantly triggered action within U.S. subnational
units that wanted to proceed along with global efforts to curb GHG emissions. Now five years
on, and with a different administration, while there continues to be no comprehensive national
climate change legislation1 , state and local governments have created many different initiatives
and laws that explicitly aim at installing climate change policies in their jurisdictions. One such
initiative, that started in February 2005 and has been promoted in many cities all over the U.S.,
1
The topic has been given considerably more attention within the Obama Administration and gradual steps
have been made in the climate change policy area. An example of this is the Supreme Court ruling that the
Clean Air Act could be applied to GHGs, giving the Environmental Protection Agency (EPA) more leeway to
set minimum standards on energy efficiency. Despite the fact that the energy bill that came out of the Senate
in July 2010 failed to include the expected cap-and-trade scheme, and now proposes little more than subsidies
for home insulation and trucks that run on natural gas (Economist 2010)
36
is the Mayors Climate Protection Agreement (MCPA). Seattle Mayor Greg Nickels launched the
initiative to advance the goals of the Kyoto Protocol through leadership and action by American
cities. Joining the initiative constitutes a voluntary act by the local governments that sign the
agreement, and thereby commit their cities to reduce emissions to seven percent below 1990
levels by 2012 (Mayors Climate Protection Center 2010). Over 1040 cities joined this initiative
by Fall 2010. It is a prime example for subnational environmental policy formation. However,
as can be seen in figure 3.1, participation in this initiative is not distributed equally throughout
the United States. The map shows the MCPA participation rate of cities within all counties
of the 48 contiguous U.S. states. Counties where no city has signed the agreement are shown
in red tones in figure 3.1, whereas different shades of yellow to green reflect participation rates
from 1 % up to 100%. Counties where participation reaches nearly 100% of the cities that could
sign the agreement are depicted in dark green on the map.
Figure 3.1: Counties in the U.S. according to their participation in the MCPA; darker shades
indicate a higher percentage
The idea that subnational units engage themselves in new policy areas is neither new to American policy making nor limited to the field of environmental politics. Quite to the contrary, the
notion that states 2 try out innovative policies that they deem necessary within their jurisdiction
became famous through Judge Brandeis’ dissent in the New State Ice Co. vs. Liebmann Case
(1932, 310). He stated that ’There must be power in the States and the Nation to remould,
through experimentation, our economic practices and institutions to meet changing social and
economic needs’ and went on to claim that ’[i]t is one of the happy incidents of the federal system
2
This can also be extended to other subnational units such as county or city governments
Nature, Nurture or Neighbors?
37
that a single courageous State may, if its citizens choose, serve as a laboratory; and try novel
social and economic experiments without risk to the rest of the country’. This potential of states
to act as laboratories of democracy can be seen as a desired outcome of federalism (Shipan &
Volden 2006), or an ’all-purpose defense of the legitimacy of states pursuing divergent policies’
(Levy 2007). However, the general desirability of states experimenting with climate change policy is not the focus of this paper. In fact, this paper relates more to Brandeis’ comments on the
changing social and economic needs that lead to different economic practices or institutions in
different states or parts of the federation. Proceeding from the observed outcome of voluntary
subnational climate change policies and their uneven distribution over the United States, I seek
to find out which social, economic, or other needs lead to the creation of this pattern.
Therefore, this paper proposes to explain participation patterns in the MCPA at the countylevel throughout the U.S. by factors regarding the natural (Nature), socio-economic and political (Nurture) characteristics, and their geographical surroundings (Neighbors). It is hypothesized that voluntary participation in climate change mitigation efforts are highest when
the county is at risk of being adversely affected by climate change, when there are favorable
socio-economic and political conditions present, and when the surrounding counties are also
strong participators.
I test these claims for the MCPA participatory behavior of cities in counties within the 48
contiguous states and thus contribute to the literature on local voluntary climate change initiatives in two respects. First, there exists a gap in the literature on local voluntary climate change
initiatives with respect to comprehensive hypothesis testing within a large-N framework. Studies
dealing with local climate change initiatives overwhelmingly rely on descriptive or case study
evidence (Betsill 2001; Betsill & Bulkeley 2006; Engel & Orbach 2008; Engel 2009; Kousky &
Schneider 2003; Schreurs 2008; Schreurs & Epstein 2007; Urpelainen 2009). Evidence from the
large-N analysis conducted in this paper represents an important step forward in understanding
the motivations behind voluntary climate change actions. Moreover, the few studies that explore
participation within a quantitative framework (Brody et al. 2008; Lee 2009a; Vasi 2006; Zahran,
Grover, Brody & Vedlitz 2008) have thus far focused on explaining participation in the Cities
for Climate Protection (CCP) program initiated by the ICLEI - the International Council for
Local Environmental Initiatives. This program started in 1991 and was the first of its kind to
introduce the topic of climate change at the city level (Betsill 2001; Betsill & Bulkeley 2006).
While participation within the CCP is still at around 160 cities throughout the U.S., 1040 cities
38
have signed the MCPA during the past five years. Due to this difference in scope as well as the
greater saliency of the climate change issue in the past few years, participation in the MCPA
fundamentally differs from the CCP and calls for a more detailed study. So far only few authors
have focused on this initiative (Engel & Orbach 2008; Engel 2009; Schreurs & Epstein 2007;
Warden 2007; Tang et al. 2010) and the second contribution therefore concerns the topic under
study, the Mayors Climate Protection Agreement.
By analyzing the determinants of participation in the Mayors’ Climate Protection Agreement
throughout all 48 states, I aim at closing important gaps in the literature. From my OLS
and spatial analyses, I find that as far as natural system characteristics are concerned, being
a coastal county seems to drive signatory behavior. For socio-economic characteristics, bad
economic conditions in a county significantly decrease the relative number of signatories, whereas
higher democratic vote shares and a well-educated population significantly increase the relative
number of signatories. By controlling for risks related to climate change and socio-economic
incentives and disincentives of counties to sign voluntary agreements, most of the spatial pattern
is explained. Alternative model specifications and robustness checks confirm these results and
show other potential ways to model the research question at hand.
The paper is structured as follows. First, a theoretical overview of the determinants for
participation in such a voluntary climate change initiative is given. Then each of the three groups
of factors (nature, nurture, and neighbors) and deduced testable hypotheses are introduced. In
the second part, the data and operationalizations of the main concepts are provided. The results
from the analysis and robustness checks are shown in the third part. The paper concludes with
a discussion of the results and suggests further extensions.
3.2 Determinants of Participation in Voluntary Climate Change Initiatives
Policies get formulated and implemented to solve given societal problems as long as their (political) costs do not outweigh their benefits to office- and/or policy-seeking politicians. Thus,
participation in voluntary initiatives by any given jurisdiction results from specific incentives
and disincentives to join (Meyer & Konisky 2007). Incentives are favorable conditions present
within the jurisdiction that will minimize the cost or maximize the benefit from adopting such
voluntary measures. Factors are disincentives or stressors if they increase costs for the locality
due to adopting voluntary measures.
Nature, Nurture or Neighbors?
39
I argue that there are three dimensions that need to be taken into account when talking
about the voluntary adoption of climate change policies. First and foremost, climate change
is an environmental problem that might cause serious damage for those that are likely to be
affected by it. Thus, the first factor for explaining participation in voluntary climate change
initiatives that should be considered is natural risks (Nature) from climate change associated
with a locality. Secondly, combating climate change and its implications might incur costs
for a locality (e.g., from curbing CO2 emissions), but might also have beneficial health effects
(Jacobson 2010). Depending on such socio-economic and political conditions (Nurture), elected
representatives will deliberate about whether to engage in voluntary climate change initiatives
or not. Thirdly, since climate change policies are not yet common practice in the localities’
toolbox, learning from or emulating other localities - as suggested by Brandeis’ dissent earlier
- might take place. I therefore argue that one also has to pay attention to what neighboring
localities (Neighbors) are doing on the issue to explain participation patterns.
3.2.1 Natural System Factors (Nature)
Recent observations of global climate have shown that climate change is unequivocal (IPCC
2007a). However, the consequences of global warming, such as rising sea levels, do not impact in
a similar way across the globe. Whereas nations such as the Maldives might cease to exist due
to further rises of the Indian Ocean, in other regions warmer climates could lead to increased
crop yields and agricultural productivity, as would be the case in many northern European and
North American countries. This considerable disparity in global climate change effects can also
be observed within such a large country as the United States, where climatic conditions are very
different throughout, with cold and dry conditions in the Northeast compared to very hot and
dry conditions in the Southwest. The map in figure 3.2 shows the IECC climate zones within
the United States, ranging from 1-7, with 1 as very hot and moist, and 7 as very cold.
Possible scenarios of the impact of climate change for the United States include more intense
hurricanes with corresponding increases in wind, rain, and storm surges, as well as drier conditions in the Southwest (Karl et al. 2009). Reduced precipitation and increased evaporation
might lead to an increase in droughts in these regions. Moreover, sea level rises due to global
warming are an especially urgent topic at the Gulf and East Coast of the U.S.. Karl et al.
(2009) estimate that a global increase in sea levels of 2 feet would result in a relative rise of 2.3
feet in New York and about 3.5 feet in Galveston, TX, but only about 1 foot in Neah Bay in
40
Map of DOE’s Proposed Climate Zones
Dry (B)
Moist (A)
Marine (C)
7
6
4
6
5
5
4
3
All of Alaska in Zone 7
except for the following
Boroughs in Zone 8:
Bethel
Dellingham
Fairbanks N. Star
Nome
North Slope
Northwest Arctic
Southeast Fairbanks
Wade Hampton
Yukon-Koyukuk
Warm-Humid
Below White Line
3
2
2
2
Zone 1 includes
Hawaii, Guam,
Puerto Rico,
and the Virgin Islands
1
March 24, 2003
Figure 3.2: Climate zones by county for the U.S.: from 1(very hot, moist) to 7(very cold) (Source
Department of Energy)
Washington state. Estimates of damages from rising sea levels come to the conclusion that in
the Gulf coast area alone, about 2700 miles of major roadway, and 246 miles of freight rail lines
could be permanently flooded within the next 50-100 years(Kafalenos et al. 2008). Moreover,
10 major ports are located in the area.
However, as stated above, different regions are affected differently by climate change. For
example, figure 3.3 indicates that summers in the two Midwestern states of Illinois and Michigan
are expected to feel progressively more like summers that are currently experienced in states
south and west of those states (Hayhoe et al. 2009). For agriculture and crop yields, changes
might not be as severe and harvest might even benefit from warmer temperatures. Risks and
opportunities from global warming are spread unevenly throughout the U.S. (Mendelsohn 2001),
and I thus expect participation in voluntary climate change initiatives to also vary.
When it comes to environmental policies generally, it is assumed that policy-makers draft
policies or propose actions in accordance with existing problems within their jurisdiction (Daley
& Garand 2005; Ringquist 1994; Potoski & Woods 2002), or in expectation of future adverse
effects if such a policy is not enacted. Studying local wetlands bylaws under the Massachusetts
Wetlands Protection Act Meyer & Konisky (2007) find strong evidence that environmental need
plays an important role. Brody et al. (2008)’s study on the determinants of CCP also shows
that counties with a higher risk of climate change impacts are more likely to join the CCP
network. Accordingly, I assume that in counties that are more likely to be seriously affected by
Nature, Nurture or Neighbors?
41
Figure 3.3: Model projections of summer average temperature and precipitation changes
in Illinois and Michigan for midcentury (2040-2059), and end-of-century (20802099)(Source (Hayhoe et al. 2009)
rising sea levels, severe droughts, floods or hurricanes, participation rates in voluntary climate
change agreements will be higher compared to counties that are not affected or those that might
even benefit from a changed climate. Thus, the severity of the problem in the locality at risk is
supposed to influence participation in voluntary initiatives.
Hypothesis 1: Participation in voluntary climate change agreements is higher the
more the county is exposed to the effects of climate change.
3.2.2 Socio-Economic and Political Factors (Nurture)
In deciding whether to engage themselves voluntarily in climate change policies, politicians are
assumed to not only take the environmental need or appropriateness of a solution into account.
42
Rather, as office-seeking politicians (Mueller 2003), I expect policy-makers to evaluate whether
there is support for such a solution to be gained or obstacles to be faced within their constituencies (Oates & Schwab 1988)3 . Structural conditions within the constituencies determine
whether demand exists for such policies. Hence, if structural conditions are favorable (incentives) within the jurisdiction, more local governments are expected to participate in the MCPA.
Which composition of a constituency is more likely to support climate change policies, and what
structural conditions lead to a lower probability of voluntary climate change policy adoption?
The following two paragraphs will deal with these questions and deduce testable hypotheses
from them.
Incentives for participation in MCPA
In the literature on people’s environmental attitudes, socio-economic factors play a central role
(Dunlap & Mertig 1992). In their 2002 study, Theodori & Luloff (2002, 471) put it quite comprehensively: ’young individuals, the more highly educated, people with higher incomes, and those
with liberal political ideologies were more likely than their opposites to maintain proactive positions on environmental issues’.4 O’Connor et al. (2002) specifically look at the climate change
issue in their 2002 study on 623 individuals in central Pennsylvania. They find a particular strong
effect for education, whereas income and age do not turn out to be significantly related to policy
support for climate change issues. Moreover, Democrats were more likely than Republicans
to support government efforts to reduce emissions, however, as far as voluntary measures were
concerned, the difference between Democrats and Republicans was not significant (O’Connor
et al. 2002). In their study on local commitment within the Cities for Climate Protection (CCP)
Program, Zahran, Brody, Vedlitz & Grover (2008) find that in localities characterized by high
percentages of environmentally concerned and ideologically center-left inhabitants, politicians
selectively benefited from participation in the CCP campaign. Based on these findings in the
literature, effects of income, education, partisanship, and age are expected to matter in different
ways for participation.
3
In a similar way to Oates & Schwab (1988)’s model where regulators set environmental standards at the level
which maximizes the utility of the median voter in the constituency.
4
Although Dunlap & Mertig (1992) concentrates more on air and water pollution issues and Theodori & Luloff
(2002) focus on positions on general environmental issues, I assume that similar socio-demographic factors
matter for the climate change issue.
Nature, Nurture or Neighbors?
Wealth
43
Endowment with resources, or wealth, is a factor that matters in policy-making (Gray
1973). One line of argumentation points to a greater capacity to compensate losers from potential
regulation. As far as environmental policies are concerned, research has shown that with more
wealth there comes additional demand for environmental quality (Beron et al. 2003; Cornes &
Sandler 1996). Due to its high income elasticity, economists have conceptualized environmental
quality as a luxury good (Baumol & Oates 1988). Demand for climate change policies should
therefore be higher in localities with a higher income. More public demand for such policies
translates into less costs and potentially even benefits for policy makers associated with policy
adoption, which in turn should make adoption more likely.
Hypothesis 2: Counties that are better endowed with monetary resources are more
likely to engage in voluntary climate change initiatives.
Education
Education is a major factor explaining individual level environmental attitudes.
Theodori & Luloff (2002) find that better educated individuals were more likely to contribute
money or time to an environmental group, to read environmental magazines, or to vote for or
against a political candidate because of his or her position on the environment. O’Connor et al.
(2002) speak of the ’awareness of consequence’ that comes with education and which will then
translate into the correct interpretation of implications and solutions associated with climate
change. Analogously, I therefore expect a higher number of educated people per locality to
translate into more demand, and also more policy initiative on climate change issues.
Hypothesis 3: Counties with a higher human capital endowment are more likely to
engage in voluntary climate change initiatives.
Partisan orientation Turning to the political component, the ideological dispositions of the
population should generally be taken into account for any policy decision. In the U.S., the
Democrats and their political leaders have been more supportive of efforts to curb greenhouse
gases than the Republicans (O’Connor et al. 2002). In general, I expect the vote patterns in
general elections within the locality to matter for participation, with a higher proportion of
votes for the Democrats leading to a higher demand for voluntary climate change policies.
Hypothesis 4: The higher the proportion of voters for the party representing environmental issues is, the higher the participation in voluntary climate change initiatives within the county.
44
Age
Literature on the connection between age and environmentalism (e.g. Buttel 1979) claims
that environmentalism can be seen as an age- or generation-based phenomenon rather than as an
education- or class-based movement. Hornback (1974) finds that age is inversely associated with
’pro-environmental bias’ and positively associated with ’anti-environmental bias’. At the time
of writing of both studies, this connection between age and environmental attitudes was quite
strong within the United States even when controlling for education. However, this clear-cut
intergenerational gap might have vanished over the years. As Buttel (1979, 251f.) wonders:
’As the cohort of youth that took part in the mushrooming of environmentalism on college campuses and other youthful milieus during the late 1960’s progresses through careers and builds up
attachments to society, community, and family, will their levels of pro-environmentalism wane?
Or as they replace their generally anti-environmental elders in positions of power and authority,
can we thereby expect greater institutional expressions of environmental awareness and concern?’
Overall, it is far from obvious why a generation that actively fought long and hard for more
environmental protection should lose its environmental attitudes as it grows older. That said,
the long-term nature of climate change as an environmental problem could explain a difference
in attitudes. In their lifetimes, older people are less likely to be adversely affected by climate
change than young people, and should therefore not be as concerned.5
Hypothesis 5: The younger the people living within a locality are, the higher is
participation in voluntary climate change initiatives.
Disincentives for participation in MCPA
The factors mentioned above lead to more favorable conditions, with policy-makers accruing less
costs and potentially even reaping benefits for the initiation of voluntary climate change policies.
However, there are also disincentives that make voluntary policies more costly for policy-makers
and potentially suppress action toward more climate friendly policies.
Economic development Regulations on environmental protection inflict costs on companies
within the locality that have to comply with those regulations, and worsen their competitive
position vis-a-vis enterprises in other jurisdictions. Several studies (e.g. Meyer (1995); List et al.
(2003)) have found regulations to have a negative impact on economic growth and development.
5
Related to this, but on a global scale is Sandler (1997)’s work concerning intergenerational public goods provision
in the context of climate change.
Nature, Nurture or Neighbors?
45
Pagoulatos et al. (2004)’s results indicate however, that although more stringent environmental
regulations have a short-run cost, they can be beneficial over time. The rationale behind this
effect is that if strict regulations are already in place, then the risk to businesses facing additional
regulations may be lower. Since there is still great uncertainty about policy instrument choice in
the realm of climate change policies, I assume that in localities experiencing economic distress,
participation in voluntary climate change policies is less likely.
Hypothesis 6: Participation in subnational climate change regulation decreases
with higher economic distress in the county.
Pollution intensity Of all the greenhouse gases (GHGs), Carbon Dioxide (CO2 ) as emitted
by human activity is usually given most attention. Therefore, global mitigation policies as
well as local voluntary efforts, focus on reducing or stabilizing the concentration of CO2 in the
atmosphere (Bernauer & Schaffer 2010). CO2 -intensive industrial production might be faced
with high abatement costs from voluntary action that could make them less competitive. Betsill
(2000) in her four city case studies as well as Zahran, Brody, Vedlitz & Grover (2008) in their
county-level analysis on CCP membership, find that dependence on carbon intensive activities
and industries is a restraint on the development of local climate change policies. I therefore
expect less voluntary action in localities where there are a lot of CO2 -intensive industries.
Hypothesis 7: Participation in subnational climate change regulation decreases
with the importance of CO2 intensive industries in the county.
3.2.3 External Influences (Neighbors)
The last two sections considered factors related to the natural environment and to socio-economic
and political characteristics of localities and linked them to expected participation patterns.
Both explanations assume that the decision to participate in voluntary climate change efforts
and the resulting patterns are solely based on locality-specific attributes. Above, I cited Brandeis’ argument about ’states as policy laboratories’. This argument implies that policy experimentation by a state can lead the way to a solution that other states might want to emulate.
Studies of this kind of diffusion of policy innovations among U.S. states has a long tradition and
has more recently received renewed scholarly interest (Walker 1969; Gray 1973; Berry & Berry
1990; Berry & Baybeck 2005; Baybeck & Huckfeldt 2002; Boehmke & Witmer 2004; Daley &
46
Garand 2005; Shipan & Volden 2006, 2008; Volden 2006). Concerning participation in voluntary
climate change agreements, the laboratories argument would suggest that unit-external influences might also determine participation. As climate change as an issue for subnational units is
a relatively new phenomenon, it is conceivable that localities take into consideration what other
localities are doing on the issue before deciding on whether to participate or not. Since I am
not able to look at individual decision-making processes, but deal with aggregated participation patterns in this paper, I am interested in whether there is explanatory power to be gained
by accounting for spatial associations of localities. Hence, the question posed here is whether
interdependence between units’ participation patterns exists above and beyond their locational
and socio-economic characteristics. To this end, two possible conceptions of interdependence
are introduced. The first, geographical proximity, taps on the notion of closeness, whereas the
second, leaders of innovation, refines the geographical proximity argument to account for the
special role of leaders of policy innovation.
Geographical proximity Localities are more likely to emulate or to learn policies from their
geographic neighbors (Shipan & Volden 2006, 2008; Millimet et al. 2002). This is what we expect
from Tobler (1970)’s first law of geography, which states that: ’Everything is related to everything
else, but near things are more related than distant things’. Spatial dependence should conform to
this basic theorem, and observations that are close to each other should reflect a higher degree of
dependence. Such interdependencies between counties would manifest themselves in reactions
of cities’ participation decisions in one county in view of their peers’ actions in neighboring
counties. As far as strategic interactions between states in the field of environmental policies are
concerned, studies have found evidence for such strategic considerations of other states’ policies
in policy-making (Levinson 2003; Fredriksson & Millimet 2002; Konisky 2007). These strategic
considerations of cities would result in more similar MCPA participation outcomes in neighboring
counties compared to those farther away. However, cities within a county could either be more
likely to emulate the policies of their peers (positive spatial dependence) or specifically avoid
adoption of such voluntary policies (negative spatial dependence). In most of the literature on
environmental regulations or treaties (Beron et al. 2003; Perrin & Bernauer 2010), free-riding
behavior (negative spatial dependence) is expected. However, since the participation and scope
of the agreement is voluntary and can be determined by the respective locality, I rather assume
a positive effect of spatial interdependence.
Nature, Nurture or Neighbors?
47
Hypothesis 8: The higher MCPA participation is in neigboring counties, the higher
is a county’s own MCPA participation
Leaders in innovation High MCPA participation rates may not only arise from participation
rates in neighboring jurisdictions, but may emanate from geographical closeness to one of the
leaders of policy innovation. Such leaders are cities that were among the first movers within the
climate change policy field, and therefore serve as a likely example to be imitated by nearby
localities. The role of leaders or pioneers for the process of policy diffusion was already a central
question in Walker (1969)’s classic study. Jaenicke (2005, 145) claims that leaders legitimize
subsequent adoptions by other jurisdictions by demonstrating the political feasibility of the
innovation.
Leaders of innovation in this study are cities that were among the first movers within the
climate change policy field in the U.S. and therefore serve as a likely example to be imitated
by nearby localities. The then Seattle Mayor Greg Nickels had initiated the Mayor’s Climate
Protection agreement in early 2005 by sending around emails and advertising the initiative on
Seattle’s webpage. The original goal was to mirror the Kyoto Protocol, which became binding
for 141 nations on the 16th February 2005, and have at least 141 Mayors to sign his initiative
to locally commit to reducing CO2 emission. By the June 2005 annual conference of USCOM
in Chicago, this goal was reached and 141 Mayors had signed the MCPA. On this meeting, the
MCPA was formally endorsed as a policy by the United States Conference of Mayors. These
first 141 cities have committed themselves very early in the process and can therefore be seen
as leaders of innovation. Without their swift reaction on Mayor Nickels agreement, momentum
might not have gained during this crucial first months. I therefore expect that cities that are
located near one of those original 141 have learned about the MCPA very early on in the process
and could potentially imitate and learn from those leaders.
Hypothesis 9: The closer a county is to one of the leaders of innovation, the higher
is its participation rate.
3.3 Data
The previous section identified three competing factors of influence and derived hypotheses. To
test the hypotheses derived in the previous section, I use U.S. counties as the unit of analysis.
Although the decision whether or not to participate in the MCPA is made by the city (i.e. the
48
mayor), the county level was chosen because of feasibility reasons. Collecting city-level data for
the 48 contiguous states and all 2700 cities within them was beyond the scope of this project
because in contrast to counties, data is not readily available in spreadsheet form. By aggregating
the cities within each county, I am not able to analyze individual policy-makers’ decisions, but
can comprehensively compare structural factors in all counties throughout the U.S to obtain
first estimates of which factors make participation more likely. In the United States, the U.S.
Census Bureau lists 3140 counties or county-equivalent administrative units in total. I excluded
Hawaii and Alaska and used only counties in the 48 contiguous states for the analysis.6
3.3.1 Dependent Variable
The dependent variable of interest gives information on the proportion of cities in each county
that have signed the U.S. Conference of Mayors Climate Protection Agreement out of those
that could have potentially signed it. To determine which cities were selected, I have used a
GIS shapefile from the U.S. Bureau of the Census of all cities and townships in the U.S. As
far as the size of cities in the sample is concerned, 2700 cities with 10,000 inhabitants or above
were included in the analysis. Most city-level research uses a considerably higher threshold of
30,000 or 50,000 inhabitants and consequently include less cities (Brody et al. 2008; Shipan &
Volden 2006, 2008). In order to allow for a comprehensive analysis of cities’ participation at the
county level, I chose the lowest possible threshold which still guarantees sufficient political and
administrative capacities to engage in local climate change policies.
The collection of the signature data of the MCPA for the dependent variable was
done through the website of the Mayors Climate Protection Center (www.usmayors.org/
climateprotection/). I have collected the participation records as of November 2008. This
specific date was chosen because it coincides with the election of Barack Obama as President
of the United States. I presume that the logic of signing a voluntary climate change agreement
has changed after the election simply due to the situation of uncertainty about national climate
policy. In other words, the certainty of no national climate change policy was not present anymore, since Obama and the Democratic Party were expected to enact climate change policies
at the national level. Since the cross-sectional nature of the data does not allow a test of different participation logics of policy-makers over time, I only include MCPA participation until
November 2008. Table 3.1 shows the city-level distribution of participation in the MCPA. Of
6
Most of the research on the county-level uses the 48 contiguous states Pagoulatos et al. (2004); Brody et al.
(2008) and excludes Hawaii and Alaska
Nature, Nurture or Neighbors?
49
the 2700 cities with over 10,000 inhabitants, 631 or 23.4 % had signed the MCPA by November
2008.7
Table 3.1: City-level distribution of MCPA participants
MCPA
signed
not signed
Total
No.
631
2,069
2,700
%
23.4%
76.6%
100.0%
In a next step, the 2700 cities were then allocated to their specific county using the FIPS
identifier and were merged with different county-level data sets. Due to the merge, the 631
cities that had signed are now located in 318 counties (or 10.2 %)8 of all 3109 counties. Taking
a closer look at the 2791 other remaining counties, it has to be noted that in a large proportion
there are no cities over 10.000 inhabitants. These counties would therefore have a value of zero
on the dependent variable. However, it would not be possible to distinguish a case where no city
in a specific county is in the sample from a case where there are cities over 10 000 inhabitants
in the county, but no city has signed. Since these describe two very different outcomes, the
analysis would potentially be biased by using all 3109 counties.
Table 3.2: Distribution of counties
MCPA status
cities have not signed
cities have signed
Total
City over 10.000 inhabitants in county
0
1
Total
No.
%
No.
%
No.
%
2,000 100
791 71.3 2,791 89.8
0
0
318 28.7
318 10.2
2,000 100 1,109 100 3,109 100
Table 3.2 identifies counties that include at least one city that is over the cut-off of 10,000
inhabitants, and shows those where at least one city signed the MCPA. To avoid this bias, I
therefore only include the 1109 counties (c.f. table 3.2) where at least one city is present and has
a chance of signing the MCPA. Figure 3.4 illustrates this for the northeastern part of the U.S.
Note that the counties where there is no city in the sample are light red. Accordingly, the black
7
8
Appendix table 3.14 further shows how participation is spread across the 48 states.
Dealing with counties, it has to be noted that they - sometimes dramatically - differ in size. For example, the
range goes from Loving county in Texas that has 46 inhabitants to Los Angeles County in California that has
around 9 million inhabitants
50
dots are the cities in the sample; counties where no city has signed the MCPA are dark red; the
different shades of yellow and green indicate counties where differing proportions of their cities
have signed. In figure 3.4 there are some clusters of ’green’ counties up in Massachussetts and
New Hampshire, whereas there are some ’red’ clusters in Pennsylvania.
Figure 3.4: Detailed map of participation and cities in the northeast; darker shades indicate a
higher percentage
For this reduced data set, the dependent variable on the county level was then calculated as
the proportion of cities that signed the MCPA of the possible cities over 10,000 citizens within
that county (%MCPA).9
3.3.2 Independent Variables
Nature
In section 3.2, I assumed that cities or states that are more likely to be affected by climate
change (by droughts, hurricanes, or an increase in sea levels) will be more prone to act than
cities or states, whose potential damage from climate change is perceived to be relatively lower.
In order to operationalize the concept of ’risk from climate change’, I use four different proxies.
The first proxy indicates whether a county is located at a coastline (coastal county) and is taken
from the National Association of Counties (NaCo). As Karl et al. (2009) conclude, the largest
9
Results from other possible operationalizations of the dependent variable, such as the count of cities, or a binary
measure indicating a 1 when at least one city in the county has signed and 0 otherwise are presented in the
robustness section.
Nature, Nurture or Neighbors?
51
probable climate change impact in the U.S. might arise from rising sea levels. Thus, coastal
counties should be directly affected. This becomes even more important with Rappaport &
Sachs (2003)’s observation of the U.S. as a coastal nation, where 50% of the population live
somewhere near a coast and a disproportionately large share of income and enterprises can be
found there.
Other natural risks emanating from climate change come from the intensifying of already
observable natural disasters. Hurricanes, as well as droughts and floods are predicted to become
more intense and frequent as the climate warms. I proxy the risk from future adverse effects by
using data on climate-related natural disasters such as hurricanes, tornadoes, floods, droughts
or storms10 in each of the counties from the Spatial Hazard Events and Losses Database for the
United States (SHELDUS) (Institute, Hazards & Vulnerability Research 2009). I use ’property
damage’ and ’fatalities from natural disasters’ within counties during the years 2000-2007 as the
second and third measure. I thus expect that a recent history of extreme weather-related events
leads to more voluntary participation.
The fourth measurement taps into the concept of regions’ benefitting from warmer climate.
I use IECC (2007) Code Climate Zones ranging from 0(very cold) – 14(hot-humid). Among
other indicators, these data include information on Heating Degree Days (HDD) and Cooling
Degree Days (CDD). Data are scaled so that a higher value means a less negative and potentially
beneficial impact from climate change. I would expect that participation in counties that are
less affected is lower and thus a negative impact on voluntary participation.
Nurture
Incentives
Each county population has their own preferences and attitudes toward environ-
mental regulation and these socio-economic and political factors were linked to the MCPA participation rate in section 3.2. Socio-economic and political characteristics expected to positively
influence participation in voluntary GHG mitigation initiatives are wealth, education, partisan
orientation, and age.
I proxy the wealth of the community by the per capita income (ICPSR 2008). For reasons
outlined earlier, I suppose that the higher the per capita income is in a county, the higher the
participation rate.
10
For example, earthquakes were not included in the data since there is no direct relation with climate change.
52
Education, or the human capital within a county, is measured using several indicators. For the
empirical test of hypothesis 3, I firstly use the percentage of persons with college education (% of
people that have had 4 years of college education) from the ’COUNTY CHARACTERISTICS,
2000-2007’ data set by ICPSR (2008). Here I also assume that the more educated the population of a county is, the higher the participation rate is. To tap at a slightly different concept, I
further include a measure on how many universities or colleges are in the county. Universities
might help to address climate change in their broader communities (Betsill 2001). Knuth et al.
(2007, 487) argue that universities can ’provide expert knowledge, research experience, and faculty and student time to communities lacking the necessary skills and resources to reduce their
GHG emissions’. I use the count of universities and colleges in counties from the Higher Education General Information Survey (HEGIS) XVIII: Institutional Characteristics of Colleges and
Universities (ICPSR 1984) scaled by the number of cities in the county. I generally include only
4-year universities or colleges that had a liberal arts program due to the fact that 2-year colleges
or those without liberal arts program often concentrate on vocational or professional education.
However, in this context, my theoretical rationale has to do with the humanistic background of
such universities to be more involved in the development of the community. Unfortunately, I
was not able to obtain more recent data on the institutional characteristics of colleges and universities. However, I do not expect a great deal of variation in 4-year universities and colleges,
since their reputations are established over long time periods.
In order to test my hypothesis 4, I proxied how liberal counties are by the share of votes for
the Democrats (% Dem vote) in election data on the 2004 presidential election. These data were
taken from the US Election Atlas (http://uselectionatlas.org/RESULTS/). The assumption
is that Democrats are more environmentally friendly (Volden 2006; O’Connor et al. 2002).
The median age of the county population was also taken from ICPSR (2008). I expect that a
younger population renders a county more liberal in general and should also make the city more
likely to support voluntary initiatives against climate change.
Disincentives
To proxy economic conditions in a community, I use the unemployment rate in
the county (Unemployment). I use the county level average from the County Characteristics
Dataset (ICPSR 2008). I expect that bad economic conditions would lead to less participation.
The U.S. produces more coal than it consumes (Fisher 2006) and since emissions from coal
are more CO2 -intensive than other energy sources, voluntary climate change regulations would
Nature, Nurture or Neighbors?
53
hit producers of coal and coal-powered energy hard. I would therefore expect less involvement
in the MCPA in counties with high coal production. I obtained data on coal production and
employment per county from the U.S. Energy Information Administration (EIA 2007).
CO2 intensity in a county proxies potential abatement costs from voluntary climate change
policy. The higher CO2 emissions are, the less participation in voluntary agreements is expected.
In October 2010, county-level CO2 data from the Vulcan Project (Gurney et al. 2009) was made
available. It is based on a number of data sources and relies to a great extent on data collected
by the EPA (and therefore relies on the integrity of the EPA data and models) through the Clean
Air Act (Gurney et al. 2009). I use per capita tonnes of carbon in the county to approximate
abatement costs in the locality.
Neighbors
In the spatial context employed here, the specification of the channels of interdependence on
the theoretical level implies the use of a certain specification of the empirical weights matrix
to account for spatial interdependence in the empirical model. The weights matrix and, thus,
the structure between the units of interest must be specified prior to an analysis of dependence
between observations and is, therefore, a crucial step in estimating a spatial model (Anselin 1988,
2004). Due to the inherent subjectivity in determining the spatial connectivities, Pluemper &
Neumayer (2010) advise scholars to show that their results hold for different weighting matrices.
I follow this suggestion in the sense that I use three connectivity matrices with slightly different
conceptions of geographical nearness.11 Gleditsch & Ward (2001) differentiate between strict
contiguity, which is what I measure with my first weights matrix, and proximity, which relates
to a more loose definition of contiguity that also incorporates units that do not necessarily
physically share a border. The second and third matrices belong in this category and measure the
geographic distance between counties. These two types of locational information – contiguity and
distance – are the ones most commonly used to quantify relative positions of units of observation
in space (Beck et al. 2006). The fourth measure then measures the concept of leaders of policy
innovation.
Geographical proximity In the empirical literature on diffusion, geographical proximity has
always played a major role in explaining policy diffusion since it is widely acknowledged to
11
To avoid problems pertaining to identification, the weights should be truly exogenous to the model (Manski
1993). The exogeneity of geography is unambiguous.
54
be a primary force shaping the opportunity for interaction (Walker 1969; Gray 1973; Tobler
1970; Gleditsch & Ward 2001; Berry & Berry 1990; Franzese & Hays 2006). Using ArcGIS,
geographical proximity is measured in three different ways here.
The first weights matrix measures strict contiguity. It is an N x N (1012 x 1012)12 county
contiguity matrix with non-zero entries for spatially contiguous counties. The second one then
measures the distances dij from the centroid of each county i to all the other counties in radians.13
Since I want to explore whether nearer counties are more similar, the inverse distance ( d1 ) was
calculated, so that counties that are closer to one another receive larger weights. Therefore in
each of the cells wij there is a measure of proximity between the counties. The third matrix uses
the similar set up as the second but employs a cut-off at 100km. This means that if counties were
more than 100km apart, a zero entry would result, while for each distance below the threshold,
the original distance was entered in the matrix. Each weighting matrix was row - standardized
(each row entry was divided by the row sum, to add up to 1) for all further analyses as is
standard in the spatial econometrics literature (Franzese & Hays 2006; Gleditsch & Ward 2001;
Ward & Gleditsch 2008).14
While the county contiguity matrix is limiting in the sense that it gives equal weight to all
bordering neighbors, the inverse distance matrix without a threshold assumes that each unit
in the system has some, albeit small impact on each other. Due to these deficiencies in the
conceptualization of proximity, specifying connectivities in an inverse distance matrix with a
cut-off appears to be the most viable option from a theoretical point of view. Nevertheless, all
three conceptualization of geographic distance will be tested in the spatial analysis section.
Leaders of Innovation
To proxy for the concept of leaders of innovation stated in hypothesis
9, I have gathered data on the first 141 cities that signed the agreement in early 2005. Figure 3.5
shows the map of all counties in the sample with all cities within them. Note that the red cities
are the first 141 that signed before the initiative was officially endorsed by the U.S. Conference
of Mayors (USCOM). Using ArcGIS, I have calculated the nearest distance of each county’s
centroid to one of the 141 cities. Since these cities then serve as external leaders that explain
12
The reduction from the original 1109 counties comes from the fact that there are isolated counties (i.e. counties
that have no contiguous neighbors) that were excluded from the analysis.
13
To calculate these point distances, GIS employs a formula using polar coordinates (φ, λ) to represent latitude
and longitude expressed in radians. These then can be converted to other units such as miles or km (c.f.
www.spatialanalysisonline.com/output/html/DistanceOperations.html).
14
For a more comprehensive and critical discussion concerning the topic of row standardization, see Pluemper &
Neumayer (2010)
Nature, Nurture or Neighbors?
55
the variation in participation, I had to exclude them from the analysis to avoid bias. I then
recalculated a new dependent variable for the signatories minus the first 141. This leads to a
reduced data set of 1085 counties, since in 24 counties, a city among the first 141 was the only
city that had signed. As stated in the theoretical part, for this concept of proximity, I expect
that cities that are located near one of the original 141 learned about the MCPA very early on
in the process, and had the potential to imitate and learn from those leaders. Accordingly, the
sign of the inverse distance variable is expected to be positive.
Figure 3.5: Counties with at least one city over 10.000 inhabitants; depicted in red are leaders
of innovation
3.4 Analysis and Results
To analyze the data at hand, I rely on a mixed strategy. Firstly, I take an exploratory look
at the data and visualize distributions of key variables across the U.S. This provides a first
hint regarding the possible clustering of values in specific regions. I then use more formalized
measures of spatial cluster analysis to determine whether a spatial analysis strategy is warranted.
Confirming the need to account for spatial interdependence in the analysis, I present in turn the
results for all three theoretically derived explanations (Nature, Nurture and Neighbors) for the
participation pattern in voluntary climate change initiatives. These results are then affirmed
using several robustness checks on different specifications of the dependent variable and different
estimation strategies. The section ends with a discussion on possible extensions.
56
3.4.1 Exploratory Analysis
Conveying data to the reader in numerical format using tables is one way, however, for a more indepth exploratory analysis, I follow the suggestion given by Ward & Gleditsch (2008) and start
my analysis by taking a look at maps. How does the participation pattern in the MCPA relate
to selected independent variables specified in the theoretical part? Can clusters of certain values
on different variables be detected, and do they significantly differ from what we would expect
from randomness? The two figures 3.9 and 3.10 in the appendix show the distribution of values
throughout the sample for four different variables. The dependent variable % MCPA is illustrated in subfigure 3.9(a), one explanatory variable on fatalities from natural hazards (Nature)
in subfigure 3.9(b), one explanatory variable depicts the distribution of votes for the Democrats
(Nurture: Incentives) in subfigure 3.10(a) and finally the distribution of unemployment rates
(Nurture: Disincentive) throughout the U.S. is given in 3.10(b).
Figure 3.9(a) shows that spatial clustering is clearly observable on the dependent variable.
Clusters of high values are recognizable in large parts of Washington State, New Jersey, New
York, and the New England states on the East Coast, in Florida as well as California on the West
Coast and in Illinois and Minnesota in the Midwest. A clustering of low values can be spotted
in parts of Colorado, Wyoming, and Nebraska as well as in the southern states of Louisiana and
Mississippi. A very first glance at the data shown in the map therefore seems to support the
conjecture made in section 3.2, that participation patterns of neighbors are more similar than
those of others farther away. Proceeding now to look at the two other explanations of MCPA
participation patterns, Nature and Nurture, I seek to determine whether a first exploratory
analysis can indicate sources for the spatial clustering in the dependent variable. Figure 3.9(a)
serves as the baseline.
In the case of the natural explanatory factors, we can see that indeed, in the southwestern
part of the U.S., especially in southern California and southern Arizona, higher participation
patterns and a high vulnerability to natural hazards seem to coincide. The high concentration
of fatalities in these areas mostly comes from severe heat waves and droughts in the years 2005
(for Arizona) and 2006 (for California). A similar observation can be made for southern Florida,
where most of the fatalities are due to hurricanes and tornadoes. Here also, participation in the
voluntary climate change initiative is quite high. However, northern Florida and Alabama along
the Gulf Coast, who are also affected by these natural hazards, do not seem to participate as
enthusiastically as their southern counterparts.
Nature, Nurture or Neighbors?
57
In subfigure 3.10(a), which shows clustering of the % of votes for the Democrats as a proxy
for a liberal and environmentally conscious electorate, we can see that clusters at the east coast
in the New England states as well as in California and Washington state on the west coast seem
to mirror high relative MCPA participation (subfigure 3.9(a)). However, a high concentration of
democratic votes along the Mississippi in states such as Mississippi, Arkansas, and Tennessee do
not seem to correlate with a high proportion of participation in the MCPA in these areas. One
explanation for the non-participation in the MCPA could be that there are more disincentives
than there are incentives to participate. Subfigure 3.10(b) shows that exactly this region around
the Mississippi is plagued by high unemployment rates, which were related to less participation
in section 3.2. In contrast, in other regions where unemployment rates are also quite high,
such as Michigan, California, and Washington, MCPA participation rates are nevertheless high.
Whether these regions have other incentives to participate cannot be deduced from this visual
evidence, but will be taken into account in the statistical analyses below.
One can see that a lot of the spatial clustering of the dependent variable might be attributed to
clustering in the explanatory variables. However, from such a cursory look, we cannot conclude
much, and as Ward & Gleditsch (2008, 11) correctly point out, ’humans are adroit at recognizing
patterns, even where no patterns exist’. To see whether spatial clustering is present, we have to
rely on more formal measures of spatial association, such as Global and Local Moran’s I (Anselin
1988). Local Moran’s I is a measure to detect clusters with values similar in magnitude on the
one hand and clusters of counties with very heterogenous values on the other. The statistic is
calculated as follows
n(yi − y) �
Ii = �n
wij (yj − y)
2
i=1 (yi − y)
(3.1)
j∈Ji
Figures 3.11 and 3.12 in the appendix show the associated z-standardized values of the local
Moran’s I statistic for the dependent variable and the three explanatory variables; fatalities,
% Democrats, and unemployment rate. The values are based on a county contiguity weights
matrix. Significant results are either depicted in red or in blue shades, and indicate whether
similarity or dissimilarity in values between the county and its neighbors is greater than what we
would expect by chance (Anselin 1988). A highly positive and significant Z score for a county is
illustrated in red and means that surrounding counties have similar values, either high (HH) or
low (LL). If there are several high Z scores next to each other, we speak of a cluster of high or
58
low values. A highly negative and significant Z score is shown in blue and means that a county
has dissimilar counties bordering it. So it either has a high value while the neighborhood is full
of low values (HL) or it has a low values relative to a high value neighborhood (LH).
As far as the spatial clustering in the dependent variable in subfigure 3.11(a) is concerned, two
or three of such high value clusters can indeed be spotted in California, Minnesota, Florida, and
New England. In these regions there exist more similar high values than we would predict by
chance. Apart from those larger clusters, there are several counties that have exceptionally high
MCPA participation values, but are surrounded by counties that have very low participation.
So it can be said that some spatial clusters of high MCPA participation exist. However, there
are less than would have been expected from our first cursory look at the distribution.
I turn now to the explanatory variables to see whether we can relate some of the clustering
in the dependent variables to exceptional clusters present in one of the explanatory variables.
Subfigure 3.11(b) indeed shows that the high values in Arizona and southern California in
subfigure 3.9(b) were exceptional and form one large cluster of counties with high numbers of
fatalities surrounded by counties with similarly high values. The same holds true for southern
Florida. Here, a cluster of high vulnerability to natural hazards corresponds to a cluster of
high MCPA participation in the region. Similarly, the significant cluster of high vote shares
for the Democrats in New England as well as in the Bay Area of California that can be seen
in subfigure 3.12(a), corresponds to a cluster of high MCPA participation counties in figure
3.11(a). As far as the Moran’s I Z scores from the unemployment rate in subfigure 3.12(b) are
concerned, no apparent similar clustering compared to the MCPA participation can be observed
from this visual inspection except again for southern Florida, where a cluster of counties with
comparatively low unemployment rates can be detected in an area where MCPA participation
is exceptionally high.
This exploratory analysis has provided an overview of the data at hand and served as a first
step in determining whether interdependence – as suggested by hypotheses 8 and 9 – truly could
be an important driver of participatory patterns in voluntary climate change initiatives. To
conclude the exploratory analysis, there is some evidence that there might be spatial dependence
in participation. However, the patterns we observe might not be caused by contagion. As we have
seen, another potential cause for the spatial pattern in the dependent variable are similarities in
values in observable explanatory variables (correlated X’s) as well as unobservable phenomena,
like common culture, preferences, or perceptions (Franzese & Hays 2008b). Spatial analysis can
Nature, Nurture or Neighbors?
59
help answering whether the spatial pattern of the dependent variable is due to spatial dependence
in the data or spatial clustering in the explanatory variables. The following section first presents
the results concerning the relation of Nature and Nurture to participation in the MCPA and
then focuses on adequately testing for the Neighbors component.
3.4.2 Statistical Analysis and Results
Nature and nurture: regression analysis
In the following, I first test the arguments brought forward in the theoretical section concerning
the Nature and Nurture components. To this end, I have estimated several ordinary least squares
(OLS) regression models on participation in the Mayors Climate Protection Agreement, which
are depicted in table 3.3. The dependent variable in my analysis ranges from 0 to 100 and gives
the percentage of cities within the county that have signed over those that could potentially
have signed.15 For the estimation of these models, robust standard errors were used (White
1980; Wooldridge 2009).16
The first column of table 3.3 shows a test of hypothesis 1 concerning the impacts of natural
risk factors on participation in voluntary agreements. We can see that the coefficient on the
climate region variable is negative and strongly significant. This result is in accordance with the
theoretically specified conjecture that one would expect counties in regions which might benefit
from climate change (a higher value on the climate region variable) to have lower participation
rates compared to counties in those regions that are more likely to be negatively affected. Similarly, being a coastal county and, therefore, more exposed to climate change risk shows to have a
huge impact on MCPA participation. As regards to the two measures to proxy the vulnerability
towards natural hazards, the coefficient on the property damage variable does not turn out to
be significantly different from zero, while a higher number of fatalities through natural hazards
is positively related to more participation. The difference in relative impact of the two natural
hazards measures might be due to the fact that property damage is not seen as threatening as
reports about fatalities, and property damage alone might not raise awareness for consequences
from climate change.
15
16
Summary statistics for all variables used in the regressions can be found in appendix table 3.13.
Diagnostic tests on the OLS residuals have detected heteroscedasticity of the residuals meaning the assumption
V ar(�i ) �= σ 2 is not fulfilled. The non-constant variance of the residuals does not bias the coefficients, but
renders them inefficient (Maddala 2001). A remedy for this problem is the use of robust standard errors
(White 1980).
60
Table 3.3: Regression table (Baseline)
(1)
Nature
-1.70∗∗∗
(0.32)
0.00
(0.00)
19.00∗∗∗
(2.76)
1.40∗∗∗
(0.48)
Climate Region
Property Damage
Coastal County
Fatalities
% 4yrs College
% Dem Vote
Lib arts scaled by cities
p.c. income
Median Age
Unemployment Rate
p.c. tonnes carbon
Population
Pop Growth (00-05)
% Manufacturing Empl
Coal Production
Industry dependent
% Loc/State Gov Empl
EPA non-attainment
25.72∗∗∗
(2.66)
1109
0.07
32.54
Constant
Observations
Adjusted R2
s.e. of estimate
Standard errors in parentheses
∗
p < 0.10,
∗∗
p < 0.05,
∗∗∗
p < 0.01
(2)
Nurture
0.90∗∗∗
(0.22)
0.88∗∗∗
(0.10)
4.20∗∗∗
(1.17)
-0.07
(0.28)
-0.17
(0.30)
-2.62∗∗∗
(0.64)
-0.05
(0.03)
-0.00
(0.01)
0.12
(0.15)
-0.01
(0.16)
0.00
(0.00)
-0.68
(2.14)
-0.10
(0.24)
0.01
(2.26)
-16.26
(14.13)
1080
0.27
28.63
(3)
Full
-0.18
(0.31)
0.00
(0.00)
7.80∗∗∗
(2.76)
0.74∗
(0.44)
1.03∗∗∗
(0.23)
0.81∗∗∗
(0.10)
4.09∗∗∗
(1.13)
-0.23
(0.27)
-0.22
(0.30)
-2.63∗∗∗
(0.65)
-0.05
(0.03)
-0.03∗
(0.02)
0.05
(0.15)
0.09
(0.16)
0.00
(0.00)
-0.26
(2.11)
-0.10
(0.24)
0.58
(2.24)
-10.68
(14.58)
1080
0.28
28.49
(4)
comb’I
(5)
comb’II
5.75∗∗
(2.41)
0.54
(0.42)
0.81∗∗∗
(0.23)
0.79∗∗∗
(0.09)
4.12∗∗∗
(1.17)
-0.03
(0.25)
-0.31
(0.29)
-2.58∗∗∗
(0.62)
-0.05
(0.03)
5.72∗∗
(2.40)
-9.88
(11.19)
1109
0.27
28.75
0.85∗∗∗
(0.14)
0.78∗∗∗
(0.08)
4.35∗∗∗
(1.23)
-2.35∗∗∗
(0.58)
-23.90∗∗∗
(4.20)
1109
0.27
28.76
Nature, Nurture or Neighbors?
61
The second column in table 3.3 then provides a first test of hypotheses 2-7, taping at the
socio-economic and political conditions in the county that are assumed to either provide incentives or disincentives for participation. Starting with hypothesis 2, we can see that counter to
expectation, wealth (as measured by per capita income) does not have the expected positive
influence on participation. Instead, the coefficient is negative and not statistically significant.17
An ad-hoc explanation for this non-finding might be that most of the pro-environmental effect
of income is picked up by other variables, such as education or age. I will return to this nonfinding in more detail below. For hypothesis 3, the suggested strong and positive influence of
education on MCPA participation is confirmed by both indicators. First, a higher % of the
population in a county that has a college education leads to higher participation rates. Second,
a higher concentration of universities or colleges with liberal arts programs within the county
18
also has a sizable impact on participation. Another characteristic of the communities in
the county that is assumed to increase MCPA participation is the % of Democrats supporters
within the county. The coefficient on the variable % Dem Vote is positive and highly significant,
confirming hypothesis 4. Democratic constituencies seem to be successful in demanding more
voluntary participation in those counties. I now turn to hypothesis 5 stipulating that the higher
the median age in the county is, the less likely is participation. Although the coefficient on
median age bears the theoretically expected sign, it is not statistically significant. This seems
to indicate that younger constituencies are not more concerned about climate change than their
older counterparts. Further potential incentives leading to high participation were tested by the
% of employment in state or local government. This variable is supposed to proxy a higher capacity and professionalism of governmental agencies within the county that might lead to more
participation in voluntary agreements due to more information about voluntary climate change
initiatives. However, no such effect could be found.
With respect to the hypotheses about disincentives to participation, hypothesis 6 is confirmed
by evidence from table 3.3. The coefficient on unemployment shows a huge negative and significant effect of bad economic conditions on participation. A high unemployment rate in a region
might, therefore, indicate that politicians are interested in policies to ameliorate the economic
situation rather than subscribing to voluntary climate change commitments. A further disincentive might arise from high abatement costs from voluntary climate change agreements that the
17
18
Using other indicators, such as total personal income, does not change the substantive result.
Using also universities without a liberal arts program (only 390 of 1863 did not offer a liberal arts program)
does not change the substantive result, however, as expected, the effect size is slightly smaller.
62
county would have to suffer in case of participation. The results from table 3.3, however, only
weakly confirm hypothesis 7. Although the coefficient on CO2 emissions (p.c. tonnes carbon)
in the county does not reach standard levels of significance, it shows the theoretically expected
negative sign. Other control variables proxying abatement costs such as the % of manufacturing
employment, industry dependence of the county, coal production in the county, or the EPA nonattainment status19 are also not found to be significantly related to participation. This finding
is quite counter to what I expected, and it is not affected by different model specifications (not
shown here). I will return to the relatively better performance of the indicators proxying incentives compared to those looking at disincentives more extensively in the section on robustness
checks and alternative specifications.
Column three of table 3.3 then displays a full model of Nature and Nurture characteristics. The
substantive results from columns 1 and 2 remain unchanged. However, in this model, the climate
region variable is no longer significant as it seems that socio-economic and political characteristics
perform better in explaining the variation in the dependent variable. Furthermore, the coefficient
on the variable population size just reaches significance and indicates that a larger population
in a county is less likely to widely participate in the MCPA. Adding natural risk characteristics,
however, does not add a lot to the model fit as we can see from the only slight increase in the
Adjusted-R2 and the decrease in the standard error of the estimate.
The last two columns in table 3.3 then combine Nature and Nurture characteristics in a more
parsimonious way. The model in column 4 includes the variables that were deemed by theory
to be most important for participation in voluntary climate change agreements, namely vulnerability to climate change, income, education, age, political identification, economic distress, and
abatement costs. Dropping all other variables from the model does not change the model fit
drastically. In column 5, the theoretical baseline model is further reduced to include only those
variables that were significantly related to MCPA participation in the models in columns 1-4.
Again, neither model fit nor the substantive results change as a consequence of further reducing
the number of explanatory factors to only five. This gives a first indication on the robustness
of the reported effects.
One issue that regularly raises concern, e.g. when confronted with non-findings from theoretically relevant variables, is multicollinearity. In the presence of multicollinearity, coefficients
19
This control variable measures whether the county has been a non-attainment county for Criteria Air Pollutants
under the Clean Air Act in 2004, 2005, or 2006. Non-attainment counties that already have difficulty complying
with given EPA legislation are expected not to participate in voluntary GHG efforts.
Nature, Nurture or Neighbors?
63
may have the wrong sign or lower / higher than predicted magnitudes. When is multicollinearity a problem? Greene (2003) suggests calculating the variance inflation factor (VIF) for each
coefficient as a diagnostic statistic to gauge whether the variances of the estimates are adversely
affected by the intercorrelation between them. This VIF then shows how much of the estimate’s
variance is inflated by multicollinearity. Observing the VIF factors of all the models calculated
in table 3.3 shows that they are well below the commonly accepted rule of thumb of 10 (O’Brien
2007). The estimate of % 4yrs college and the one of per capita (p.c.) income had the highest
VIF factors with just over 3. In addition, a simple pairwise correlation of all the variables shows
that those two variables were correlated with .68. While this could explain the insignificance of
the coefficient on income, additional tests show that it does not. As multicollinearity does not
affect the overall results and as education and income are assumed to be theoretically relevant
for the explanation of MCPA participation, both variables are included in subsequent models.
Neighbors: Spatial analysis
To empirically test the hypothesis 9 concerning the interdependence in participation between
geographically proximate counties, I use spatial analysis. Spatial analysis can help to take
dependence into account and deal with spatially clustered phenomena. Furthermore, spatial
models allow to examine the impact that one observation has on other, proximate observations
(Ward & Gleditsch 2008). To this end, the following spatial model (Franzese & Hays 2008b) is
estimated
y = ρWy + Xβ + �
(3.2)
where y is an N x 1 vector of observations on the % of MCPA signatories, X is an N x k matrix
of observations the independent variables concerning the Nature and Nurture characteristics, β
is an k x 1 vector of coefficients on the X� s and � is an N x 1 vector of disturbances. W is the
spatial weighting matrix that reflects the connectivity between neighboring observations and ρ is
a scalar that measures the impact of MCPA participation in the neighboring (as defined by wij )
counties on the participation in county i. Wy – the spatial lag– is the weighted average of MCPA
participation in county i’s neighbors. However, because of the endogenous spatial lag on the
right-hand side, estimating the model in equation 2 with standard OLS will induce bias (Anselin
1988). More precisely, estimating the model using standard OLS regression would in the case of
expected positive spatial dependence lead to an overestimation of spatial interdependence and
64
underestimate the strength of unit-level factors (Hays 2009; Franzese & Hays 2007). Franzese
& Hays (2008a) recommend using spatial maximum likelihood (S-ML) in this case.
To test whether it is warranted to use spatial econometrics, I have checked the residuals from
the OLS regression for spatial dependence. The residuals are usually inspected to see whether
the OLS results, which assume independent observations, are affected by spatial clustering in
the dependent and independent variables. Both the global Moran’s I test statistic as well as the
Lagrange Multiplier test (Ward & Gleditsch 2008) indicated the presence of such dependencies
in the data and therefore empirically confirmed theoretical considerations about the importance
of interdependence in studying MCPA participation.
Table 3.4: Spatial ML; W based on county contiguity (0 / 1 )
(1)
Neighbors
mu
% 4yrs College
% Dem Vote
Unemployment Rate
Median Age
p.c. income
p.c. tonnes carbon
Lib arts scaled by cities
Coastal County
Fatalities
Constant
rho
Constant
Log lik.
χ2
Prob> χ2
Observations
14.49∗∗∗
(1.22)
0.21∗∗∗
-4978.37
(0.04)
1012
(2)
Combined I
0.66∗∗∗
0.76∗∗∗
-2.96∗∗∗
-0.48∗
0.01
-0.06
4.31∗∗∗
6.02∗∗∗
0.41
0.00
(0.18)
(0.09)
(0.70)
(0.29)
(0.20)
(0.07)
(0.83)
(2.27)
(0.34)
(11.35)
0.08∗
-4814.38
323.58
0.00
1012
(0.05)
(3)
Combined II
0.75∗∗∗
0.74∗∗∗
-2.65∗∗∗
(0.13)
(0.09)
(0.69)
4.55∗∗∗
5.80∗∗∗
(0.83)
(2.21)
-19.97∗∗∗
(5.04)
0.08∗
-4817.38
320.69
0.00
1012
(0.04)
Standard errors in parentheses
∗
p < 0.10,
∗∗
p < 0.05,
∗∗∗
p < 0.01
As mentioned in section 3.3.2, I use three different conceptions of geographic proximity. Each
of the three weighting matrices is tested in turns. I conduct the empirical testing of hypothesis
9 with the first conception of proximity, strict contiguity (Gleditsch & Ward 2001). The results
from the spatial-ML based on the county contiguity W are shown in table 3.4. Model 1 tests the
Neighbors hypothesis with an empty model, only including the spatial lag. We can see that the
estimate for the spatially lagged y is positive and highly significant, indicating that a county’s
MCPA participation positively covaries with the level of MCPA participation of its neighbors.
Nature, Nurture or Neighbors?
65
A one unit positive shock to the MCPA participation level in the neighbors would lead to an
immediate 0.21 increase in the county’s own participation (Franzese & Hays 2006). However,
so far I do not account for other covariates explaining MCPA participation. In models 2 and
3, I add the two baseline models from the Nature and Nurture context (models 4 and 5 from
table 3.3). The coefficients in this spatial model differ from those obtained in the non-spatial
model in that models 2 and 3 control for spatial dependence in y. For example, in model 2, the
coefficient on % Dem Vote now represents the effect of democratic votes for MCPA participation
while not only controlling for other Nurture and Nature characteristics, but also for the extent
that county i’s participation level can be described by the value of y in connected counties j
(Ward & Gleditsch 2008; Franzese & Hays 2006). As expected, all coefficients remain robust in
their direction and significance. The only change that can be observed is that, in model 2, the
coefficient on median age has a negative impact on participation and reaches conventional levels
of significance.
Table 3.5: Spatial ML; W based on Inverse Distance
(1)
Neighbors
mu
% 4yrs College
% Dem Vote
Unemployment Rate
Median Age
p.c. income
p.c. tonnes carbon
Lib arts scaled by cities
Coastal County
Fatalities
Constant
rho
Constant
Log lik.
χ2
Prob> χ2
Observations
12.82∗∗∗
(1.34)
0.30∗∗∗
-5457.89
(0.05)
1109
(2)
Combined I
0.82∗∗∗
0.75∗∗∗
-2.45∗∗∗
-0.35
-0.06
-0.04
4.24∗∗∗
5.25∗∗
0.55
-8.83
(0.18)
(0.09)
(0.69)
(0.27)
(0.20)
(0.06)
(0.85)
(2.25)
(0.35)
(10.92)
0.10∗
-5291.47
383.08
0.00
1109
(0.05)
(3)
Combined II
0.86∗∗∗
0.74∗∗∗
-2.19∗∗∗
(0.13)
(0.09)
(0.67)
4.46∗∗∗
5.17∗∗
(0.84)
(2.19)
-24.70∗∗∗
(4.94)
0.09∗
-5294.32
375.49
0.00
1109
(0.05)
Standard errors in parentheses
∗
p < 0.10,
∗∗
p < 0.05,
∗∗∗
p < 0.01
As a second test, I now use the N x N (1109 x 1109) spatial weights matrix connecting the
inverse distance between counties to estimate the spatial ML (see table 3.5). Again, an empty
model including only the spatial lag is estimated first. As in table 3.4, ρ, the coefficient on the
66
spatial lag, is highly significant in the first model and is only marginally significant at the 10%
level as soon as the other theoretically relevant explanatory variables are included. Once more,
the substantive results from the two baseline models stay exactly the same.
In table 3.6, the spatial ML is now calculated using the second distance weights matrix
with the imposed threshold of 100km.20 In section 3.2.3, it was argued that compared to
the county contiguity matrix, which gives equal weight across all bordering neighbors, and
the regular inverse distance matrix, where all units are expected to influence one another, the
inverse distance matrix with a threshold allows us to better capture spatial distances. Once
again, nothing changes with respect to the county-internal Nature and Nurture characteristics.
However, interestingly, ρ is only significantly related to more MCPA participation in the empty
model. In model 2 and 3, it does not reach conventional levels of significance, while all other
coefficients remain unchanged. This fact is not too surprising given that both other spatial
weights, although generally indicating the presence of spatial dependence, were only marginally
significant at the 10% level.21
Lastly and as a fourth test of the effects of spatial interdependence, I look at the influence of
leaders of innovation. In section 3.2.3, I have assumed that the closer a county is to one of the
141 leaders of innovation that first signed the Mayors’ Climate Protection Agreement in 2005,
the higher is its participation level. Analogously to the tables before, in table 3.7 the measure of
proximity (inverse distance) to the closest of the 141 cities is highly significant in an empty model
(model 1). The positive effect from the proximity to the leader cities on MCPA participation in
the county is huge. However, as I add theoretically important covariates to the model, the effect
again vanishes and even gets negative. This would imply that, controlling for all other covariates,
nearness to one of the leader cities is even detrimental for MCPA participation. However, this
effect is not significant in models 2 and 3. Other than that, no coefficients substantively change
direction in the models including the proximity to a leader city.
How can we interpret the findings regarding the Neighbors component in MCPA participation
patterns? Although evidence that MCPA participation patterns are positively influenced by
MCPA participation in neighboring counties was found, the spatial analysis did not find a
robust spatial effect for the different conceptions of nearness. Furthermore, Hypothesis 9 on the
20
Results from an inverse distance matrix with a threshold of 200 km can be provided, however, they do not
change results
21
Although Franzese & Hays (2008a) find from Monte Carlo Simulations that spatial ML tends to underestimate
the strength of interdependence, they also find that this is a more serious problem in small-N samples, thus,
the analysis of 1109 counties should not seriously be affected by this bias.
Nature, Nurture or Neighbors?
67
Table 3.6: Spatial ML; W based on Inverse Distance up to 100 km
(1)
Neighbors
mu
% 4yrs College
% Dem Vote
Unemployment Rate
Median Age
p.c. income
p.c. tonnes carbon
Lib arts scaled by cities
Coastal County
Fatalities
Constant
rho
Constant
Log lik.
χ2
Prob> χ2
Observations
15.32∗∗∗
(1.19)
0.16∗∗∗
-5459.57
(0.04)
1108
(2)
Combined I
0.81∗∗∗
0.77∗∗∗
-2.53∗∗∗
-0.33
-0.05
-0.04
4.19∗∗∗
5.53∗∗
0.54
-9.01
(0.18)
(0.09)
(0.69)
(0.28)
(0.20)
(0.06)
(0.85)
(2.25)
(0.35)
(10.95)
0.04
-5288.21
397.84
0.00
1108
(0.03)
(3)
Combined II
0.86∗∗∗
0.76∗∗∗
-2.28∗∗∗
(0.13)
(0.09)
(0.67)
4.41∗∗∗
5.48∗∗
(0.84)
(2.19)
-24.00∗∗∗
(4.94)
0.03
-5290.89
391.28
0.00
1108
(0.03)
Standard errors in parentheses
∗
p < 0.10,
∗∗
p < 0.05,
∗∗∗
p < 0.01
Table 3.7: Basic Models with Proximity to Leaders of Innovation
(1)
Neighbors
Proximity to
leader city
% 4yrs College
% Dem Vote
Unemployment Rate
Median Age
p.c. income
p.c. tonnes carbon
Lib arts scaled by cities
Coastal County
Fatalities
Constant
Observations
Adjusted R2
s.e. of estimate
17.21∗∗∗
14.30∗∗∗
1085
0.02
31.30
Standard errors in parentheses
∗
p < 0.10,
∗∗
p < 0.05,
∗∗∗
p < 0.01
(4.17)
(0.99)
(2)
Combined I
-4.65
0.54∗∗
0.66∗∗∗
-2.11∗∗∗
-0.43
0.27
-0.04
4.84∗∗∗
4.17∗
0.66
-8.02
1085
0.21
28.11
(4.64)
(0.22)
(0.09)
(0.60)
(0.28)
(0.26)
(0.03)
(1.29)
(2.41)
(0.43)
(10.54)
(3)
Combined II red
-3.65
0.74∗∗∗
0.66∗∗∗
-2.04∗∗∗
(4.47)
(0.15)
(0.09)
(0.57)
5.17∗∗∗
4.93∗∗
(1.17)
(2.42)
-19.99∗∗∗
1085
0.20
28.15
(4.45)
68
distance to leaders of innovation was refuted. The fact that a weakly significant spatial effect is
present in two of three weights matrices is of concern only due to the fact that the two distance
weights matrices, Inverse Distance and Inverse Distance (100km) do not appear to make a huge
difference at first glance. However, the choice of using one over the other would substantively
change the result from the spatial analysis. This result corroborates Pluemper & Neumayer
(2010)’s concern about justification and validity of one chosen set of connectivities as compared
to another. As Beck et al. (2006, 28) remark: ’The assumption that these connectivities are
known a priori is both a strong assumption and critical for the methods of spatial econometrics
to work ’. Unfortunately, precise theoretical considerations about which spatial connectivity is
the right one for a given substantive problem do virtually not exist (Anselin 2004), and even
with better theories, some arbitrariness will remain (Pluemper & Neumayer 2010).
Although hypothesis 8 on the influence of participation in neighboring counties on a county’s
participation could not be unambiguously confirmed, results pertaining to the two other explanations of MCPA participation, Nature and Nurture, were supported by the OLS as well the
spatial ML analysis. Robust results are that favorable conditions for a high participation in the
MCPA are highly educated and liberal constituencies. Coastal counties that are vulnerable to
the effects of global warming are also consistently found to increase participation in voluntary
climate change agreements. A high proportion of unemployed persons means that policies concerning climate change are not the first priority within the county and lead to less participation.
However, income did not have a statistically significant positive effect on participation, although
in theory it was seen as an important determinant for environmental consciousness. Also, abatement costs, as proxied by CO2 emissions, which substantively should have a deterrent effect,
do not seem to play a role in predicting participation. The following robustness section rigorously checks whether these main results hold for different model specifications and estimation
methods.
3.4.3 Robustness Checks
The first part of the robustness section looks at different model specifications with respect to the
inclusion of interactive terms to the regression and their implication for the substantive results
discussed in the previous section. In the second and more extensive part, I then use three
alternative estimation methods to explain participation in local climate change agreements and
compare the results with the ones obtained in the main analysis above.
Nature, Nurture or Neighbors?
69
Model specification
In this subsection, I directly follow up on the puzzling result obtained from the main analysis
that income did not have a statistically significant effect on MCPA participation, although in
theory it was seen as an important determinant for environmental consciousness. To that end,
I test whether income might not have a direct effect on participation, but if the effect of income
could be conditional on the % of voters for the Democratic Party. Models 1 and 2 in table 3.8
show the results from adding an interactive term to the baseline regression models. It can be seen
that while the coefficient on % Democrats stays significant and similar in magnitude compared
to the baseline, the coefficient on the interactive term does not reach statistical significance and
shows no effect. To better grasp the result, figure 3.6 visualizes the marginal effect of p.c. income
on MCPA participation at varying percentages of democratic vote. As suspected, we can see a
slightly negative marginal effect of income on MCPA in areas where the constituency is leaning
more to the Republicans and approaching zero for higher values of % Democrats. However,
the effect never reaches statistical significance. This result confirms the results obtained in the
-1
Marginal Effect of p.c. income
on \% MCPA
-.5
0
.5
1
previous section regarding the null effect of income on MCPA participation.
0
20
40
60
80
100
\% Dem Vote
Dashed lines give 95% confidence interval.
Figure 3.6: Interaction effect between income and % Democrats in the county
The second interactive term also has to do with the % of democratic votes in counties. In
hypothesis 4, the pro-environmental effect of democratic votes was related to the positioning
of the democratic leaders on the issue. However, I would not have expected this effect of the
% of Democrats on MCPA participation to be so persistently high. The reason for this is that
the electorate of the Democratic Party is quite heterogeneous. Due to their stance on social
security, many low income or unemployed individuals belong to their vote base (CNN exit polls
70
Table 3.8: Regression table with Interaction terms
% 4yrs College
% Dem Vote
Unemployment Rate
Median Age
p.c. income
p.c. tonnes carbon
Lib arts scaled by cities
Coastal County
Fatalities
%Dem x Income
(1)
%Dem*Income
0.81∗∗∗
(0.23)
0.79∗∗∗
(0.09)
-2.58∗∗∗
(0.61)
-0.30
(0.30)
-0.05
(0.26)
-0.05
(0.03)
4.08∗∗∗
(1.16)
5.72∗∗
(2.43)
0.55
(0.42)
0.00
(0.01)
(2)
%Dem*Income
0.93∗∗∗
(0.19)
0.78∗∗∗
(0.09)
-2.46∗∗∗
(0.60)
-0.18
(0.23)
4.15∗∗∗
(1.16)
5.79∗∗
(2.43)
0.58
(0.42)
0.00
(0.01)
%Unemp x %Dem
Constant
-9.72
(11.13)
1109
0.27
28.76
Observations
Adjusted R2
s.e. of estimate
Standard errors in parentheses
∗
p < 0.10,
∗∗
p < 0.05,
∗∗∗
p < 0.01
-20.01∗∗∗
(5.60)
1109
0.27
28.76
(3)
Unemp*%Dem
0.73∗∗∗
(0.23)
1.59∗∗∗
(0.21)
4.05∗∗∗
(1.48)
-0.44
(0.30)
-0.02
(0.26)
-0.04
(0.03)
4.26∗∗∗
(1.21)
5.34∗∗
(2.43)
0.59
(0.41)
(4)
Unemp*%Dem
0.81∗∗∗
(0.14)
1.50∗∗∗
(0.20)
3.84∗∗
(1.51)
-0.15∗∗∗
(0.03)
-38.36∗∗∗
(12.47)
1109
0.28
28.61
-0.14∗∗∗
(0.03)
-54.73∗∗∗
(8.47)
1109
0.28
28.62
4.36∗∗∗
(1.29)
4.93∗∗
(2.44)
0.61
(0.41)
71
-2
Marginal Effect of \% Dem Vote
on \% MCPA
-1
0
1
2
Nature, Nurture or Neighbors?
0
5
10
Unemployment Rate
15
Dashed lines give 95% confidence interval.
Figure 3.7: Interaction effect between % Democrats and unemployment in the county
2008). Among these voters, environmental concerns may not rank highly, and I therefore expect
that in areas where there is high unemployment, but also a high proportion of votes for the
Democrats, the pro-environmental effect of the Democratic Party is less pronounced. Models 3
and 4 in table 3.8 show the baseline models with the inclusion of an interactive term between
the % of democratic votes and % unemployment. As expected, the interactive term is negative
and also significant. Figure 3.7 nicely illustrate how the effect of % democratic vote is positive
and very strong for low levels of unemployment. However, as unemployment increases, the
positive effect becomes less strong and eventually even turns into a negative one for very high
levels of unemployment. As far as all other covariates in the models are concerned, they remain
unchanged. In addition, the model fit has slightly increased compared to the baseline models.
Estimation
Logistic Regression In a similar study, Brody et al. (2008) look at county-level CCP participation patterns and operationalize their dependent variable to take the value one if there is a city
within the county that has committed to the CCP agreement and zero otherwise. Although this
reduces information or variance on the dependent variable, I use a different operationalization
of my dependent variable as a robustness check on the results obtained from the OLS regression
above. Table 3.9 presents the results of a logistic regression on the existence of at least one city
in the county that is a participant in the Mayors Climate Protection Agreement. The estimated
models are the same as for the regression analysis with two slight changes. First, the university
/ colleges variable is now also binary, indicating whether a county has a university with a 4-year
72
Table 3.9: Logistic Regression whether at least one city in the county has signed MCPA
(1)
Nature
0.53∗∗∗
(0.06)
-0.12∗∗∗
(0.03)
-0.00
(0.00)
1.13∗∗∗
(0.19)
0.10∗∗
(0.04)
MCPA
No.of cities
Climate Region
Property Damage
Coastal County
Fatalities
% 4yrs College
% Dem Vote
Uni/Coll
p.c. income
Median Age
Unemployment Rate
p.c. tonnes carbon
Population
Pop Growth (00-05)
% Manufacturing Empl
Coal Production
Industry dependent
% Loc/State Gov Empl
EPA non-attainment
-1.49∗∗∗
(0.22)
0.23
-513.25
1109
Constant
Pseudo R2
Log lik.
Observations
Standard errors in parentheses
∗
p < 0.10,
∗∗
p < 0.05,
∗∗∗
p < 0.01
(2)
Nurture
0.28∗∗∗
(0.09)
0.05∗∗
(0.02)
0.09∗∗∗
(0.01)
0.74∗∗∗
(0.24)
0.00
(0.03)
-0.00
(0.03)
-0.26∗∗∗
(0.09)
-0.02∗
(0.01)
0.01
(0.02)
0.02
(0.02)
0.01
(0.02)
-0.00
(0.00)
-0.12
(0.25)
-0.01
(0.02)
0.03
(0.23)
-5.82∗∗∗
(1.54)
0.41
-384.59
1080
(3)
Full
0.28∗∗∗
(0.09)
-0.01
(0.04)
0.00
(0.00)
0.44∗
(0.26)
0.08
(0.05)
0.06∗∗∗
(0.02)
0.08∗∗∗
(0.01)
0.76∗∗∗
(0.24)
-0.01
(0.03)
-0.01
(0.03)
-0.25∗∗∗
(0.09)
-0.02∗
(0.01)
0.01
(0.02)
0.02
(0.02)
0.02
(0.02)
-0.00
(0.00)
-0.10
(0.25)
-0.01
(0.02)
0.04
(0.23)
-5.54∗∗∗
(1.57)
0.41
-381.76
1080
(4)
comb’I
0.38∗∗∗
(0.07)
(5)
comb’II
0.42∗∗∗
(0.07)
0.30
(0.22)
0.10∗∗
(0.04)
0.04∗
(0.02)
0.08∗∗∗
(0.01)
0.81∗∗∗
(0.22)
0.03
(0.03)
-0.03
(0.03)
-0.24∗∗∗
(0.08)
-0.02∗
(0.01)
0.39∗
(0.22)
-5.21∗∗∗
(1.23)
0.39
-402.33
1109
0.05∗∗∗
(0.01)
0.08∗∗∗
(0.01)
0.82∗∗∗
(0.22)
-0.25∗∗∗
(0.08)
-5.83∗∗∗
(0.55)
0.39
-406.75
1109
Nature, Nurture or Neighbors?
73
program or not. Second, a variable was added on how many cities there are in the county. The
rationale behind the inclusion is, of course, that the likelihood that at least one city signs the
MCPA increases with more cities in a county. As expected, the coefficient on the number of cities
is positive and highly statistically significant. As far as the natural characteristics are concerned,
the effect of being a coastal county does not reach statistical significance in all models, whereas
fatalities become a significant predictor of whether there is at least one city in the county that
has signed the MCPA. The only significantly changed variable is the per capita carbon emission.
A higher emission value per capita now significantly decreases the likelihood of having a city in
the county that has signed the agreement. The difference in results from the logistic regression
compared to OLS obviously lies in the interpretation of the dependent variable. Whereas p.c.
emissions might not serve as a good predictor of the percentage of MCPA signatories, it turns
out to be a significant predictor whether there is at least one city that signs in a county or not.
I will address this slight difference in interpretation in more detail in the following alternative
conceptualizations of the dependent variable.
Quantile (Median) Regression In the statistical analyses above, it was mentioned that the
dependent variable is rightly skewed. Therefore, the homoscedasticity assumption is not met
and robust standard errors were used to circumvent the problem. However, while this remedy
renders perfectly valid standard errors, it can be regarded as fixing the nuisance associated
with the distribution of the dependent variable rather than explicitly modeling the substance
of it. Traditional regression analysis focuses on the relationship between a response variable
and its predictor variables by modeling the conditional mean of the response (Hao & Naiman
2007). However, in instances of high skewness, only relying on the mean as a measure of
central tendency might fail to capture other informative trends in the response function (Hao
& Naiman 2007). The right skewness of the distribution of MCPA participation makes the
mean (18.2) considerably larger than the median (0). In such cases, median regression might
be an alternative approach. Median regression estimates the median instead of the mean of
the dependent variable, conditional on the values of the independent variables (For a more
complete and formal introduction in quantile regression models c.f. Koenker & Bassett 1982;
Hao & Naiman 2007; Wooldridge 2009). Table 3.10 compares the result for the two baseline
models estimated using standard OLS regression in models 1 and 3 with the models from the
median regression using Koenker-Bassett standard errors (Koenker & Bassett 1982). For the first
74
model with all the theoretically relevant covariates, we observe some differences between OLS
and median regression, whereas in the second baseline model, no huge changes are noticeable.
Similar to the linear framework, coefficients can be interpreted as the change in the median of
Table 3.10: OLS (Mean) Regression vs. Quantile (Median) Regression with Koenker-Basset (kb)
(1982) standard errors
% 4yrs College
% Dem Vote
Unemployment Rate
Median Age
p.c. income
p.c. tonnes carbon
Lib arts scaled by cities
Coastal County
Fatalities
Constant
Adjusted R2
Observations
(1)
Mean (Comb’I)
0.81∗∗∗
(0.23)
0.79∗∗∗
(0.09)
-2.58∗∗∗
(0.62)
-0.31
(0.29)
-0.03
(0.25)
-0.05
(0.03)
4.12∗∗∗
(1.17)
5.75∗∗
(2.41)
0.54
(0.42)
-9.88
(11.19)
0.27
1109
(2)
q50 kb
0.12∗∗
(0.05)
0.25∗∗∗
(0.02)
-0.33∗
(0.19)
-0.34∗∗∗
(0.08)
0.62∗∗∗
(0.06)
0.00
(0.01)
1.68∗∗∗
(0.24)
4.54∗∗∗
(0.64)
0.67∗∗∗
(0.10)
-13.91∗∗∗
(3.10)
1109
(3)
Mean (Comb’II)
0.85∗∗∗
(0.14)
0.78∗∗∗
(0.08)
-2.35∗∗∗
(0.58)
(4)
q50 kb
0.56∗∗∗
(0.03)
0.27∗∗∗
(0.02)
-0.36∗∗
(0.17)
4.35∗∗∗
(1.23)
5.72∗∗
(2.40)
1.01∗∗∗
(0.22)
3.56∗∗∗
(0.56)
-23.90∗∗∗
(4.20)
0.27
1109
-16.79∗∗∗
(1.28)
1109
Standard errors in parentheses
∗
p < 0.10,
∗∗
p < 0.05,
∗∗∗
p < 0.01
the dependent variable for a one unit change in the independent variable (Buchinsky 1998). A
remarkable change can be seen for the coefficient on unemployment, which increases from -2.58
to -0.33. For the substantive interpretation, this means that for an unemployment rate increase
of 1 %, average MCPA participation would decrease by 2.58. However, predicting the median,
the decrease in participation from more unemployment would not be as substantial (0.32). Note
also how the coefficient on income changes to the theoretically expected positive and significant
relation, when predicting the median instead of the mean. Due to the right-skewed distribution,
OLS or the conditional-means function reflect to a greater extend what is going on in the upper
tail of the distributions (biased to predicting participation) rather than what is going on in the
Nature, Nurture or Neighbors?
75
middle of the distribution. This has lead to the differences we observe comparing the baseline
regression models with the ones obtained from median regression in table 3.10 Although there
are some changes in statistical significance and effect sizes of the coefficients, it nevertheless can
be seen that – in spite of the fact that 50% did not receive the treatment and had zero values –
the main results concerning the influence of Nature and Nurture still hold.
Count models Finally, the substantive interest lies in the explanation of MCPA participation
and, therefore, using the count of MCPA signatories in a county as the dependent variable is
another possibility. For data on counts, the Poisson distribution usually serves as a baseline
to represent the variability in the count data. However, the Poisson model mostly does not
fit in practice since in many applications the restriction that the conditional mean equals the
conditional variance does not hold (Long 1997). In such cases, the negative binomial regression
model is the right choice. Tests to account whether the Poisson model or the Negative Binomial
fit the MCPA count data at hand found significant overdispersion and suggested a better fit
of the Negative Binomial model as can be seen in figure 3.8. The observed proportion (count
of cities that have signed the MCPA) is well described by one that we would expect from a
negative binomial. Table 3.11 accordingly shows the results of a negative binomial regression on
the count of cities that have signed the MCPA in the county. The two combined models from
the baseline regression were estimated to allow for comparisons. To account for the different
compositions of counties, the number of cities in the county is also added as an explanatory
variable. Both models show significant overdispersion (i.e. σ 2 > 1) and justify the modeling
0
.2
Proportion
.4
.6
.8
choice.
0
2
4
6
8
10
Count MCPA
mean = .5744; overdispersion = 3.238
observed proportion
neg binom prob
poisson prob
Figure 3.8: Actual count distribution against negative binomial and poisson distribution
76
The results from table 3.11 generally reflect the results obtained from the simple OLS regression, with the one exception that the number of universities or colleges with liberal arts programs
is not statistically significant any more. All other explanatory variables bear the theoretically
expected signs and statistical significance levels. Of course, the substantive interpretation of the
coefficients changes in comparison to the OLS regression. For a unit change in the respective
explanatory variable xk , the expected count changes by a factor of exp(βk ), keeping the other
variables constant (Long 1997). As an example, going from the minimum value of % Democratic
vote to the maximum value while keeping the other variables at their median results in a total
effect for % Democratic vote of 1.54 on the expected MCPA count.
These results again strengthen my confidence in the obtained substantive results from the
OLS and spatial regressions. One additional test, which could deal with the high number of
counties with no MCPA signatories, is to estimate a zero-inflated negative binomial regression
model (Sattler & Bernauer 2011). These models explicitly model the occurrence of ’excess’ zero.
The process generating the zeros might, however, depend on other factors than the process for
strictly positive outcomes (Winkelmann 2008). Zero-inflated negative binomial models estimate
two equations, one for the count model and the other for predicting the excess zeros. This
modeling of the count outcome allows the researcher to specify a separate model to account for
the zeros. In the case of the MCPA participation, I would suspect that those variables that are
seen as disincentives for participation are the ones to predict excess zeros best. In table 3.12,
I therefore included CO2 emissions, the unemployment rate, whether the county is dependent
on industrial production, and whether there is coal production in the county as predictors in
the logit (inflation) part. The count models have not changed from the ones in table 3.11.
Again, the substantive direction and significance of the coefficients remains robust with the
exception of the coefficient on the unemployment rate, which becomes insignificant in the count
equation. Turning to the inflation equation predicting excess zeros, the theoretically derived
disincentives all perform well in predicting whether a county would be beyond the zeros. A high
unemployment rate as well as CO2 emissions increase the probability of observing a zero. All
measures bear the expected sign and are statistically significant. Also, the Vuong test to see
if the zero-inflated negative binomial is a significant improvement over the standard negative
binomial model (Long 1997) takes on a large positive value and a large z value, favoring the
zero-inflated negative binomial model over the standard negative binomial (Greene 2003).
Nature, Nurture or Neighbors?
77
Table 3.11: Negative Binomial Regression on the count of MCPA signatories in the county
(1)
Nature
0.16∗∗∗
(0.02)
-0.07∗∗∗
(0.02)
0.99∗∗∗
(0.11)
-0.00
(0.00)
0.02
(0.02)
Count MCPA
No.of cities
Climate Region
Coastal County
Property Damage
Fatalities
% 4yrs College
% Dem Vote
No. Lib Arts Uni/Coll
p.c. income
Median Age
Unemployment Rate
p.c. tonnes carbon
Population
Pop Growth (00-05)
% Manufacturing Empl
Coal Production
Industry dependent
% Loc/State Gov Empl
EPA non-attainment
-1.15∗∗∗
(0.13)
-0.28
(0.22)
0.18
-885.91
1109
Constant
lnalpha
Pseudo R2
Log lik.
Observations
Standard errors in parentheses
∗
p < 0.10,
∗∗
p < 0.05,
∗∗∗
p < 0.01
(2)
Nurture
0.08∗∗∗
(0.02)
0.03∗∗∗
(0.01)
0.05∗∗∗
(0.01)
-0.07
(0.04)
0.01
(0.01)
-0.01
(0.02)
-0.18∗∗∗
(0.05)
-0.03∗∗∗
(0.01)
0.00
(0.00)
0.02∗∗
(0.01)
-0.01
(0.01)
-0.00
(0.00)
-0.08
(0.16)
-0.02∗
(0.01)
0.22∗∗
(0.11)
-2.83∗∗∗
(0.78)
-1.00∗∗∗
(0.16)
0.28
-766.20
1080
(3)
Full
0.08∗∗∗
(0.02)
-0.00
(0.02)
0.36∗∗∗
(0.11)
0.00
(0.00)
-0.00
(0.04)
0.04∗∗∗
(0.01)
0.05∗∗∗
(0.01)
-0.06
(0.04)
0.00
(0.01)
-0.02
(0.02)
-0.16∗∗∗
(0.05)
-0.02∗∗∗
(0.01)
0.00
(0.00)
0.01∗
(0.01)
-0.00
(0.01)
-0.00
(0.00)
-0.08
(0.16)
-0.02∗
(0.01)
0.26∗∗
(0.12)
-2.47∗∗∗
(0.75)
-1.14∗∗∗
(0.16)
0.29
-761.00
1080
(4)
comb’I
0.10∗∗∗
(0.02)
(5)
comb’II
0.11∗∗∗
(0.02)
0.37∗∗∗
(0.11)
0.44∗∗∗
(0.11)
0.02
(0.02)
0.03∗∗∗
(0.01)
0.04∗∗∗
(0.00)
-0.03
(0.04)
0.01
(0.01)
-0.02
(0.02)
-0.14∗∗∗
(0.05)
-0.03∗∗∗
(0.01)
-2.77∗∗∗
(0.65)
-0.89∗∗∗
(0.19)
0.27
-793.18
1109
0.04∗∗∗
(0.01)
0.04∗∗∗
(0.00)
-0.02
(0.03)
-0.15∗∗∗
(0.04)
-3.49∗∗∗
(0.33)
-0.82∗∗∗
(0.20)
0.26
-799.07
1109
78
Table 3.12: Zero Inflated Negative Binomial Regression on the count of MCPA signatories in
the county
(1)
Combined I
Count MCPA
No.of cities
% 4yrs College
% Dem Vote
Unemployment Rate
No. Lib Arts Uni/Coll
Coastal County
Median Age
p.c. income
Fatalities
Constant
inflate
p.c. tonnes carbon
Unemployment Rate
Coal Production
Industry dependent
Constant
lnalpha
Constant
Log lik.
Voung test
Prob> z
Observations
0.11∗∗∗
0.04∗∗∗
0.04∗∗∗
-0.08
-0.02
0.44∗∗∗
(0.01)
(0.01)
(0.01)
(0.06)
(0.03)
(0.12)
-3.71∗∗∗
(0.39)
0.10∗∗∗
0.03∗∗∗
0.04∗∗∗
-0.07
-0.04
0.39∗∗∗
-0.02
0.01
0.03
-3.37∗∗∗
(0.01)
(0.01)
(0.01)
(0.06)
(0.03)
(0.12)
(0.02)
(0.01)
(0.02)
(0.75)
0.26∗
2.51∗∗
0.00∗
5.58∗
-26.23∗∗
(0.14)
(1.18)
(0.00)
(3.30)
(12.54)
0.26∗
2.51∗∗
0.00∗
5.47∗
-26.05∗∗
(0.14)
(1.17)
(0.00)
(3.26)
(12.44)
-0.88∗∗∗
-789.89
3.13
0.00
1109
(0.17)
-0.92∗∗∗
-787.83
3.12
0.00
1109
(0.17)
Standard errors in parentheses
∗
p < 0.10,
∗∗
p < 0.05,
∗∗∗
(2)
Combined II
p < 0.01
Nature, Nurture or Neighbors?
79
The zero-inflated negative binomial model translates the theoretical split between incentives
and disincentive for participation nicely into an empirical model. It further can account for the
difference in interpretation that came out of the logistic regression in the sense that the variables
measuring theoretically derived disincentives can explain the excess zeros and, therefore, the
high number of counties where no city has signed the agreement. Simultaneously, the negative
binomial count model predicts the number of MCPA signatories by focusing on the incentives
for participation plus the natural risk variables.
In conclusion, it can be said that the robustness section does not only provide clear evidence
in support of the general robustness of the initial findings concerning hypotheses 1-7, but also
shows that and how alternative specifications can fit the theoretical set up.
3.5 Conclusion
This paper set out to explain county-level participation rates in voluntary climate change initiatives throughout the U.S. by factors regarding a county’s natural (Nature) and socio-economic
and political (Nurture) characteristics, as well as their geographical surroundings (Neighbors).
It was hypothesized that voluntary participation in climate change mitigation efforts are highest when the county is at risk of being adversely affected by climate change, when there are
favorable socio-economic and political conditions present, and when the surrounding counties
are also strong participators. I have tested these claims for the MCPA participatory behavior
of cities in counties within the 48 contiguous states in the period 2005-2008.
As far as the natural system characteristics are concerned, being a coastal county seems to
drive signatory behavior. For socio-economic and political characteristics, drivers of participation are a well-educated and liberal population as well as a high percentage of democratic voters
within the county. Furthermore, the presence of universities significantly increases the relative
number of signatories. Disincentives for participation include especially bad economic conditions in a county. Depending on model specification, proxies for potential abatement costs are
also significantly related to participation. Findings from spatial analysis indicate some evidence
of spatial interdependence in participation, meaning that some of the variation in a country’s
participation level can be attributed to participation levels in connected countries. However, of
the four different specifications of geographic influence tested, only two support the conjecture
of a significant effect of spatial interdependence.
80
How can we interpret these ambiguous results on the spatial interdependence in participation?
First, it generally seems that most of the spatial clustering in the dependent variable is explained
by the independent variables other than the spatial lag. This is positive news for the Nature
and Nurture components. Their influence on MCPA participation has also been shown to
be very robust to model specification and choice of estimation method. With respect to the
spatial dependence that prevailed in two of the spatial models, this effect is either due to true
interdependence between the units or caused by unobserved heterogeneity (Franzese & Hays
2007). This distinction amounts to the essence of Galton’s Problem (Franzese & Hays 2007; Jahn
2006; Braun & Gilardi 2006; Pluemper & Neumayer 2010), namely to the difference between true
interdependence in decision-making as opposed to spatial clustering or unobserved heterogeneity.
From this analysis, a first indication was attained that interdependence has to be taken further
into account in future research on MCPA participation. To adequately model interdependence
in climate change decision-making, research at the city level as well as research including the
temporal dimension is warranted.
This paper is the first contribution looking at the determinants of participation in the Mayors’
Climate Protection Agreement throughout the 48 states. Evidence from the large-N analysis
conducted in this paper represents an important step forward in understanding the motivations
behind voluntary climate change actions.
Nature, Nurture or Neighbors?
81
3.6 Appendix
Table 3.13: Summary statistics
Variable
MCPA
% MCPA
Count MCPA
% Dem Vote
p.c. income
% 4yrs College
Uni/Coll
Unis scaled by cities
Lib arts scaled by cities
No. Lib Arts Uni/Coll
Population
Pop Growth (00-05)
Loc/State Gov Empl
Unemployment Rate
Industry dependent
Median Age
p.c. tonnes carbon
Pop Loss 1980-2000
% Mining
Manufacturing Empl
EPA non-attainment
Property Damage
Climate Region
Fatalities
Coastal County
%Dem x Income
%Unemp x %Dem
Mean
0.3
18.2
0.6
42
30.4
20.9
0.6
0.7
0.6
1.1
22.3
4.6
14.7
5.2
0.3
36.4
7.3
0.1
0.7
11.5
0.4
69.4
7.5
0.7
0.2
1301.8
223.6
Std. Dev.
0.5
33.7
1.6
12.2
7.2
8.7
0.5
1.8
1.1
2
48.1
7.2
30.6
1.6
0.5
3.9
14.6
0.3
2.1
24.4
0.5
464.6
3.3
2.5
0.4
613
104.3
Min.
0
0
0
7.1
12.2
6.9
0
0
0
0
1.1
-11.3
0.9
2.3
0
22.2
1
0
0
0.1
0
0
1
0
0
117.1
19.2
Max.
1
100
22
89.2
93.4
63.7
1
39
20
29
994.8
46.9
540.5
16
1
52.9
258.1
1
26.5
504.5
1
6136.7
14
39
1
7664.4
956.8
N
1109
1109
1109
1109
1109
1109
1109
1109
1109
1109
1109
1109
1086
1109
1109
1109
1109
1109
1109
1080
1109
1109
1109
1109
1109
1109
1109
82
Table 3.14: Pattern of signature of MCPA in all states
MCPA
1
0
Total
State name
No.
%
No.
%
No.
%
Alabama
6
1.0
46
2.2
52
1.9
7
1.1
26
1.3
33
1.2
Arizona
3
0.5
29
1.4
32
1.2
Arkansas
112
17.7
239
11.6
351
13.0
California
4
0.6
30
1.4
34
1.3
Colorado
11
1.7
14
0.7
25
0.9
Connecticut
1
0.2
2
0.1
3
0.1
Delaware
1
0.2
0
0.0
1
0.0
District of Columbia
70
11.1
78
3.8
148
5.5
Florida
8
1.3
53
2.6
61
2.3
Georgia
2
0.3
14
0.7
16
0.6
Idaho
34
5.4
168
8.1
202
7.5
Illinois
11
1.7
59
2.9
70
2.6
Indiana
12
1.9
22
1.1
34
1.3
Iowa
5
0.8
28
1.4
33
1.2
Kansas
5
0.8
28
1.4
33
1.2
Kentucky
3
0.5
31
1.5
34
1.3
Louisiana
6
1.0
4
0.2
10
0.4
Maine
8
1.3
14
0.7
22
0.8
Maryland
25
4.0
24
1.2
49
1.8
Massachusetts
20
3.2
72
3.5
92
3.4
Michigan
25
4.0
58
2.8
83
3.1
Minnesota
2
0.3
35
1.7
37
1.4
Mississippi
14
2.2
50
2.4
64
2.4
Missouri
4
0.6
3
0.1
7
0.3
Montana
3
0.5
13
0.6
16
0.6
Nebraska
4
0.6
4
0.2
8
0.3
Nevada
7
1.1
6
0.3
13
0.5
New Hampshire
41
6.5
84
4.1
125
4.6
New Jersey
4
0.6
14
0.7
18
0.7
New Mexico
25
4.0
65
3.1
90
3.3
New York
19
3.0
44
2.1
63
2.3
North Carolina
2
0.3
7
0.3
9
0.3
North Dakota
20
3.2
146
7.1
166
6.1
Ohio
5
0.8
35
1.7
40
1.5
Oklahoma
10
1.6
31
1.5
41
1.5
Oregon
8
1.3
63
3.0
71
2.6
Pennsylvania
3
0.5
5
0.2
8
0.3
Rhode Island
5
0.8
28
1.4
33
1.2
South Carolina
1
0.2
8
0.4
9
0.3
South Dakota
4
0.6
44
2.1
48
1.8
Tennessee
21
3.3
167
8.1
188
7.0
Texas
1
0.2
34
1.6
35
1.3
Utah
1
0.2
2
0.1
3
0.1
Vermont
9
1.4
28
1.4
37
1.4
Virginia
24
3.8
34
1.6
58
2.1
Washington
1
0.2
13
0.6
14
0.5
West Virginia
14
2.2
59
2.9
73
2.7
Wisconsin
0
0.0
8
0.4
8
0.3
Wyoming
Total
631 100.0 2,069 100.0 2,700 100.0
Nature, Nurture or Neighbors?
83
Text
(a) MCPA participation rates in counties, calculation based on quartiles, darker shade indicates more participation
Text
(b) fatalities of natural hazards; calculation based on quartiles; darker shade indicates higher number of fatalities
Figure 3.9: Comparison of the pattern of % MCPA and fatalities from natural hazards in the
1109 counties
84
Text
(a) %Democratic vote in counties; brown 3 50%; darker brown 3 70%
Text
(b) unemployment rate in counties; calculation based on quartiles; darker shade indicates higher unemployment
Figure 3.10: Comparison of the pattern of % Democrats and unemployment rate in the 1109
counties
Nature, Nurture or Neighbors?
85
HH
HH
HH
HH
HH
HL HHHHHH
HH
HH
HL
HL
HL
HHHH
HLHL
HL
HH HH
HHHH
HL
HH
HHHH
HH
HH
HHHH
HH
HHHHHH
HH
LH
HH
HHHH
HL
HL
Text
HH
HL
HL
HL
HH
HHHH
HHHH
HH
(a) Local Moran’s I for participation rates in MCPA
LH
LH HL
LH
HH HH
HH
HH HH
HH
HH HH
HH
HH
HH HH
HH
HH
HH
HH
HHHH HH
HHHH
HL
HL
Text
HH
HH
HH
LH
HH
HH
HH
HH
HH
HH
HH
HH
HH
HH
HH
HH
HH
(b) Local Moran’s I for fatalities of natural hazards
Figure 3.11: Analysis of spatial clusters with Local Moran’s I statistic (W: county contiguity)
in the 1109 counties
86
LH
HL
LH
HH
LH
LH
LL LL
LH
LL
LL
HH
HH
LL LH
LL
LH
LH
HH HH
HHHH
HHHL
HH
HL
LL
HH HH
HH HH
LH LH
LH
LL
LL
LLLL
LLLL
LLLL
HL
LL
LLLL
HHHH
LH
HHHH
LL
LH
LH
HH HH
HH HH LH
LH
LH
HH HH
HHHH
LLLL
LHLH
LH
HH LLLL
HL
LL
Text
LH
LL
LL LL
LL
LL
LL
LL
LL
LL
LL
HH
HH LH
HH
LL
HL
LL
LL LLHH
LL
LH
HH
LH
LL
LH
HL
LL
LH
HHHH
(a) Local Moran’s I for %Democratic vote in counties
HH
HH
HH
HH
HH
HH
LL
LH
HHHH
HH
LH LH
LH
HH HH
HHHH HH
LL
LL
LL
HH
HHHH
HH
LH
LH
LL
LH
LL LHLH
LLLL
LH
LH HL
LL
HH LL
HH
HH
LL
LL HL LH
LL
LH
LL
LL
LL
LH
HLLL
LL
LLLL
HH
HH
LH
LL
LL
LH
HH
LH LH
HHHH
LH
HL
LL
LLLL
HL
LLLL
HH
HHHH
Text
LL
HL
LL LH
LL
HH
HH HH
HH
LHHL
HH
HL
HL
LL
HL
HHHH LL
HH HH
LH
LH
HH
LL
LL
(b) Local Moran’s I for unemployment rate in counties
Figure 3.12: (Cont’d)Analysis of spatial clusters with Local Moran’s I statistic (W: county contiguity) in the 1109 counties
87
4 Voluntary Climate Change Initiatives in the U.S.: Testing
Participation in Space and Time
Lena Maria Schaffer
4.1 Introduction
Global climate change is one of the biggest challenges facing mankind in the 21st century. Prominent attempts to deal with global climate change and to mitigate its consequences have focused
on multilateral cooperation and international institutions. The success of these international
efforts has been limited. A prime example is the Kyoto protocol. Particularly large emitters
of greenhouse gases, such as the United States, which refused to ratify the Kyoto protocol, are
generally reluctant to contribute to this global environmental effort. Given that effective cooperation between nation-states in climate politics is very difficult, the question arises whether
there are other possibilities for coping with global climate change at lower political levels. In
fact, such climate policy efforts already exist.
The largest scale effort of this kind is the U.S. Mayors Climate Protection Agreement (MCPA).
Its aim is to advance the goals of the Kyoto Protocol on the city level even if the U.S. federal
government lags behind. As of November 2010, 1044 cities have signed the agreement, representing a population of 80 million, effectively more than one quarter of the U.S. population.On
February 16, 2005, the Kyoto Protocol came into force for the 141 countries that had ratified it.
On the same day, the then Seattle Mayor Greg Nickels launched the initiative called the ’U.S.
Mayors’ Climate Protection Agreement’ to advance the goals of the Kyoto Protocol through
leadership and action by American cities. Joining the initiative constitutes a voluntary act by
the local government that signs the agreement. The MCPA stipulates that mayors should ’strive
to meet and beat the Kyoto Protocol targets in their own communities’, as well as expects local
governments to ’[u]rge their state governments, and the federal government, to enact policies
and programs to meet or beat the greenhouse gas emission reduction target’ (Mayors Climate
Protection Center 2010). Originally Mayor Nickels planned to mirror the Kyoto Protocol and
88
have at least 141 Mayors throughout the U.S. sign the initiative. By the June 2005 annual conference of the United States Conference of Mayors (USCOM) in Chicago, this goal was already
reached. At this meeting, the MCPA was formally endorsed by USCOM. The map in figure
4.1(a) shows the first 141 signatories. Over the past 5 years, this initiative has grown immensely
to now incorporate the 1044 participating cities in figure 4.1(b).
!
5
!
5!
!
5
5!
!
5
5
!
5
55!
!
!!
5
5
!
5
!
5
!
5
5
!!
5
!
5
!
5
!
5
!
5
!
5
!
5
!
5
!
.
!
5!
5
!
5
!
.
5
!
5 !
!
5
!
.
5
!
5!
!
5
!
5
!
.
!
.
!
5
!
5!
5
.
5!
!!
55
!
5!
!
5!
5!
5
!
5
!
5
!
5
!
.
!
.
!
.
!
5
!
5
!
5
!
5
!
.
!
5
!
5
!
.
!
.!
. !
.
!
.
!
..
!
.
!
.
!
.!
!!
.
.!
.
!
.
!
.
!
.
!
.
!
5
!
5
!
5
!
5
5
!
5 !
!
5
Text
!
5
!
5
!
.
.!
!
.!
.!
!!
.
.
.
.
!
. !
!
.
!
.
!
5
!
5
!
.
!
5
!
5
!
5
!
.
!
5
!
5
!
.
!
.
!
.
!
.
!
.
!
.
.!
!
.
(a) June 2005
(b) June 2010 (Source: USCOM)
Figure 4.1: Cities that have signed the Mayors Climate Protection Agreement
Since the subnational formation of climate policy institutions is a new phenomenon, we know
very little about the conditions that motivate local governments to join such efforts. Such
knowledge is important for the evaluation of whether cooperation at this level can offer a fea-
Testing Participation in Space and Time
89
sible substitute or, at least, a useful complement to global efforts. This paper examines the
determinants of cities’ willingness to join the MCPA in cities over 10,000 inhabitants in seven
Midwestern states. I develop arguments on how various factors influence local government decisions to voluntarily contribute to climate change mitigation efforts by joining the Mayors Climate
Protection Agreement. These factors include community-specific internal factors and external
factors accounting for cities’ interdependence in decision-making. As far as internal factors are
concerned, I argue that there are specific determinants leading to a higher demand for voluntary climate change policies within the city, e.g., a higher percentage of voters that support
the Democratic Party, which then make governments more likely to participate in voluntary
efforts. I further assume that the initiative to join the MCPA can also emanate from the supply side of local government. Here, the existence of a policy entrepreneur and a mayor-council
form of government are considered to increase the propensity to participate. Interdependencies
between cities are conceptualized as emanating from geographic linkages to other jurisdictions,
peer group effects, as well as from participation in social networks. My theoretical framework
is complemented by the consideration of certain events that are assumed to be associated with
a city’s decision to sign on to the MCPA.
To test my claims, I use a unique new data set on monthly signing behavior of 749 Midwestern
cities from February 2005 to June 2010. I find that a city’s participation in social networks as
well as the MCPA signing behavior of cities within the same population group increases a city’s
likelihood of participating. Although city-internal characteristics, such as the % of Democratic
votes in the city or human capital endowment, are also significantly related to a higher propensity
to sign, their substantive effects are much smaller. As far as pivotal events are concerned, I find
that the likelihood of participation significantly increases in the months leading up to a mayoral
election in the city. Insights gained from this large-N study are further supported by qualitative
evidence from an analysis of questionnaires sent to a subsample of the 749 Midwestern cities.
This paper seeks to contribute to the scholarly literature in at least three ways. First, existing
research on voluntary climate change initiatives emphasizes city-internal characteristics. This
paper considers the behavior of other cities and their effects on the dynamics of MCPA adoption.
To this end, I present a coherent framework of municipal voluntary climate change adoption that
is based on both internal and external characteristics and that considers the different channels
through which cities may be influenced by the choices of others. Hence, the main focus of
this paper is on the interdependence in the decision-making between different jurisdictional
90
units vis-a-vis internal factors. For many phenomena within social sciences, it is very difficult
to argue that we are dealing with independent observations. Far more often, policy choices
by governments depend on previous decisions of other governments. This is mostly referred
to as Galton’s Problem 1 (see Jahn 2006; Braun & Gilardi 2006). Second, this study is the first
comprehensive large-N inquiry into the determinants of MCPA adoption over time. By including
a temporal dimension, I am not only able to test different internal and external influences that
are supposed to matter for the participation in voluntary climate change policies, but to also
account for the importance of specific events that influence the likelihood of MCPA adoption.
Third, this paper complements large-N statistical evidence with qualitative evidence on cities’
motives for joining the MCPA.
The paper is structured as follows. I first provide an overview of the existing literature on
voluntary climate change initiatives and define my own contribution to this literature in section
two. In the theoretical section that follows, I discuss theoretical problems and elaborate on my
arguments before introducing the determinants of voluntary climate change initiative adoption
and presenting testable hypotheses. In section four, I then introduce my research design and the
operationalization of the main variables before presenting the results from my statistical and
qualitative analyses, and robustness checks. The paper concludes with a discussion of the main
empirical findings and suggests further extensions.
4.2 Background on Voluntary Climate Change Policies in the U.S.
This section gives a background on the ever-increasing literature on voluntary climate change
policies in the U.S. and carves out the contribution of this paper.
Research on subnational climate change policies in the U.S. has so far focused primarily on
state level activity. Scholars have examined the determinants, as well as the implications of state
policies in a multilevel governance setting (Rabe 2004, 2008; Schreurs & Epstein 2007). Rabe
(2008) examines states’ strategic choices in climate change mitigation policies with reference to
their mix of emission trends and policy adoption rates. Recognizing the importance of states’
policies as a political feasibility test for subsequent national policies, he concludes with a set
of scenarios on how to integrate subnational efforts into a functioning multilevel governance
1
Galton’s Problem arises when the independence of the units of analysis cannot be regarded as given and
potential outcomes could be linked to interdependent behavior between these units. Any conclusions drawn
from such a setting risks bias. For a historical account of the implications of this problem, see Franzese &
Hays (2007)
Testing Participation in Space and Time
91
framework. Schreurs & Epstein (2007) further raise questions concerning federalist practices
with respect to the future leadership role in climate change policies. More recent publications
look at whether and how state and local government initiatives will continue in the wake of
potential national action (Engel 2009). She concludes that although it remains to be seen what
happens to existing cap and trade regimes, state and local government leadership in climate
change policies will likely continue and intensify.
While Rabe (2004) has provided a careful categorization of the different approaches of state
climate change policy with reference to national policy, Lutsey & Sperling (2008) go one step
further and try to quantify the potential impacts of lower level government commitments in the
U.S.. They focus on the consequences of decentralized climate change policies on actual GHG
emissions, exploring the development of these emissions based on current inventories and chosen
policies. They find that if the 17 states that have set an GHG emission-reduction target were to
achieve these targets, U.S. GHG emissions would be stabilized at 2010 levels by the year 2020
(Lutsey & Sperling 2008, 683). The question of whether bottom-up activity can substitute for
absent national level climate policy is at the core of many studies. Selin & VanDeveer (2007,
22) arrive at differing conclusions concerning the significance of emission reductions emanating
from these subnational policies. They stress the importance of such programs as they allow
policy-makers to ’[. . . ] see which of the many available policy options are gaining support in
the public and private spheres’ and thereby are most likely to influence future federal policy
development.
Turning now to the impact of cities, Kousky & Schneider (2003) compare the policies of
twelve Cities for Climate Protection (CCP) members and the CO2 reduction achieved from their
implemented policies. Their results indicate that most reduction targets are fairly modest, and
they express concern that after picking the proverbial ’low hanging fruit’, free-riding problems
might re-emerge and bring mitigation efforts to a halt. Also looking into the feasibility of
climate action at the city level, Tang et al. (2010) analyze 40 local climate action plans to
explore how well they grasp the concepts of climate change and prepare for climate change
mitigation and adaptation. In their view, climate action plans indicate a serious consideration
and commitment for climate change mitigation. They find that although these plans reflect a
high level of environmental awareness, they have only limited effects on emissions. More recently,
Krause (2011b) collected information on the GHG reduction activities in 53 cities in Indiana
to gauge whether implementation of such policies is higher in communities that had committed
92
themselves to an agreement like the MCPA. One of the results from her study is that while
policies that have important co-benefits (e.g., curbside recycling) are very commonplace, more
formal and explicit climate policies are lacking. Furthermore, when considering a specific climate
protection focus (N=17), she does not find a significant difference between MCPA members and
non-members, which leads her to the conclusion that participation is only symbolic. Although
presenting evidence from one U.S. state to view MCPA participation as symbolic is a huge
improvement for research on implementation, we still do not know why some communities chose
to proclaim their efforts by joining a national framework of mayors, while others obviously
chose not to. This paper seeks to take a step forward in explicitly considering what determines
communities’ adoption of national efforts such as the MCPA.
Among the pioneering research concentrating on cities as units of analysis and their adoption
of voluntary climate change efforts, is the contribution by Betsill (2001). She studies the opportunities and obstacles with respect to the Cities for Climate Protection (CCP) program in the
United States. This program began in 1991 and was the first of its kind to introduce the topic
of climate change at the city level (Betsill & Bulkeley 2006; Betsill 2001). Most of the scholarly
work on local climate change action has thus far centered on the CCP initiative (Betsill 2001;
Betsill & Bulkeley 2004; Brody et al. 2008; Kousky & Schneider 2003; Lee 2009a; Lindseth 2004;
Vasi 2006; Zahran, Grover, Brody & Vedlitz 2008).2 Based on interviews with policy-makers
and qualitative case study evidence, Betsill (2001) argues that in order for cities to successfully
participate in this program, the issue of global climate change is best framed as a local issue.
Along the same lines, Bulkeley and Betsill (2002: 184-85) emphasize the value of networks and
local capacity for action in addressing climate change, and highlight the importance of ’the way
climate protection is framed, particularly in relation to economic incentives’. Lindseth (2004)
takes a more critical view of this result and objects to the pragmatic CCP policy of framing
climate change policies as being about solving problems locally and enjoying local benefits. In
his view, local actors should be more aware of the various dimensions of climate change.
Another important part of the literature seeks to understand the motivations for the public’s as
well as of decision-makers’ preferences for local climate change initiatives. Kousky & Schneider
(2003) use interview evidence from officials in 23 CCP municipalities across the U.S. to see
what makes communities voluntarily implement these efforts. They find that implementation is
generally not driven by demand from the population, but is rather a top-down issue initiated by
2
An interesting study that does not deal with a specific climate change initiative, but also with local environmental institutions, is the one by Meyer & Konisky (2007) on Massachussetts Wetlands Protection
Testing Participation in Space and Time
93
policy-makers. Also, the cost savings and local co-benefits from climate change policy matter for
participation in the CCP. By modeling the public good provision in the realm of climate change
policies across two levels – national and local – Urpelainen (2009) finds that local policymakers
have an informational advantage concerning local political benefits. These unobservable benefits
lead politicians to overcome the collective action problem and pursue local climate policy. As
an example of potential political benefits for the decision-maker, Urpelainen (2009, 87) refers
to energy efficiency requirements that would favor producers using advanced technologies. In a
similar vein, Engel & Orbach (2008) do not specifically concentrate on one initiative, but give
a comprehensive overview of U.S. state and local level initiatives. To explain participation in
the initiatives, they point to and discuss potential supply and demand side explanations. On
the demand side, they see the potential for participation as a symbolic statement against the
inactivity of the U.S. government as one motivation, but other sources for participation may be
altruism (’warm glow’), as well as the localization of global problems. On the supply side, they
(similar to Rabe (2004); Krause (2011b)) stress the role of political, as well as administrative
entrepreneurship in promoting such initiatives. Furthermore, they also discuss the possibility
that communities might present certain municipal policies as having a climate change component
due to the popularity of actions on climate change. This is contrary to what earlier studies
(Betsill 2001; Lindseth 2004) have found in the CCP context, and I will further elaborate on
this issue below. Despite the plethora of different possible reasons for climate policy action,
Engel & Orbach (2008) do not empirically test their claims.
While the research mentioned so far relies primarily on case study evidence or interviews with
officials, systematic quantitative large-N analyses of the subject are still rare and predominantly
focus on the CCP program (Vasi 2006; Brody et al. 2008; Zahran, Grover, Brody & Vedlitz 2008;
Lee 2009a). Drawing on research on organizational innovations, Vasi (2006) combines qualitative
and quantitative methods to examine how organizational environments and framing processes
jointly influence the diffusion of organizational innovations in the context of local climate change
policies within the CCP program. Employing event history analysis, his two key findings are that
adoption of the CCP program is determined by spatial or cultural proximity to earlier adopters,
and that organizational embeddedness in transnational frameworks also fosters participation.
Lee (2009a) uses a multilevel setting to analyze which cities participate internationally in the
CCP and other networks such as C40. Controlling for city and country-level variables, he finds
that cities’ positions in the global economy, and transportation hub characteristics significantly
94
influence participation in these networks. In a similar paper, Zahran, Grover, Brody & Vedlitz
(2008) study metropolitan areas in the U.S. and explain their participation with respect to
physical location, natural capital, production and transportation modalities, and socioeconomic
characteristics. They find that those metropolitan areas that contribute most to the problem
in terms of CO2 emissions are least likely to engage in the CCP, while those with high levels of
civic capacity are more likely to engage. An interesting finding from their study is that natural
climate change risk does not seem to matter for metropolitan engagement in the CCP. However,
in a similar study, Brody et al. (2008) use county-level participation in the CCP and find that
natural risk plays a large role in explaining participation in the CCP. While Brody et al. (2008)’s
study is one of the first large-N studies examining participation in the CCP, they approach the
issue from an urban planning background and pay only very scant attention to political and
economic determinants of climate policy adoption.
As regards more recent climate change initiatives, such as the mayors climate protection
agreement, few authors have conducted in-depth examinations. Clearly one of the first scholarly
writings on this newer phenomenon is Warden (2007)’s dissertation on the initial years of the
MCPA and the importance of networks for the proliferation of local climate policies. Through
document analysis and interviews with 17 officials (mayors, city officials, representatives from
relevant organizations), she was able to find ten themes of municipal motivation. While some
officials claimed that there was a moral urgency to act on the issue due to national inaction,
others also hinted to economic incentives due to energy efficiency policies. Interestingly, she
also finds evidence for competition between neighboring cities to be ’green’, and also stresses
the importance of the mayor’s characteristics and his leadership role on the issue. Concerning
the MCPA movement as a whole, there is another recent article by Krause (2011a), where she
looks at the adoption of the MCPA in a multilevel setting, taking states’ climate policies into
account. She looks at 1026 cities over 25.000 inhabitants and finds that state-level characteristics cannot explain the signing of the MCPA by a city. Instead, internal characteristics, such as
income, education, votes for the democrats, and if there are participating neighbors are important determinants for a city to sign. This is the first study to hint at the existence of possible
interdependence or diffusion effects in MCPA participation. Furthermore, the substantive result
that most of the variation in cities’ participatory behavior can be attributed to city- or countyinternal variables is a great step towards a better understanding of cities’ participation patterns.
However, there are some shortcomings. First, one shortcoming that Krause (2011a, 13) also
Testing Participation in Space and Time
95
addresses is that the model tested in this study is entirely cross-sectional, while the process
being studied is dynamic. Berry & Berry (1999) recommend the use of panel data or event
history analysis as the most appropriate methodology when trying to model policy adoption. A
second and related point, concerns the time period under study, which is the three years up to
May 2008. The choice of May 2008 seems to be due to practical reasons, however the initiative
grew substantially afterwards. A less arbitrary time point to use for such a cross-section might
have been the election of President Barack Obama in November 2008, since it ended the time
of certainty of no national climate change action (Schaffer 2010).
In conclusion, despite notable systematic studies on the subject (Brody et al. 2008; Vasi 2006;
Krause 2011a; Zahran, Grover, Brody & Vedlitz 2008), there are still large gaps in the literature.
This paper seeks to contribute to the scholarly literature in at least three ways. First, existing
research on voluntary climate change initiative emphasizes city-internal characteristics. This paper considers the behavior of other cities and their effects on the dynamics of MCPA adoption.
To this end, I present a coherent framework of municipal voluntary climate change adoption
based on city internal and external characteristics, considering the different channels through
which cities may be influenced by the choices of others. Second, using the first multi-state panel
design, this paper can also control for temporal variation in adoption and the importance of
specific events on the likelihood of adoption. Third, this paper complements large-N statistical
evidence with qualitative evidence on cities’ motives for joining the MCPA gained from questionnaires. In the following section, I will first develop my theoretical argument, starting with a
discussion of the theoretical problem. In a second part, the main hypotheses will be introduced.
4.3 Theoretical Framework
4.3.1 Puzzle and Theoretical Argument
Why should policymakers in local communities tackle problems related to climate change and
what is the substantive puzzle underlying this paper? First and foremost, environmental quality,
with respect to climate change, is a global public good (Hardin 1968). The production of GHG
emissions, irrespective of the location, increases the risk of adverse climatic changes, whereas
a reduction of GHG emissions contributes to the mitigation of such risks (Baettig & Bernauer
2009). The standard prescription for mitigating GHG emissions, and thus combating global
warming, is a concerted global effort combined with implementation of internationally agreed
96
measures at the level of the nation-state. In a more general way, the scholarly literature on
environmental regulation stipulates that when environmental effects are felt in the vicinity of
the source, local governments should be best positioned to control the problem. However,
’when effects are not localized, as with carbon dioxide and other precursors of climate change,
centralized control may be more appropriate’ (Kolstad 1996, 349). According to this argument,
GHG emissions should be tackled at the federal level. Therefore within the non-binding national
framework present within the United States up until November 2010, cities should have faced
a strong incentive to free-ride on the mitigation efforts of others. This describes a standard
collective action problem (Olson 1965).
Second, concerning the interjurisdictional competition within the United States, cities act
as ’growth machines’ (Molotch 1976) in order to sustain themselves. Local governments are in
competition with each other for resources and should, therefore, not voluntary burden themselves
with potentially costly tasks. In line with this notion and according to Oates (2002, 7), local
government cannot be ’entrusted with the responsibility for setting environmental standards,
because it will sacrifice the environment on the altar of economic development’. We can therefore
conclude that neither is local government the ideal level of government for tackling climate change
mitigation policies, nor do local governments have strong incentives to withstand free-riding on
the efforts of others to provide the public good. This is what conventional theory of public
good provision would predict when climate change is viewed as a global public good. However,
empirical observations show that actually, a lot of local governments are taking part in voluntary
initiatives and embrace local climate change policies. So, to what extend can climate change
policies be an issue at the local level? What are the reasons why cities engage in climate change
initiatives such as the MCPA? Finally, why is participation in the MCPA commonplace despite
the convincing theoretical arguments that predict the absence of local action on climate change
issues?
First, regarding the nexus between climate change and cities, cities are responsible for the
bulk of worldwide GHG emissions. According to Kamal-Chaoui & Robert (2009), cities consume
about 60-80% of energy production worldwide and contribute a similarly large share to global
CO2 emissions. Second, not only do cities contribute to the problem, they also have key competencies to act on climate change. Some of the core competencies of local government include
issues of land-use planning, energy supply, transportation, and waste management. All of these
are essential domains of environmentally sustainable policies and central to any climate change
Testing Participation in Space and Time
97
mitigation strategy. In this context, Glaeser & Kahn (2010) stress the impact of cities’ land-use
regulations on CO2 emissions. Traffic management can also be important as arrangements with
less congestion (e.g., introducing congestion charges) or reduced start and stop times at intersections (e.g., with roundabouts (Rouphail et al. 2006)) lead to lower CO2 emissions. Therefore,
it can be established that cities do have the possibility to actually enact relevant mitigation
policies at the local level. Furthermore, recent research on the local effects of CO2 in cities,
Jacobson (2010) looks at so called domes of high CO2 levels that build over U.S. cities. He finds
that although CO2 is usually said to have the same effect on global temperatures no matter
where it was emitted, high levels might worsen the effects of localized air pollutants like ozone
and particulates (Jacobson 2010). This would indicate that strategies to curb GHGs might also
lead to better air quality with respect to localized pollutants. There are many additional links
between climate change policy and sustainable development that might drive the development
of innovative policies at the local level. Communities may well conceptualize climate change as
having a public health dimension. The possibilities for cities to implement climate policies and
enjoy co-benefits with regards to energy efficiency, improved public health, and general quality
of life are manifold. In fact, that is what Betsill (2001)’s research on the Cities for Climate
Protection (CCP) campaign suggests, namely that cities should frame climate change as a local
issue, utilizing links between climate change policy and issues already on the local agenda.
From the above, it can be established that on the one hand, cities potentially incur costs from
enacting climate change policies due to their competitive position vis-a-vis other communities
that free-ride on their efforts. On the other hand, it can be shown that climate change policies
are complementary to many efforts on the agenda of local governments, and that possible cobenefits from engaging in those policies exist. Notwithstanding all the possibilities presented so
far, and given that it can make sense for local governments to pursue climate change policies,
especially when framing them as a local issue of concern (Betsill 2001), the question remains as
to why local government would want to frame policies as being climate change policies in the
first place?
As mentioned in section 4.2 above, there is no consensus in the scholarly literature on one allencompassing reason for why local communities act on climate change. A plethora of different
factors are assumed to be at play, and many of them have more to do with the specific situation
in the U.S. during the time the MCPA began to evolve than providing a universal explanation
of communities’ climate change mitigating behavior. Generally, I assume that two different
98
sources of motivation stand out, substantive environmental concerns, and symbolic participation.
Whereas some cities and mayors are genuinely concerned about the environment, symbolic
participants only join because of the political statement their participation makes, and the
political benefit it brings. Due to the agreement’s flexibility in timing and scope with respect
to the implementation of local climate change policies, one of the reasons for joining could be
a symbolic statement (Engel & Orbach 2008) against the current national policies, be it in the
realm of climate change policies or elsewhere. Environmental concern and symbolic participation
might of course also overlap. Furthermore, environmental concern could be motivating the
community to pressure their local government to sign, while the government’s reason could be
symbolic participation alone.
Separating these different motivations is difficult, and it has to be assumed that, more often
than not, both motivations are present simultaneously. Therefore and unless otherwise specified,
I consider them both to be at work throughout the following theoretical discussion. Of course,
empirically separating one from the other is just as daunting a task. Disentangling these different
motivations necessitates two things: first, qualitative evidence from decision-makers as well as
communities, and second, further research into the extent of actual implementation of climate
change policies at the community level (Krause 2011b; Tang et al. 2010). In section 4.5.2, I will
address the first of these two points, presenting qualitative evidence from questionnaires that
were sent to mayors throughout the U.S.
My theoretical argument derives from the above observations and concentrates on the local
government as the decision-making body that either choses to join an initiative or not. I assume
that the existence of political benefits from joining voluntary climate change initiatives is the
driving factor. Whether the political benefits are gained by sincerely acting on the climate
change cause or merely by making a symbolic statement through participation in climate change
initiatives is considered to be largely unobservable (not only in the context of a large-N analysis).
However and independent of the ’channel of motivation’, the extent to which political benefits
can be reaped by the local decision-maker is theorized to depend on conditions within the city,
external factors, as well as on initiative- and time-specific factors.
The local and city-specific factors that matter for the signing the MCPA are assumed to
emanate either from the demand side (the community) or the supply side (the city government).
For both of them I discuss factors that make participation more likely, as well as factors that
are supposed to make signing the MCPA less likely.
Testing Participation in Space and Time
99
Given my overarching research framework that deals with the importance of interdependent
choice in joining voluntary climate change initiatives, I then introduce external factors that are
assumed to facilitate the diffusion of voluntary climate change policies. I argue that there are
three main channels that matter for the interdependence of cities: neighbors, peers, and social
networks.
As already noted above, a unique feature of this paper is that it provides the first examination
of MCPA participation in a large-N context and over time. Hence, a more careful investigation
into the time period and the events occurring in it is possible. I argue that alongside – and
conditional upon – internal and external factors, there are pivotal events, such as the presidential
election of Barack Obama, that had an impact on cities’ signatory behavior.
In the subsequent section, I introduce the three groups of determinants for the adoption
of voluntary climate change initiatives and develop hypotheses about their effect on the main
outcome of interest – participation of cities in voluntary climate change initiatives.
4.3.2 Determinants of Adoption of Voluntary Climate Change Initiatives
The previous section introduced the theoretical puzzle that follows from the nature of the climate
change issue and the level of government in a federal system. I have then considered alternative
explanations as to why cities or local governments might be interested in pursuing climate change
policies. This section presents the conditions that are supposed to have lead to the patterns of
local climate change policy adoption that we currently observe. It also asks what the factors
leading to participation in climate change initiatives are.3 I argue that to explain the temporal
and spatial patterns in participation in voluntary climate change agreements three factors are
decisive: city-internal characteristics, external factors, and time and initiative-specific events.
As this is the first comprehensive study of cities’ participation in the Mayors Climate Protection
Agreement over time, I introduce each of these determinants in more detail below and provide
testable hypotheses for all of these decisive factors. However, I am especially interested in
how well external factors can explain participation in voluntary climate change initiatives and
whether there is evidence for interdependent decision-making.
3
Elsewhere I have looked at the influence of vulnerability and natural risk on the participation in voluntary
initiatives (Schaffer 2010) and have found that especially coastal counties and to a lesser extend also fatalities
from natural hazards related to extreme weather conditions tend to increase participation. However, socioeconomic and political factors were found to be better predictors of MCPA participation levels. Therefore, I
concentrate on these influences to explain the adoption and timing of cities’ climate change policies.
100
Internal determinants of policy adoption
Turning to the internal determinants of policy adoption, I start with a discusion of the demand
for local climate change policies. I expect that policy-makers can evaluate whether there is
support for climate change policies or whether their constituencies oppose such policies (Oates
& Schwab 1988).4 I then consider factors concerning the supply side, i.e. characteristics of the
local government that are assumed to either facilitate or hinder MCPA participation. As in any
market-related model, these two are, of course, interrelated.
Demand side explanations Endowment with resources or wealth is a factor that matters in
policy-making (Gray 1973). One line of argument points to a greater capacity to compensate
losers from potential regulation. A second line of argument however, stresses the relationship
between levels of income and the demand for environmental quality. Due to its high income
elasticity, economists have conceptualized environmental quality as a luxury good (Baumol &
Oates 1988). As far as environmental policies are concerned, research has shown that people with
higher incomes increasingly care about quality of life issues (Kahn 2006). As mentioned above,
climate change policies complement other local policies, such as public health or transportation
(e.g., the lowering of congestion or the provision of more public transport), thus enhancing the
quality of life (Betsill 2001). With increasing wealth, I therefore expect additional demand for
environmental quality in general (Beron et al. 2003; Cornes & Sandler 1996) and climate change
policies in particular. Consequently, demand for climate change policies should be higher in
localities with a higher median income.
Correspondingly, bad economic conditions in a city are expected to lower demand for more
climate friendly policies. Although Engel & Orbach (2008, 132) argue that ’labour unions often
support renewable energy projects because, in contrast to imported oil, they generate jobs’, cities
with high economic distress are assumed to prioritize economic development over becoming
active on climate change.
As argued already in chapter 3, education is a major factor explaining individual level environmental attitudes. Theodori & Luloff (2002) find that better educated individuals are more
likely to contribute money or time to an environmental group, to read environmental magazines,
or to vote for or against a political candidate because of his or her position on the environment.
Generally, better educated people tend to be more informed about the consequences of human
4
This is similar to Oates & Schwab (1988)’s model where regulators set environmental standards at the level
which maximizes the utility of the median voter in the constituency.
Testing Participation in Space and Time
101
activity on the earth’s climate (Betsill 2001). This effect may be due to correlates, such as a
higher income or a more liberal ideology. However, I posit that the effect of education is separate
from income and can be conceptualized as a sort of ’learning to embrace climate change’ effect,
which is similar to the one identified by Hainmueller & Hiscox (2006). Following this logic,
higher levels of education bring about exposure to information about climate change and lead
to an ’awareness of consequence’ (O’Connor et al. 2002). Therefore, I expect that demand for
local climate change policies, and therefore the likelihood of MCPA adoption increases with the
percentage of college-educated people in a city.
Furthermore, the ideological dispositions of the population are typically taken into account
for any policy decision. Generally, the environmental movement has often been associated with
policy liberalism and therefore, ’liberals are often seen as more likely to support government
solutions to environmental problems than conservatives, who are more likely to rely on private
sector or market-based solutions’ according to Daley & Garand (2005, 619). During the past
decade, the Democrats and their political leaders in the U.S. have been more supportive of
efforts to curb greenhouse gases than the Republicans (O’Connor et al. 2002). I thus expect
a higher demand for climate change policies in cities with a higher proportion of Democratic
voters.
In sum, I assume that mayors who want to be re-elected behave like any other office-seeking
politician. Hence, if voters within the local jurisdiction demand action on climate change, local
governments will be more likely to adopt more stringent environmental politics (Oates & Schwab
1988). I further expect that local decision-makers know the preference of their constituents
(Oates 1998; Urpelainen 2009). More public demand for climate change policies should therefore
translate into potential political benefits for the policy makers that are associated with policy
adoption, which in turn should make adoption more likely.
Hypothesis 1a: The propensity of a city to adopt voluntary climate change policies
increases with higher demand for climate change policies
Supply side explanations Supply side explanations that matter for the adoption of local climate
change policies concern the local government and more specifically, the mayor or the city council
as policy-makers. As mentioned in the section on demand side explanations, local policy-makers
are assumed to react to the demand for climate change policies in their constituencies and draft
policies accordingly. To be able to respond to the demands of the population, administrative
102
capacity is an important factor. There is also the possibility that the local government acts
according to its own agenda i.e., getting involved in local climate change policies as a policy
entrepreneur even in the absence of a manifest demand by the population (Engel & Orbach 2008;
Engel 2009). Furthermore, political benefits from voluntary climate change policy adoption
are not uniformly distributed. Depending on the form of government, political benefits from
participation may vary.
In the following paragraphs, I discuss how the supply of local climate change policies is
influenced by the form of government, administrative capacity, and the local government’s own
environmental agenda. Each of these factors either acts as a facilitator to voluntary action or
as an institutional barrier that make MCPA adoption less likely.
Form of local government: One feature of the executive branch of city governments that
is frequently analyzed and debated is the difference between a ’reform’ council-manager or a
mayor-council system (Davis & Hayes 1993; Feiock et al. 2003; Lubell et al. 2009). The introduction of the council-manager abolished the position of the mayor as the chief executive.
Instead, the elected council hires a professional city manager for daily administrative decisions
in the council-manager system (Lubell et al. 2009). Mayors, in this arrangement, have largely
representative functions and less political and administrative power. Therefore, mayors in towns
with the council-manager arrangements are referred to as ’weak mayors’. In contrast, ’strong
mayors’, who are directly elected, have more political and administrative power, but face elections. Therefore, they also have to be more responsive to their electorate. Generally, Kwon
et al. (2009) argue that the mayor-council form creates a political incentive for officials to adopt
visible policies, which they can claim credit for. In the case of voluntary climate change policies,
and especially in the context of the Mayors Climate Protection agreement, I assume that due to
gains in visibility and popularity, strong mayors are more likely to participate in the agreement.
Hypothesis 1b: The propensity to sign the MCPA is higher for a city with a mayorcouncil as compared to one with a council-manager form of government.
Administrative capacity: City-level research has shown that larger cities with great administrative capacity are among the first to develop an interest in climate change solutions (Kern &
Gotelind 2009). With respect to the MCPA, the original proposal that initiated the agreement
was drafted by Seattle’s Office of Sustainability and Environment (Warden 2007). While larger
cities have specific personnel that are dedicated to traditional environmental protection issues,
Testing Participation in Space and Time
103
such as sustainability and air pollution, smaller cities may not have such specialized administrative capacity. I expect that cities with higher administrative capacity are more likely to be
informed about climate change and its possible city-specific impacts. Therefore, the likelihood
of joining the MCPA should increase with the level of administrative capacity in the city.
Hypothesis 1c: The propensity of cities with high administrative capacity increases
the likelihood of voluntary climate change adoption.
Policy entrepreneur : According to Mintrom (1997, 739), policy entrepreneurs are political
actors that promote specific policy ideas, identify problems, network in policy circles, and build
coalitions to implement such policies. The importance of policy entrepreneurs for the promotion
and implementation of voluntary climate change initiatives is one of the most consistent findings
in the relatively new literature on these initiatives (Betsill 2001; Krause 2011a; Rabe 2004, 2008;
Vasi 2006). As the word ’policy’ entrepreneur already suggests, it is assumed that the mayor as
the decision-maker receives political benefits from the pursuit of his own agenda. For example,
if the mayor has previously been involved in the promotion of environmental policies at the city
level, I expect him to also act as a policy entrepreneur in the realm of voluntary climate change
policies. Cities that have been active promoters of local environmental policies are considered
to have a higher propensity to engage in voluntary climate change policies.
Hypothesis 1d: The existence of a policy entrepreneur in the city increases the
likelihood of signing the MCPA.
External determinants of voluntary action
The theoretical argument above posits that city-specific factors alone do not suffice to explain
adoption decisions in voluntary climate change policies. Research on voluntary climate change
initiatives has so far either neglected the influence of external factors altogether or has tested
them only as controls (Krause 2011a).5 In this paper, I therefore want to explicitly account
for unit-external factors that may lead to interdependent policy choices. In essence, I argue
that cities, when deciding whether to sign up to the MCPA, consider what other cities have
done on the issue. Such interdependencies between cities would therefore manifest themselves
in reactions of local governments in view of their peers’ actions (Brueckner 2003).
5
A notable exception is Vasi (2006). In his event history analysis on the adoption of the CCP program, he
controls for the effects of geographical neighbors’ adoption of the CCP.
104
Interdependent choice is the basis of the field of policy diffusion studies, where diffusion can be
dened as ’a process where choices are interdependent, that is, where the choice of a government
inßuences the choices made by others and, conversely, the choice of a government is inßuenced
by the choices made by others’ (Braun & Gilardi 2006, 299). The question of how policies
diffuse between American states has a long tradition (Walker 1969; Gray 1973; Berry & Berry
1990; Mooney & Lee 1995) and has more recently received renewed scholarly interest (Berry &
Baybeck 2005; Baybeck & Huckfeldt 2002; Boehmke & Witmer 2004; Daley & Garand 2005;
Shipan & Volden 2006, 2008; Volden 2006; Nicholson-Crotty 2009). In a large variety of issue
areas, most studies have found that diffusion is more likely to happen between neighboring rather
than distant states. The positive effect of a neighbor’s adoption on one’s own policy has been
found, e.g., for anti-smoking legislation (Shipan & Volden 2006, 2008), Indian gaming (Boehmke
& Witmer 2004), lottery adoption (Berry & Berry 1990; Berry & Baybeck 2005), and welfare
programs (Volden 2007).
The discussion on how policies diffuse has evolved over the years, starting with the assumption
of a simple imitation process at work. For example, Gray (1973) states that the ’gain in adoptions
is due to nonadopters’ emulation of adopters’ (Gray 1973, 1176). Berry & Baybeck (2005, 505)
finds that decision-makers learn from others to the extent that ’when confronted with a problem,
decision makers simplify the task of finding a solution by choosing an alternative that has proven
successful elsewhere’. Although different conceptualizations of diffusion mechanisms have been
proposed by different fields, i.e. sociology (Rogers 1995), economics (Young 2009), and political
science (Braun & Gilardi 2006; Simmons et al. 2008), these conceptualizations are very closely
related. Furthermore, while Shipan & Volden (2008) as well as Simmons et al. (2008) establish
four main mechanisms of diffusion; learning, economic competition, imitation, and coercion,
Gilardi (2010, 651) recently stated that ’there is agreement that competition, learning and social
emulation are the main drivers of diffusion [. . . ].’
In general, the specific kind of mechanism that applies may differ according to which units
(e.g., cities, countries, or people) are examined and what kind of policy diffuses. To give an
example, although Shipan & Volden (2008) also use cities as their unit of analysis, I do not
necessarily share their argument concerning the theoretical distinction between what is emulation
and what is learning. In their conception of learning, the opportunity to learn is the decisive
factor, and therefore, a city’s likelihood to adopt increases when the policy in question is adopted
broadly throughout the state. In contrast, emulation only happens if comparatively smaller cities
Testing Participation in Space and Time
105
imitate their larger (in population size) neighbors. Besides the small conceptual difference in
the definition, I would posit that in the case of signing a voluntary climate change agreement,
small cities instead look at other small cities for information on signatory behavior. This makes
sense in the area of climate change policies because contrary to anti-smoking bans, the problems
and solutions differ for different city sizes. This would even hold if one would regard voluntary
climate policies as purely symbolic. It is intuitive to assume that policy-makers in smaller cities
experience different political benefits than those in larger ones. Whether voluntary climate
change policy adoption is caused by learning or emulation is very difficult to identify, since both
processes are a consequence of new information on how other cities position themselves on the
issue. Furthermore, given that the object that is supposed to be diffused from city to city is in its
substantive content a voluntary policy, coercion does not seem to apply. Economic competition
in the direct sense is also not entirely plausible given that cities have discretion over the scope
of the climate policies they implement. However, following up on Warden (2007)’s work, there
may be competition between cities to be the green champion within a region.
Concluding the discussion on the differing mechanisms of diffusion, I assume that learning,
emulation and – to a certain extent – competition are at work. However, since I am mainly
interested in the comparison of unit-internal and unit-external factors for MCPA adoption on
the one hand, and testing different conceptions of contingency on the other, disentangling these
mechanisms is not the goal of this paper.
In the following part of this theoretical section, I claim that there are three main channels
through which interdependence in the adoption of voluntary climate change policies operates;
closeness to neighbors, similarity of peers, and participation in national networks.
Neighbors
The positive influence of geographical neighbors on policy adoption is one of the
most consistent elements in state-level (e.g. Walker 1969; Berry & Berry 1990) and city-level
diffusion studies (Krause 2011a; Vasi 2006; Shipan & Volden 2008). In a similar vein, I argue
that local governments are more likely to adopt policies if neighboring cities already have them
in place. Serving as their geographic peer group as well as general competitors for resources,
cities compare themselves to cities that are geographically close. Neighbors can influence local
policy adoption through internal demand, i.e. if the constituency observes climate change policy
adoption in surrounding communities, they could also try to lobby for similar action within their
106
community. The rationale lies in the fact that more social interactions happen in geographically
close settings.
Hypothesis 2a: Participation in nearby cities increases the likelihood that a city
will participate.
Peers
Although it is probably easier for decision-makers to observe whether the cities that
are located geographically closer to their own locality adopt climate change policies, these surrounding cities might not be the relevant peer group to refer to. Interdependence is theorized
to occur due to perceived similarities between subnational units. I assume that cities know who
their peers are as they interact in multiple policy dimensions. In general, cities either learn from
the experiences of their peers or compete with them for resources, e.g., federal grant money.
Therefore, when considering whether or not to sign up to a voluntary climate change initiative,
they refer to their peers’ previous actions on the issue. Rincke (2005) argues that it may be
rational for office-seeking governments to choose policies that are similar to those implemented
in benchmark jurisdictions. In the context of voluntary climate change initiatives, I argue that
such benchmark jurisdictions or relevant peer groups consist of cities with a similar population
size and those in the same income bracket. This assumption is fairly straightforward and has
something to do with the problems that a city government faces as well as its capacities to
solve these problems. In the realm of climate change policies, larger cities might be more concerned about urban sprawl, while smaller communities might concentrate on renewable energy
solutions. A richer community might decide to invest more in alternative energy production,
whereas less well-endowed communities might want to save energy cost. When it comes to the
adoption of voluntary climate change initiatives, I therefore expect that cities take into account
what similar cities have already done.
Hypothesis 2b: The likelihood that a city signs the MCPA increases with the share
of cities of the same population group that have joined the MCPA.
Hypothesis 2c: The likelihood that a city signs the MCPA increases with the share
of cities of the same income group that have joined the MCPA.
Social networks
Besides these similarities that are assumed to help diffuse voluntary climate
change policy adoption, participation in networks is also thought to increase the likelihood of
Testing Participation in Space and Time
107
joining climate change agreements. In theory, social networks connect participants that otherwise may not interact. Therefore, networks emphasize the relationship between participants
rather than the participant’s own inherent characteristics (Leitner & Sheppard 2002). However,
as Gore (2010, 35) points out, ’the loose, flexible character of some networks makes it difficult to
understand how networks and network participants operate and how they influence the actions of
other network members’. An essential means for conceptualizing social networks is the presence
of an opportunity or platform to exchange knowledge, e.g. about solutions to local problems,
and to encourage learning from other participants. Meetings of the U.S. Conference of Mayors
(USCOM) provide such opportunities for the cities that attend.6 I expect that local governments
that are part of such national networks are generally interested in new ideas and the exchange
of information about environmental policies with representatives of similarly interested communities. As actors in environmental policy networks, local governments learn about best practices
and communicate successes or failures of climate change policies. Also, several studies have
found that advocacy organizations, as integral component of these networks, facilitate policy
adoption (Mintrom 1997; Shipan & Volden 2006). Therefore, participation in national networks
can be expected to spread knowledge about a new policy to a city, and participation in networks
increases the likelihood of the adoption of innovative voluntary climate change initiatives.
Hypothesis 2d: A cities’ participation in networks of cities increases the likelihood
of MCPA adoption.
Events
In a temporal setting, there are also time-specific events to be considered that have an impact
on the whole system or parts of it. These events are assumed to influence decision-making
concerning voluntary climate change initiatives. Generally, one can differentiate between unitinternal events and external shocks. Unit-internal events, such as a mayoral election, only
concern the city in which the election is being held, whereas external shocks apply to all units.
Mayoral elections In section 4.3.2 mayors were considered to be office-seeking politicians that
care about re-election. In order to get re-elected, it was further assumed that mayors implement
policies in accordance with the city-specific demand for these policies, supply-side considerations,
6
In social network theory, the whole concept of an affiliation network is based on the idea that actors are linked
through event participation (Wasserman & Faust 1994).
108
and external factors influencing the decision to implement. A mayoral election or more precisely
the campaign before such an event represents the time when the mayor is especially in need
of accumulating political capital by reacting to the needs of the citizens and making public
gestures. Therefore, I expect that in cities where mayors have not yet signed the MCPA, the
closer a mayoral election is, the more likely it is that a city will join the MCPA.
Hypothesis 3a: The propensity of cities to participate in the MCPA is higher in
the time period before a mayoral election.
Energy efficiency and conservation block grants Section 4.3.1 dealt with the fundamental
puzzle underlying the observed high participation in voluntary climate change initiatives, and
the theoretically expected collective action problems underlying the provision of a global public
good. Large voluntary groups, such as the MCPA, should theoretically only be able to overcome
free-riding behavior if they can offer certain excludable benefits through participation (Mueller
2003). Olsen (1971, 51) stipulated that: ’only a separate and selective incentive will stimulate a
rational individual in a latent group to act in group-oriented ways’. In the case of the CCP, Brody
et al. (2008, 34) establish such separate and selective incentives as software and analytic services,
as well as strategic plans that enable cities to inventory and track their GHG emissions. It has
been hard to argue that such selective benefits emanate from MCPA participation however,
as this agreement is even less exclusive than the CCP, which at least charges membership
fees. However, I argue that two related events slightly altered the incentive structure of MCPA
participation and thereby changed the existing ’rules of the game’. The two events relate to
legislation concerning the so-called energy efficiency and conservation block grants (EECBG).
This legislation entered Congress for debate in July 2007 and was signed into law on 19th of
December 2007 as part of the Energy Independence and Security Act (EISA). However, it was
funded for the first time by the American Recovery and Reinvestment Act (Recovery Act) of
2009, where the EECBG Program was appropriated $3.2 billion.7 This program is intended to
assist U.S. cities and counties ’to develop, promote, implement, and manage energy efficiency
and conservation projects and programs designed to a) reduce fossil fuel emissions; b) Reduce
the total energy use of the eligible entities; c) Improve energy efficiency in the transportation,
building, and other appropriate sectors; and d) Create and retain jobs’ (U.S. Department of
Energy 2010). As the national mayors advocacy group, USCOM had long been lobbying for this
7
Further information can be obtained at http://www1.eere.energy.gov/wip/eecbg.html
Testing Participation in Space and Time
109
funding and proclaimed the legislation in 2007 to be huge success (Mayors Climate Protection
Center 2007). To obtain funding from the competitive grants within the EECBG, cities have
to apply with an energy efficiency-related project proposal.8 Due to the success of the MCPA
within the U.S. Conference of Mayors and nationwide, I assume that participation in the MCPA
could act as a signal to the government that a city has expertise in the realm of climate change
policies in general, and energy efficiency policies in particular. I therefore expect participation
to increase with the advent of legislation on the EECBG program.
Hypothesis 3b: The likelihood of cities to participate in the MCPA increases with
the initiation of the EECBG program.
Obama administration As elaborated above in section 4.3.1, I assume that one of the motivations for local politicians to join the MCPA was to deliver a statement against the inaction
on the climate change issue by the federal government of President George W. Bush. Before
the general election of 2008, such a motion is supposed to have created political benefits in
constituencies that lean towards the Democratic Party or such that are discontent with the federal government. With the election of Barack Obama, the candidate of the Democratic Party, I
hypothesize that this motivational logic came to a halt, in the sense that the new administration
was first granted a period of adjustment to tackle the issue of national climate change policies.
This period, however, would have ended by the summer of 2010 with Barack Obama and the
Democrats failing to bring meaningful climate change legislation through Congress. However,
since this coincides with the time period I am looking at, I expect therefore that the election of
Barack Obama lead to a decrease MCPA signature due to the expected national approach on
the topic.
Hypothesis 3c: The likelihood of cities to participate in the MCPA decreases with
the election of Barack Obama as the President of the U.S..
This concludes the theoretical part of this paper. I have argued that voluntary policies on
the local level are not only influenced by local factors such as the resource endowments of
a community, but are also determined by a city’s interdependence with others. The following
paragraphs gives an introduction into the research design used to empirically test the hypotheses.
8
For example, a project involving the ’developing, implementing, and installing on or in any government
building of onsite renewable energy technology that generates electricity from renewable resources (solar and
wind energy, fuel cells, and biomass)’ (U.S. Department of Energy 2010)
110
4.4 Research Design
4.4.1 Case Selection
Selection of States
While I would have liked to conduct a study on the temporal and spatial variation in voluntary
climate change policy adoption for all cities within the U.S., severe data limitations on city data
availability (c.f. section 4.4.4) as well as concerning cities’ MCPA adoption dates (c.f. section
4.4.2) prompt me to test the temporal and spatial dynamics within a significantly smaller subset
of states. The states that I chose for this analysis of MCPA adoptions are the seven Midwestern
states of Illinois, Indiana, Michigan, Minnesota, Missouri, Ohio, and Wisconsin.9
In order to be able to generalize findings from this study, I wanted to make sure not to exclusively analyze states that are either leaders (e.g., California, Florida, or the Northeastern states
of the U.S.) or laggards (e.g., Texas) on the climate change issue. Since the focus of this study
is on socio-economic and interdependence factors, climate change vulnerability needs to be kept
constant. For instance, this is why I do not include coastal as well as non-coastal states in this
study. Compared to other parts of the U.S., the risk from climate change is comparatively low
in the Midwest (Karl et al. 2009), except from a potential worsening of already existent extreme
weather-related natural hazards. Therefore, I can assume that participation in voluntary climate
change agreements are determined by socio-economic and political characteristics of cities or the
result of interdependence between neighbors or peers.
I now introduce the states in my sample.10 Table 4.1 provides a comparison of the Midwestern
states on several relevant indicators. In subtable (a), the average per capita GDP is listed for
all seven states and the U.S. alongside the ranking of the state across all 50 states. Whereas
Illinois and Minnesota are above average, the remaining five states are below. Considering the
ranking of all 50 states, the range exhibited by my seven states – from 10 (MN) to 40 (MO)
– seems to be quite large. A similar picture presents itself in subtable (b), which lists the
states according to their energy consumption in 2007. Total Energy Consumption is calculated
as the sum of all energy sources (fossil and non-fossil) in each state. Here we can see that
Indiana as well as Minnesota are among the consumption leaders, whereas Michigan and Illinois
consume comparatively less energy. Heavy industry, especially the steel industry, accounts for
9
The only directly contiguous Midwestern state missing is Iowa, which was excluded due to its mainly agricultural
economy.
10
Data are from Climate Analysis Indicators Tool (CAIT US) version 4.0. (2010)
Testing Participation in Space and Time
111
Indiana’s lead on this indicator. Three states have above U.S. average energy consumption,
while four are below. The third indicator in subtable (c), the GHG intensity of the economy,
is also very relevant in the context of climate change mitigation policies. GHG intensity of the
economy (or GHG per GDP) is a measure of greenhouse gas emissions per unit of economic
output and reflects both the state’s overall economic structure as well as the level of energy
efficiency (Climate Analysis Indicators Tool (CAIT US) version 4.0. 2010). It therefore proxies
the abatement costs incurred by each state of either engaging voluntarily in climate change
mitigation or from potential nation-wide regulations in this area. We can see that Indiana,
again due to its heavy industrial production, leads the table, whereas Minnesota has the least
GHG intensive economy. Once again, the U.S. average lies in between the Midwestern states in
my sample. The selected states are not extreme outliers on the economic and abatement cost
indicators. Instead, they conveniently cover the range of values that would have been obtained
from to a more extensive U.S. study.
Table 4.1: Comparison of the 7 Midwestern states and the U.S. average
(a) GDP in 2007
State
MN
IL
US
WI
OH
IN
MI
MO
$ p.P.
40990
40287
37931
35161
33703
33263
32936
32359
(b) Energy Consumption in 2007
Rank
10
13
–
27
33
35
38
40
State
IN
MN
OH
US
MO
WI
IL
MI
Tonnes Oil p.P.
11.5
9.1
8.9
8.5
8.4
8.3
8
7.6
Rank
10
20
23
–
27
28
34
38
(c) GHG intensity of Economy 2007
State
IN
MO
OH
MI
US
WI
IL
MN
tCO2e/ Mill $
1189.3
756.2
723.4
573.9
538.5
537.5
483.8
475.7
Rank
8
18
20
24
–
26
32
33
Table 4.2: Climate change related policies on the state-level (Source: Pew Center on Global
Climate Change)
Action Emission Portfolio
E.E. Res. Green Build. Overall
Plan
Targets
Standards Standards Standards
(max. 21)
IL
IN
MI
MN
MO
OH
WI
x
x
renewable
x
x
x
x
x
x
alternative
renewable
renewable
alternative
renewable
x
x
x
x
x
x
x
x
14
9
11
15
7
11
15
112
With respect to present voluntary climate change programs, table 4.2 shows selected climate
action efforts at the state level. Generally, we can see that there exists a lot of variation in
the implementation of different Green Buildings and Energy Efficiency instruments and policies
across the seven states. The instrument that most of the states have introduced is a climate
change action plan. Only Indiana and Ohio have not yet drafted such a plan. Illinois, Minnesota,
and Michigan even have state-wide emission targets for GHGs in place. Renewable or alternative
Portfolio Standards are implemented in all states except Indiana, whereas the related concept of
Energy Efficiency Resource Standards, which encourages the more efficient generation and use
of electricity and natural gas, is used in only four states so far. Green building standards for
state buildings are in force in Indiana, Michigan, Ohio and Wisconsin. Overall, the table on the
seven states’ voluntary engagement in climate change policies clearly shows that there is a lot
of variation in the voluntary approaches taken by these states. In addition, it emphasizes once
more the adequacy of my sample choice. The chosen contiguous subregion of states is, to some
extent, representative of the larger U.S. with respect to economic indicators, abatement costs,
as well as engagement in voluntary climate change policy.
Selection of Cities
With respect to the cities within these seven states, I have decided to use a threshold of 10,000
inhabitants for my main analyses. Although this decision increases the number of observations
for which I need data considerably, it is crucial for a proper study of the Mayors Climate Protection Agreement. Especially in recent years, small to mid-size cities (10,000-30,000 inhabitants)
have increasingly signed on to the MCPA. Explicitly accounting for this development increases
the generalizability of my findings. However, to corroborate my findings from this large sample, I also generate a second sample that contains only those cities that have 30,000 or more
inhabitants.
Applying these two cut-offs to data from the National Atlas of the United States (2004), which
is based on population figures from the 2000 Census11 , yielded 749 cities over 10,000 inhabitants
and 242 cities over 30,000 inhabitants in the seven states.12 Subfigures 4.2(a) and 4.2(b) show
11
Although population estimates on the city level are calculated for each year (large cities) or each five years
(small cities), I have decided to use the 2000 Census data instead of the less precise estimates for the 30,000
and 10,000 citizens thresholds discussed above.
12
These are only cities or towns and no census-designated places (CDP). CDP’s are concentrations of population
that the U.S. Census Bureau uses for statistical purposes. They resemble smaller incorporated places in size,
but lack a separate municipal government, and therefore these places were not considered in this study on the
adoption of voluntary climate change policies by local governments.
Testing Participation in Space and Time
_
[
113
_
[
_
[
_
[
North Dakota
_
[
_
[
Minnesota
_
[
_[
[
_
_[
_ [
_
_[
_[
[
_[
[
_
[
_
_[
[
__
[
_
[
South Dakota
_
[
_
[
_
[
Wisconsin
_
[
_
[
_
[
_
_[
[
_
[
_
[
_
[
_
_
[ [
__
[
_[
[
_
_ _[
_[
[
_[
[
_
[
_
[
_
[
_
[
_
_
[
[
_ [
[
_
_
[
_
[
_[
_
[
[
_[
_
[
_[
_
_
_
[
[
_
[
_
[
_
[
Iowa
_[
[
_
Nebraska
Michigan
_
[
_
[
_[
[
_
_
[
__
_[
_
_[
_[
[
[
_[
[
_
[
_
[
_
[
_
[
_
[
_
_[
[
_
[
_
_
[
_
[
_ _
[
[[
_
_[
[
[
_
___
[
_
[
_[
[
_ [[
Pennsylvania
_
[
_
[
_
[
_
[
_
[
_
_[
[
_
[
_
[
Missouri
Ohio
_
[
_
[
_[
[
_
_
[
_
[
_
[
_
[
Kansas
_
[
_
Indiana [
_
[
Illinois
_
[
_
[
_
_[
_[
_[
[
[
_
_
[
_[
[
_
_
[
West Virginia
_
[
_
[
_
[
Kentucky
Virginia
Oklahoma
Arkansas
Tennessee
Text
(a) 749 cities >=10.000 inhabitants
_
[
North Dakota
_
[
Minnesota
_
_[
[
[
_
_
[
_
_[
[
Wisconsin
South Dakota
_
[
Michigan
[
_
_
_[
[
_
[
Nebraska
_
[
_
[
_
[
_
[
Illinois
_
[
[[
_
_
Indiana
_
[
Missouri
_
[
_
[
_
[[
_
_
[
Pennsylvania
Ohio
_
[
_
[
Kansas
_
[
_
[[
_
_[
_
[
[[
_
_[
[
_
[
_
_
[
_
[
_
_
[
_
[
_[
[
_
_
[
Iowa
_
[
_
[
_
[
[
_
_[
[
_
_[
[
_
_
[
West Virginia
_
[
Kentucky
Oklahoma
Virginia
Arkansas
Tennessee
Text North Carolina
(b) 242 cities >=30.000 inhabitants
Figure 4.2: Sample of cities and signatories in the seven Midwestern states
all the states and cities in my two samples. Red stars indicate cities that have signed the MCPA,
all other cities are displayed as blue dots.
This section has introduced two samples of cities. In the following section, I describe how the
data that are used to test the observable implications hypothesized in the theoretical part were
collected.
114
4.4.2 Data Collection: Date of MCPA Adoption
The U.S. Conference of Mayors Climate Protection Center, which administers the MCPA, only
provide information about which cities have joined the agreement on their homepage (www.
usmayors.org/climateprotection/), but not when they joined. Therefore, one of the main
challenges regarding empirical tests of the adoption of voluntary climate change initiatives was
to collect data on the mayor’s signature date. Contacting various officials at the Mayors Climate
Protection Center multiple times in 2008 and 2009 with regard to these dates was not successful.
Most e-mail queries or messages left on answering machines were either not returned or with
substantial delay. The first answer that I received, 6 months after my original request and after
two reminders, was as simple as: ’About your request, we do not make the signatory dates public’.
Luckily, I finally obtained the information I needed thanks to the cooperation of various sources
– to whom I am all extremely grateful. The following paragraphs are dedicated to a discussion
of the complicated and time consuming, but ultimately successful data collection process.
First, in March 2009, professor Dana Fisher, whom I was visiting in New York helped me
establish a contact within USCOM, who had also worked for the climate protection center earlier.
I was then able to interview him and to discuss with him the possibility of obtaining records
of the dates of each city’s adoption of the MCPA. He pointed out that the organization had
not even collected the exact days during the first year, when the MCPA was still administered
in Seattle. While they started collecting the dates from 2007 onwards, they were not going to
release these dates for fear of exposing some of their members to public criticism over their lack
of pro-active climate policy action. What he could provide me with, however, was a list with
the first 141 cities that had signed the agreement during the initial 5 months in 2005. This
list of 141 cities was presented at the 2005 Annual Meeting of the USCOM in Chicago, when
the MCPA was officially endorsed by the conference. This provided me at least with a starting
point, as I now knew which cities were the first leader cities that signed until June 2005.
Second, Toby Warden, who had written her dissertation at the University of California, Irvine
about the early years of the MCPA (Warden 2007), kindly shared information on the earlier
years with me. Although she did not have enough information about the starting period at the
time she wrote the dissertation, she had downloaded some valuable information on who had
signed in 2006 and 2007 from the webpage of the Mayors Climate Protection Center.
Third, the people I interviewed at the Sierra Club – one of the oldest environmental groups in
the United States – fortunately were very cooperative in sharing the data they had collected on
Testing Participation in Space and Time
115
the signature dates of various cities. They have a subgroup called the Cool Cities Program, led
by volunteers around the U.S. It is a collaboration between community members, organizations,
businesses, and local leaders to implement clean energy solutions. One of the steps in becoming
a cool city is that the local campaign sends formal letters to the mayor asking him to sign the
MCPA and to record when they had done so. Therefore, they also had collected dates when
cities joined. The national Cool Cities campaign director, Stephanie Cutts, whom I interviewed
in March 2009, was kind enough to share the data they had collected until August 2008.13
Furthermore, the Cool Cities Program now also features a homepage with all cities and – where
the Sierra Club knows it – the date of adoption of the MCPA.
Last but not least, over the past years, I have downloaded and saved the data provided on
the MCPA homepage concerning the signatories to the agreement for each month from August
2008. So if a new member was added to the MCPA homepage, I could observe it and would
know the approximate month of signature. Although there is a slight lag between the signature
and the upload on the homepage, these nevertheless provided valuable information (as I point
out below) when none of the other sources said anything about the participation date.
Since the above sources did not provide all the data points necessary for the analysis of
MCPA adoption in the seven Midwestern states, I decided to contact each of the 155 cities
that were listed to have signed the MCPA by March 2010. To that end, I wrote an e-mail
explicitly asking for the month and the year the city had signed the agreement. Furthermore, I
drafted a questionnaire that included additional questions about city characteristics, reasons for
participation in the MCPA, as well as some question concerning their assessment of the overall
initiative. The sample questionnaire as well as the e-mail are provided in the Appendix. For this
data collection, a new e-mail address (climateresearch@ethz.ch) and a U.S. phone number
were created to give city officials the possibility to call back or ask questions about the research
project. It was stressed in the e-mail that priority of the research was to obtain the date of
MCPA adoption. Having collected the phone numbers and e-mail addresses for the 155 cities,
the first e-mails were sent out in April 2010.14 After one week, we had feedback from about 15
cities, which send us the relevant information, and about 6 or 7 questionnaires. After two weeks,
a reminder e-mail was sent to the cities that had not replied, which led to an increase in replies.
By the end of April, we then started calling cities and asking whether they had received the
13
Although Ms. Cutts wanted to sent me an updated version in 2009, there was a change of directors in mid-2009
and unfortunately I could never establish contact to the new director
14
Research assistance by Gwen Tiernan in contacting the cities is greatly acknowledged.
116
information concerning the research project and whether they would be interested in providing
this information. Whereas most of the e-mail addresses directly went to the mayor, who might
have been too busy to answer, the telephone numbers provided by USCOM usually led to the
secretary or city clerk, who then referred us to the right person to contact. We then resend
our query, since many had not received our request for information. Therefore, the personal
contact over the phone led to an increase in information about the date of adoption and by May
2010, we had information from about 60% of the 155 cities. We then sent out a third and last
reminder indicating the end of the research project and asked again for the city’s cooperation
concerning MCPA adoption. Finally, we filtered out the ones where we had the exact date of
adoption from the Sierra Club data and then predominantly called the ones we did not have
this information. In the end we had obtained about 85% of the data through our study or the
Sierra Club data.
To avoid problems caused by potential uncertainty in the recollection of information, I crosschecked the dates indicated by the mayor or city staff with the information I had from the other
sources. In most of the cases those were either identical or deviated by about one month. I
attribute this difference to the fact that there is a slight time lag, e.g., between the signature of
the mayor and the time the information that the city has signed the MCPA is presented on the
organization’s homepage. However, in some rare (< 5) instances the information provided was
more than two months apart. I then went to the cities homepage or looked at local newspaper
posting to see whether there was an announcement about the MCPA and could resolve all
inconsistencies in this way.
From the remaining circa 15% that never replied to us or were not listed in the Sierra Club
data, the month of adoption could be directly taken from the data downloaded from the Mayors
climate protection center homepage each month, for the cities that had signed in 2008 and 2009
(11 cities). A potential problem as mentioned earlier is the slight difference with the information
of when the Mayor signed the agreement (as reported by the city) and when it was posted on the
homepage. I do not regard this as a huge problem for the reliability of the results, since it only
applies to a small portion of the data. However, for 15 out of 155 cities, the date of adoption
could not be made available or did not lie within two months. Since I had some information
regarding their participation with gaps from three to six months, I decided to impute the missing
data. A variable impute was created, which indicates whether the information on the record was
imputed or not. An example would be that two of the early adopters (the first 141 cities) did not
Testing Participation in Space and Time
117
know exactly when they had reacted to Mayor Nickels call to sign the MCPA. Hence, I entered
the date May 2005 as a conservative estimate of their signature, since they had definitely signed
by the time the USCOM conference took place in June 2005 and the agreement was endorsed by
the conference. Correspondingly, if the city did not have any recollection, but the other sources
informed me that the local government had not yet signed in February, for example, but was
listed to be a participant in May, the month before, April, was entered as the adoption date. The
longest period for which I do not have data is the one between August 2007 and February 2008.
For the three cities that fall into this 6 months category, I again used a conservative estimate.
January 2008 was put down as their date of adoption. Coinciding with the winter meeting of
the USCOM, it is likely that these cities actually signed up at the time of the meeting. However,
it is also probable that adding these three more cities biases the network participation variable
for this event. I return to these issues in the analysis section.
Notwithstanding these potential drawbacks, the data gathered through the direct contact
with the cities and backed up by other sources allow for a first comprehensive test of the factors
influencing cities to engage in voluntary climate change initiatives. The following section now
takes a first look at the data generated by this effort.
4.4.3 Dependent Variable
The dependent variable of interest indicates whether a city has signed the Mayors Climate
Protection Agreement in a given month or not. As mentioned above, basic information about
which cities had signed this agreement within the seven Midwestern states was collected using
www.usmayors.org/climateprotection/. Table 4.3 shows the distribution of MCPA participant and non-participant cities as of July 2010. As one can see from table 4.3, over all seven
states, 165 out of 749 cities or 22.0% have signed the Mayors Climate Protection Agreement.
The highest percentage of signatories can be found in Minnesota and the lowest percentage
in Ohio. Turning to the temporal data concerning the signature behavior of the 749 cities15 ,
figure 4.3 shows the cumulative number of signatories throughout the seven states over the time
period from February 2005 until July 2010.
We can see that the cumulative MCPA adoptions approximate the familiar s-shaped curve
from studies on the diffusion of innovations (Walker 1969; Berry & Berry 1990; Rogers 1995).
Although steadily increasing in all of the years, one can observe a steeper increase starting by
15
Since the larger 749 cities sample constitutes the main sample, the following paragraphs deal especially with
this more inclusive sample.
118
Table 4.3: Signatories (cities > 10k) per state by July 2010
State
Illinois
Indiana
Michigan
Minnesota
Missouri
Ohio
Wisconsin
Pooled
No. of cities
202
69
92
83
64
166
73
749
Signed
45
13
25
28
14
25
15
165
Not signed
157
56
67
55
50
141
58
584
% signed
22.3
18.8
27.2
33.7
21.9
15.1
21.5
22.0
Figure 4.3: Adoption of the Mayors climate protection agreement (2005-2010)
June 2006 and increasing especially during the year 2007 until about May 2008. These two
years can be thought of as the height of the climate change discourse in the U.S. In 2006,
the documentary ’An Inconvenient Truth’ by Al Gore opened at U.S. box offices on May 24,
shattering existing visitor box office records for documentaries. Although seemingly belonging
in the realm of popular culture, the influence of this movie on people and decision-makers’
perception of climate change and – relevant in this context – the accompanying influence it
had on local climate change policies should not be underestimated. Kellstedt et al. (2008) and
Boykoff (2007) in the U.S. and the U.K. and, more recently, Sampei & Aoyagi-Usui (2009) in
Japan have studied mass media coverage of climate change discourses and its influence on public
awareness of the issue, and they find that Al Gore’s movie had a sizeable impact on people’s
perception as well as on the media discourse. Furthermore, during these years, popular media
Testing Participation in Space and Time
119
outlets, such as Time magazine, featured climate change issues prominently (Kluger 2006; Walsh
2008). Although the release of the Stern Report in September 2006 also had a sizeable impact
on the climate change discussion in the U.K., it was not as widely discussed in the U.S. (Boykoff
2007). However, when the Intergovernmental Panel on Climate Change (IPCC) began to release
a series of Fourth Assessment Reports (AR4) from each of their Working Groups (IPCC 2007a)
in February 2007, it received huge media coverage in the U.S. These media and scientific events
were complemented by news on extreme weather events in the U.S., such as the 2006 drought in
the Southwest and the announcement of the National Oceanic and Atmospheric Administration
(NOAA) in January 2007 that 2006 had been the warmest year on record.16 There are certainly
other very important events that shaped public opinion, but I consider these to be the main
events for the U.S., and a visual inspection of figure 4.3 seems to support the idea that heightened
public awareness for climate change during these years considerably spurred participation in the
MCPA. However, as argued above, whether this increase was due to a genuine concern on the
side of local governments or whether the predominant motive was to gain political benefits by
making a local statement against the federal government’s lack of climate policy action is very
hard to say.
I now divide the adoption propensity over time into further important subgroups. Figure
4.4 shows adoption across the seven Midwestern states in the sample. The y-axis shows the
percentage of cities that have signed the MCPA over all cities in that state. We can see that
Minnesota has the highest percentage of cities signed from early 2007 onwards and reaches a
level of just over 32% by the end of the observation period. Illinois’ trajectory starts out with a
comparatively low number of cities that have signed the agreement, but then steadily increases
to overtake Ohio in 2007, Wisconsin in 2009, and finally ends up at the similar level as Missouri
at around 21%. Whereas Ohio’s percentage of signed cities is just a bit lower until 2007, from
2008 onwards, there is almost no increase, whereas in other states such as Michigan the % of
signed states continues to increase.
In section 4.3.2, I have claimed that there are two conceptions of peer groups, whose actions
on the climate change issue cities closely observe – those of similar income groups and those
that are of similar size. Subfigure 4.5(a) depicts the evolution of MCPA signatories for each of
the four income quartiles. We can see that the percentage of cities that signed of all cities in the
respective income group generally shows a similar pattern over time. Whereas at the beginning,
16
available at http://www.noaanews.noaa.gov/stories2007/s2772.htm
120
Figure 4.4: % of cities who signed the MCPA by state (2005-2010)
a relatively higher proportion of cities of the middle-high category signed the MCPA, from 2007
onwards cities from the low-middle income group have the highest %. Overall, one can see that
all cities seem to co-evolve in similar patterns. However, a very different pattern of development
emerges in subfigure 4.5(b). Here, states are split up into 5 categories, and the evolution of the
percentage of cities that signed over the years 2005-2010 is shown. We can see very clearly that
each of the population groups follows an individual trajectory and that they only intersect at the
very beginning, i.e. in 2005. Also, it can be noted that in the population category of the largest
cities, i.e. cities with over 250.000 inhabitants, 12 out of 12 cities or 100% of the cities signed
the MCPA by February 2009 – the last one being Detroit. The percentage of cities that have
signed increases continuously in all population groups. However, it seems that the proportion
of the smallest cities (between 30k and 50k) that have signed does not grow very fast. Just over
10% of this population group had signed by the end of the observation period. This could be
first evidence that the population groups do indeed evolve differently. Whether this is due to
city-specific factors, e.g., that more specialized knowledge of environmental policies and climate
change makes larger cities more susceptible to new policies, or whether cities’ choices are driven
by the choices of similar sized cities, cannot be determined, yet.
Testing Participation in Space and Time
121
(a) by per capita income group
(b) by population group
Figure 4.5: % of cities who signed the MCPA (2005-2010)
4.4.4 Independent Variables
To test the hypotheses stipulated in the theoretical framework, measurable indicators need to
be found for the specified concepts. I briefly discuss the operationalization of the independent
variables in the following paragraphs.
122
Internal Factors
As proposed in section 4.3, to determine the adoption of voluntary climate change mitigation
initiatives on the part of local governments, internal, city-specific factors have to be taken into
account. These internal determinants of MCPA participation are hypothesized to either emanate
from the demand or supply side. All of the internal variables unless otherwise specified were
obtained using City Data (http://www.city-data.com). This webpage provides the Census
2000 data as well as 2008 estimates from the U.S. Bureau of the Census for all cities over 10.000
inhabitants. Since these data are not available in spreadsheet format, a script had to be written
to access data of the relevant cities.17
Demand side To proxy demand for climate change policies within the city, I rely on socioeconomic and political indicators that are linked to either a higher or lower demand in section
4.3. Socio-economic and political indicators that are assumed to increase demand for MCPA
participation are high levels of education, income, and % of supporters of the Democratic Party
within the community. I use the % of people living in the community that have obtained
a bachelors or a higher degree (% bachelors & up) as an indicator for education. I proxy the
wealth of the community by the median per capita income in the city (p.c. income). To measure
ideology within the community, I rely on the county level result for the Democratic Party in
the 2004 and 2008 elections % Dem Vote 18 . All of the above variables are supposed to increase
demand and, therefore, the likelihood of participation in the MCPA.
Determinants that I assume to lower demand for local climate change policies are primarily
linked to bad economic conditions in a community. I proxy such bad economic conditions via
the unemployment rate in the city (Unemployment) and with the percentage of residents living
in poverty (Poverty level ). High abatement costs within a city were also considered to lower
demand from the constituency. Abatement costs are measured by the variable Industry dep
that takes on the value 1 if the main sector of employment in the city has to do with industrial
manufacturing. In chapter 3, I assumed that a higher median age of the population would have
a negative impact on the participation in climate change policy initiatives. Therefore, I also add
a control variable that measures the median age in the city Median Age and expect a negative
coefficient.
17
18
Research assistance by Krzysztof Wojtaniec to collect data for all 749 cities is greatly acknowledged.
County level data on %Dem Vote in 2004 and 2008 was collected from the U.S. Election Atlas http:
//uselectionatlas.org/
Testing Participation in Space and Time
123
Supply side To test the influence of the form of local government on the likelihood of participation in the MCPA, Mayor-council takes the value 1 for a mayor-council or ’strong mayor’
form and 0 otherwise. Comprehensive data on this variable was not accessible for all cities, but
only for the subsample of cities of 30,000 and more inhabitants. Data were obtained from Vasi
(2006). For this variable, I assume that a city with a ’strong mayor’ form of government will be
more likely to participate in the MCPA.
In hypothesis 1c, I stipulate that a higher administrative capacity is associated with a higher
propensity of MCPA participation. I measure the concept of administrative capacity with two
variables: the percentage of local government employment of total employment in the city % loc
gov empl and the size of the city measured by its population Population. While the first measure
taps at the essence of administrative capacity by measuring its size in relation to other sectors
of employment, the population proxy captures the ’informedness’ about climate change and
sustainability issues aspect of capacity that I discuss in the theoretical part. The larger a city,
the more specialized is the staff and the more likely they will know about new policy areas and
solutions. I therefore expect that both variables are positively linked to MCPA participation.
Hypothesis 1d links the existence of a policy entrepreneur within the city to a higher probability of participation in the MCPA. Since it is hard to obtain data on individual mayors and
their policy preferences, I therefore proxy policy entrepreneurship through a city’s membership
in ICLEI-Local Governments for Sustainability. This association of local governments is committed to advancing climate protection and sustainable development and has worked with cities
on these topics since the 1990s. I assume that if a city is already a member in ICLEI (which
charges membership dues), the existence of a policy entrepreneur in the city and, linked to that,
the likelihood of MCPA participation is higher. Information concerning a city’s membership in
ICLEI was obtained from Vasi (2006) and updated to include members from 2006 onwards with
data from ICLEI-USA.19
External Factors
Neighbors
In the empirical literature on diffusion, geographical proximity has always played a
major role in explaining policy diffusion (Walker 1969; Berry & Berry 1990). Correspondingly, I
suspect that for the units of analysis – the cities in the seven states – contiguity is an important
channel of policy diffusion. I therefore test whether the influence of a truly external factor like
19
available at http://www.icleiusa.org/programs/climate/about-iclei/members/member-list
124
Figure 4.6: Neighbor relations within 100km
geography explains interdependence in the signature behavior of cities with regard to the MCPA.
In order to test hypothesis 3a on policy diffusion via the geographical contiguity of neighboring
cities, a point distance weights matrix with a threshold of 100km was first created via ArcGis.
This matrix was then transformed into a binary matrix indicating with a 1 if the city is a
neighboring city (within 100km) or not. Such a matrix was created for both samples, the one
with the 749 cities over 10,000 inhabitants and one including 242 cities over 30,000 inhabitants.
Most studies of policy diffusion between cities are rather imprecise as to how exactly they define
neighboring cities and mostly use cities in a neighboring county as being neighbors (Krause
2011a; Vasi 2006). However, since counties differ enormously in size, this operationalization is
not the same as having a clearly specified area to include the neighbors in for each observation.
This area or, more precisely, the connections between neighboring cities within 100km20 are
shown in figures 4.6 and 4.7. To obtain a measure for each of the 749 (242) cities that indicates
how many neighbors had signed the MCPA up until the last time period, the binary weights
matrix was multiplied with the outcome of interest in each of the 66 time periods. In the analysis,
20
I thank Stefan Schuetz for help in visualizing the connections.
Testing Participation in Space and Time
125
the count of neighbors for each city that had signed the MCPA up until the time period before
(t − 1) is therefore expected to influence a cities’ propensity to participate in the MCPA.
Figure 4.7: Neighbor relations within 100km, reduced sample >=30k
Peers
In the theoretical part, I have assumed that cities interact with one another in multiple
policy dimensions and that the choices they make concerning participation in MCPA are dependent on their peers’ actions on the issue. Furthermore, it was supposed that cities are more
likely to adopt policies from cities that are ’similar’ to them. Peers are therefore assumed to
be cities that are similar in terms of income and size. The underlying concept of ’similarity of
interest’ is operationalized as follows. To measure the behavior of other cities in the same income
group regarding the MCPA, I have constructed four income groups along income quartiles. The
variable % inc group signed then captures how many cities within the same income group had
signed the MCPA up until the previous time period. Similarly, the variable % pop group signed
informs about the percentage of cities that have signed within the respective population group
(10-30k, 30-50k, 50-100k, 100-250k, and over 250k). It is assumed that the more cities within
the same income or population group have signed up until the previous time period, the more
likely is a city’s signature of the MCPA.
126
Social Networks
To proxy the concept of social networks, I downloaded participation lists from
the U.S. Conference of Mayors webpage (http://usmayors.org/meetings/) for the biannual
meetings between 2005 and 2010.21 The meetings take place in January in Washington, D.C.
(Winter Meeting ), and in June at alternate places (Annual Meeting ). At these gatherings,
mayors not only meet other mayors, but have the chance to talk to, e.g., Administration officials
or Congressional leaders. Generally, a variety of issues that affect cities across the country
are discussed in various policy committees, which come up with new resolutions that are then
presented to the entire members that attend. There are a lot of social networking opportunities
for mayors during the 3-4 days of the meetings. Furthermore, topics relating to voluntary climate
change policies, such as energy efficiency or building codes, are regularly discussed. I therefore
expect that mayors learn about what other cities are doing and what can be done on the issue at
these meetings, and I thus submit that cites that attend bi-annual meeting of the USCOM are
more likely to participate in the MCPA. Correspondingly, the variables Winter Meeting and
Annual Meeting indicate for each city whether the mayor attended the specific meeting.
Events
The data on mayoral elections were obtained from the website of the U.S. Conference of Mayors
(http://usmayors.org/elections/). These data were downloaded at various points in time
and entered into spreadsheets, which were then merged into the data set containing the basic
city-level Census data. Since I expect that mayors are generally more likely to engage themselves
in voluntary climate change initiative right before an election, the variable Mayoral campaign
takes the value 1 for the five months leading up to a mayoral election in the city and is expected
to have a positive and significant influence on the propensity to sign the MCPA.
To proxy the influence that 2007 legislation on the Energy Efficiency and Conservation Block
Grants as well as the 2009 endowment with financial resources for cities had on the incentives
of cities to join, I created two variables EECBG July 07 and EECBG Feb 09. These take the
value 1 for the five months from the announcement that Congress started debate about these
block grants until it became clear that there would be no money appropriated for these block
grants in 2008. Similarly, the variable EECBG Feb 09 taps at the effect that the news from the
appropriation of financial resources within the American Recovery and Investment Act had on
21
Unfortunately, the participation records from the June 2005 meeting were not available
Testing Participation in Space and Time
127
cities’ likelihood to join the MCPA. It takes the value 1 for the 5 months after February 2009
and zero otherwise. I expect a positive and significant coefficient on both variables.
In hypothesis 3c, I stipulated that the impact of Obama’s election in November 2008 lead to
a different logic of participation. To account for this event, I add a variable that is zero in all
time periods before November 2008 and takes on the value 1 afterwards. In accordance with
hypothesis 3c, a negative coefficient is expected.
This concludes the discussion of my independent variables. The following paragraphs introduce the statistical method that is used to empirically test my theoretical claims.
4.4.5 Statistical Method
The unit of analysis used in this paper is the city-month. The time period under analysis are the
66 months from the initiation of the Mayors Climate Protection Agreement in February 2005
to July 2010. For each city-month, the dependent variable signature indicates whether the city
has signed the MCPA in this month or not (0/1). If a city becomes a signatory to the MCPA in
one month, the event is considered to have occurred and, consequently, the observation leaves
the data set.
In analyzing the adoption of a policy, many scholars have used event history analysis (EHA)
(c.f. Berry & Berry 1999). Event history data for the discrete-time process at hand (adoption
of policy or not) would correspondingly denote the dependent variable as a series of binary
outcomes indicating whether a city signed the agreement in the time period or not (0 / 1).
Therefore, for discrete-time data it can be noted that ’the dependent variable, although different
in form from the actual duration time, conveys the same information as the duration
time’ (Box-Steffensmeier & Jones 2004, 70). This fact then allows us to estimate such as
binary time-series-cross-section (BTSCS) data using standard logit or probit analysis as grouped
duration data (Beck et al. 1998). However, as Beck et al. (1998) point out, these techniques
consider the observations to be temporally independent. They argue that this assumption is
not true for many applications. Estimating standard logit on BTSCS data in the event of
temporal dependence leads to estimates that are consistent, but inefficient. The standard errors
are wrong and inflate t-values (Beck et al. 1998). They recommend the use of cubic splines to
account for potential time dependence. In a recent paper, Carter & Signorino (2010) suggest
a simpler alternative, i.e., the use of a cubic polynomial approximation by including time (t),
time squared (t2 ), and time cubed (t3 ) in the model. This fix should help to control for time
128
effects in the data that are due to the fact that the propensity to sign the MCPA in one month
is related to the city’s behavior and propensity in the past months. I therefore decide to follow
Bernauer, Kalbhenn, Koubi & Spilker (2010) and Perrin & Bernauer (2010) in using Carter &
Signorino (2010)’s approach.
In the following section, I present and discuss the results from my analysis of cities’ adoption
of the MCPA from 2005-2010 using my new data set.
4.5 Analysis and Results
What are the factors that influence a city’s decision to partake in voluntary climate change
agreements like the MCPA? In section 4.3, I have presented three groups of determinants –
internal city-specific factors, external factors, and specific events – that are supposed to matter
for participation in voluntary climate change initiatives. The first part of this section shows
the main results from the statistical analysis. In the second part, I present qualitative evidence
from the questionnaires and give an overview over the factors that mayors stress as having been
important in their decision to join the MCPA.
4.5.1 Statistical Analysis
Main results
Table 4.4 presents the main results explaining MCPA participation. Proceeding in the order of
the theoretical part, I first test hypotheses 1a-d in model 1 in table 4.4.
Internal factors
Hypothesis 1a postulates that a higher demand for voluntary climate change
policies increases the propensity of a local government to participate in the MCPA. As hypothesized, a higher % of democratic votes significantly increases the likelihood of participation. The
effect of the % of people in the city having at least obtained a bachelor degree – a proxy for
the exposure to information about climate change – is also positive, but does not reach standard levels of significance. The same is true for the positive coefficient on median per capita
income in the city. I now turn to the factors that decrease demand for climate change policies
and, correspondingly, are hypothesized to also decrease the propensity for MCPA participation.
Here, a high unemployment rate as well as a higher median age have the predicted negative and
significant effect, whereas dependence on industry and manufacturing is statistically insignificant. Another indicator that was associated with lower demand is the Poverty level. However,
Testing Participation in Space and Time
129
Table 4.4: Main Results
(1)
Internal
Internal Factors
p.c. income
% Bachelor&up
% Dem Vote
Population
Median Age
Unemployment
Industry dep
Poverty level
% loc gov empl
ICLEI member
0.00
0.01
0.03***
0.00***
-0.05**
-0.12**
-0.26
0.03**
0.04
1.90***
(2)
External
(0.00)
(0.01)
(0.01)
(0.00)
(0.02)
(0.05)
(0.21)
(0.01)
(0.09)
(0.49)
External Factors
No.of Neighbors
% inc group signed
% pop group signed
Winter Meet.01-08
Winter Meet.01-07
Annual Meet.06-07
Annual Meet.06-06
-0.04
-0.18***
0.05***
2.24**
1.85**
1.40
2.49***
(0.03)
(0.06)
(0.01)
(1.06)
(0.78)
(1.18)
(0.58)
Events
Mayoral campaign
Obama Elected
EECBG Feb 09
EECBG July 07
t
t2
t3
Constant
0.11**
-0.00
0.00
-6.16***
Log lik.
Observations
-1017.89
43228
Standard errors in parentheses
* p < 0.10, ** p < 0.05, *** p < 0.01
(3)
Events
(0.05)
(0.00)
(0.00)
(1.01)
0.06
0.00
-0.00
-6.12***
-993.27
42851
(0.05)
(0.00)
(0.00)
(0.44)
(4)
Combined
-0.00
0.02*
0.02***
0.00
-0.03
-0.12**
0.02
0.03*
0.05
1.68***
(0.00)
(0.01)
(0.01)
(0.00)
(0.02)
(0.05)
(0.23)
(0.02)
(0.11)
(0.61)
-0.04
-0.18**
0.04***
1.89
1.62**
1.46
2.51***
(0.03)
(0.07)
(0.01)
(1.22)
(0.77)
(1.09)
(0.63)
1.46***
0.20
-0.42
-0.08
(0.26)
(0.39)
(0.40)
(0.27)
1.43***
-0.16
-0.42
-0.05
(0.25)
(0.42)
(0.41)
(0.29)
0.06
-0.00
-0.00
-6.47***
(0.05)
(0.00)
(0.00)
(0.39)
0.04
0.00
-0.00**
-6.59***
(0.05)
(0.00)
(0.00)
(1.06)
-1054.17
43600
-947.79
42484
130
it is surprising that a higher % of people in the city living in poverty seems to lead to a higher
likelihood of signing the MCPA. This counterintuitive result might somehow be due to the fact
that a higher number of poor people live in large cities that usually are very active on voluntary
climate change policies.
Concerning the supply side explanations for MCPA participation, hypothesis 1c stipulates
that a higher administrative capacity positively influences the propensity of cities to participate.
The two variables proxying the administrative capacity of a city are the % of people working for
the local government and the size of the city measured by its population. While both show the
theoretically predicted positive influence on the likelihood of participation, only the coefficient on
population is statistically significant. Large cities usually have a higher administrative capacity
in the sense that they tend to have specialized personnel that knows about recent developments,
e.g., with regard to local climate change policies.
Testing hypothesis 1d on the local government’s previous engagement with local climate
change politics, we can see that a city’s membership in ICLEI is indeed associated with a
higher likelihood of joining the MCPA. This confirms the theoretical assumption that cities that
already have a background in the local implementation of climate change mitigation policies are
more likely to also engage their communities in further initiatives.
External Factors
I now turn to the external factors that are considered to impact on a local
government’s propensity to adopt voluntary policies. In model 2, we can see that – counter
to theoretical expectations – the number of neighbors that have signed the agreement already
seems to have a negative effect. This result could hint to a free-riding logic with respect to
MCPA participation. With an increasing number of neighbors joining, cities may chose to
free-ride on the efforts of others. While this would generally be a realistic scenario given the
global public good character of environmental quality with respect to global climate change,
this seems to be an unlikely scenario in the case of voluntary policies, that might even be
purely symbolic. Therefore, another possible explanation might be that political benefits for the
local government from MCPA participation decrease the more neighbors have already signed.
However, the coefficient on the neighbors variable is not statistically significant at conventional
levels and could also be equal to zero.
In contrast, the two different conceptions of peer effects between similar income (hypothesis
2b) and population groups (hypothesis 2c) are both significantly associated with MCPA partic-
Testing Participation in Space and Time
131
ipation. Regarding the peer similarities through city size, the positive coefficient means that a
city is more likely to participate in the MCPA the higher the percentage of cities that had signed
the agreement up until the previous time period is. This confirms the influence of similar size
on participation. This effect seems to dominate the ’simple’ neighbors effect when testing them
together. Testing the robustness of the main results, I will provide further evidence concerning
the null finding for the influence of geographic neighbors.
Hypothesis 2d stipulates that city governments that actively participate in social networks of
cities are more likely to sign up for the MCPA. Proxied by the local government’s attendance
record in USCOM biannual meetings, we can see that the participation in the two winter meetings of 2007 and 2008 significantly increased the likelihood of MCPA adoption. The strongly
significant result for the winter meeting 2007 might be explained by the fact that the second day
of the meeting was specifically dedicated to energy and climate protection issues.22 A related
topic on the meetings’ agenda was the call for an ’Energy and Environmental Block Grant.’
One year later, the then called Energy Efficiency and Conservation Block Grants were again
discussed at great lengths culminating in a letter to the House of Representatives’ budget committee to appropriate $2 billion for the EECBG Program, signed by about 140 participants.23
These discussions also might have influenced attending cities’ mayors or representatives in their
decision to join the MCPA. Furthermore, participating in the annual meeting in Las Vegas in
2006 also increases the likelihood of signing the MCPA. Again, the energy crisis of 2006 and
rising fuel prices were the main topics of this meeting, which may again have been pivotal in
persuading attending mayors to participate in the agreement. Surprisingly, the attendance of
the 2007 annual meeting in Los Angeles does not appear to have spurred MCPA participation
despite having ’Global Warming & Climate Protection’ as an agenda item.24 In sum, it can be
stated that the hypothesized positive effect of interacting with other cities is strongly supported
by model 2.
Events
Model 3 in table 4.4 includes the four indicators measuring the influence of events
on participation. In hypothesis 3a, I have postulated that mayors are more likely to sign the
MCPA during the time leading up to a mayoral election to show leadership and initiative. Indeed,
the coefficient on Mayoral campaign is statistically significant and positively associated with a
22
Further information accessible at http://www.usmayors.org/75thWinterMeeting/
Further information accessible at http://www.usmayors.org/76thWinterMeeting/
24
All other network variables were dropped during estimation because they completely determined failure.
23
132
higher participation propensity. Counter to expectations, the Obama Administration leads to
a higher likelihood of participation. However, the coefficient is not significantly different from
zero. The same applies for the remaining events concerning potential incentives for participation
in the form of funding for cities via the Energy Efficiency and Conservation Block Grant. Both
coefficients do not reach statistical significance and are negatively related to the adoption of
voluntary climate change policies. This counterintuitive effect might be either due to the fact
that the time period in which positive effects of these grants kicked in were not correctly specified,
or MCPA participation was not seen as an advantage to obtain these grants. Therefore, only
hypothesis 3a on the relationship between a mayoral election date and the likelihood to sign the
agreement received statistical support in the model that included only the events.
Robustness of main results
Model 4 in table 4.4 serves as the baseline model and includes all theoretically specified influences
for MCPA participation. Combining all factors shows that some of the internal determinants
loose in significance while the external factors remain largely unchanged. That is, even when
controlling for the effects of city internal characteristics and time-specific events, city-external
determinants – the participation behavior of peers in the same population (income) group – are
strongly associated with a higher (lower) likelihood of MCPA adoption.
A further observation from table 4.4 is that the coefficients on the cubic polynomials (t, t2
t3 ) are hardly statistically significant throughout models 1-4. The question arises whether one
might want to exclude them from these models altogether. However, a likelihood ratio test
comparing models 1 and 2 of table 4.5 against the hypothesis that all three coefficients on the
cubic polynomials are jointly zero led me to come to the conclusion that I need to include
them. Furthermore, it can be seen that estimating the combined model without t, t2 t3 does not
change the results greatly. The effect of the Obama administration now has a significant and
negative impact on the likelihood of adoption. This effect is picked up by the cubic polynomials
in the baseline model. Conversely, the negative and significant coefficient on % inc group signed
vanishes.
Having established the need to control for time effects, 4.8 shows a non-monotonic hazard
that first increases and then turns around. The approximated hazard rate indicates how likely
’failure’ or, in this case, the signing of the MCPA is to occur as time progresses (Beck et al. 1998).
The non-monotonic shape depicted in figure 4.8 is a common feature in research on the diffusion
Testing Participation in Space and Time
133
Table 4.5: Robustness of main results I; exclucding t and reduced model
(1)
Combined
(2)
w/o t
Internal Factors
p.c. income
% Bachelor&up
% Dem Vote
Population
Median Age
Unemployment
Industry dep
Poverty level
% loc gov empl
ICLEI member
-0.00
0.02*
0.02***
0.00
-0.03
-0.12**
0.02
0.03*
0.05
1.68***
(0.00)
(0.01)
(0.01)
(0.00)
(0.02)
(0.05)
(0.23)
(0.02)
(0.11)
(0.61)
-0.00
0.02*
0.02***
-0.00
-0.03
-0.10*
0.02
0.03*
0.05
1.74***
(0.00)
(0.01)
(0.01)
(0.00)
(0.02)
(0.05)
(0.23)
(0.02)
(0.11)
(0.61)
External Factors
No.of Neighbors
% inc group signed
% pop group signed
Winter Meet.01-08
Winter Meet.01-07
Annual Meet.06-07
Annual Meet.06-06
-0.04
-0.18**
0.04***
1.89
1.62**
1.46
2.51***
(0.03)
(0.07)
(0.01)
(1.22)
(0.77)
(1.09)
(0.63)
-0.01
0.00
0.04***
2.05*
2.04***
1.61
2.63***
(0.03)
(0.03)
(0.01)
(1.20)
(0.78)
(1.10)
(0.61)
Events
Mayoral campaign
Obama Elected
EECBG Feb 09
EECBG July 07
1.43***
-0.16
-0.42
-0.05
(0.25)
(0.42)
(0.41)
(0.29)
1.32***
-0.78**
0.08
0.07
(0.24)
(0.33)
(0.40)
(0.29)
t
t2
t3
Constant
0.04
0.00
-0.00**
-6.59***
(0.05)
(0.00)
(0.00)
(1.06)
Log lik.
Observations
-947.79
42484
Standard errors in parentheses
* p < 0.10, ** p < 0.05, *** p < 0.01
-6.24***
-956.83
42484
(1.00)
(3)
reduced
0.02***
0.02***
(0.01)
(0.01)
-0.10*
(0.06)
0.04***
(0.01)
1.68***
(0.63)
-0.16**
0.04***
(0.07)
(0.01)
1.61**
(0.78)
2.50***
(0.62)
1.47***
(0.25)
0.06
0.00
-0.00
-8.09***
(0.05)
(0.00)
(0.00)
(0.60)
-952.73
42484
0
.01
.02
.03
.04
.05
134
0
20
40
60
time
99% confidence interval
expected signature behavior
Figure 4.8: Baseline hazard
of innovation or policy adoption. The cumulative s-shaped curve of percentage of adopters over
time that is usually shown (c.f. figure 4.3) corresponds to the bell-shaped hazard rate (Rogers
1995). Reasons why the underlying hazard in figure 4.8 is not completely bell-shaped could be
that, due to right censoring occurring in July 2010, the height of the curve is either not yet
reached for the cities in my sample, or is just about to decline. Given that many of the smaller
cities were still joining the agreement in 2010, this might be an explanation.
Returning to table 4.5, model 3 is a reduced model that includes only significant predictors
of the likelihood of MCPA participation from the baseline model. The substantive coefficients
remain the same, however, the levels of statistical significance slightly changes for some variables.
This might be due to issues of multicollinearity. In the present of such multicollinearity, standard
errors are inflated, which leads to more conservative interpretations of an estimates’ significance.
Since the coefficient size should not be affected in such a large-N framework (Carter & Signorino
2010) and the inclusion of the variables is theoretically warranted, this should not provide a
major problem for the results. However, to be cautious about the potential influences on the
substantive results and a check on the robustness of the main results especially concerning the
external factors, I have estimated the baseline model and added each of the contingency factors
Testing Participation in Space and Time
135
in turns in table 4.6. Model 2 shows that when only the number of neighbors that signed
the MCPA is included as the geographical contiguity effect, the coefficient also does not reach
statistical significance. Similarly, the coefficients on % inc group signed and % pop group signed
do not change at all if included separately into the model. However, the coefficients on some of
the city-specific variables change. This might be due to the exclusion of relevant external factors
or multicollinearity. Since the fundamental results do not change, this does not warrant further
concerns.
There are two further checks of the baseline results that are shown in table 4.7. Here I first
exclude the variables pertaining to the influence of (social) city networks on the likelihood of
MCPA participation. One might argue that in the context of the sample including all cities over
10.000 inhabitants, these networks are not open for smaller cities. This is true in that USCOM
usually only represents the interest of larger cities, even though every city can participate in
the Mayors Climate Protection Agreement. Also, small cities can still be informed about the
policies associated with the MCPA even if they do not attend the meetings themselves. Model 2
in table 4.7 confirms that excluding these variables does not change the results at all. In section
4.4.2, it was mentioned that some of the city’s MCPA adoption dates had to be imputed due to
lack of precise information. Model 3 now excludes those imputed cities and their values from
the estimation. Once again, we can see that the results do not change.
As a final check on the baseline models’ ability to account for internal city-specific factors,
external determinants, as well as time-specific events, I have reduced the sample size to only
include cities over 30.000 inhabitants. For this smaller sample, we can now test the hypothesis
1b, namely that the propensity to sign the MCPA is higher for a city with a mayor-council as
compared to one with a council-manager form of government. In model 1, this assumption is
confirmed by the positive and significant coefficient on Mayor-Council. In models 1-4 in table
4.8, we can see that the results for most of the other variables are similar to the ones obtained
from the larger sample in table 4.4. However, combining the different determinants of MCPA
adoption in model 4, some of the coefficients turn insignificant. For example, of the external
factors in model 2, only participation in the two meetings is significantly related to an increase
in the likelihood of joining the MCPA. None of the contingency or interdependence factors reach
conventional levels of significance. Generally, this seems to point to a potentially more important
role of the city-specific factors in cities with more than 30.000 inhabitants. We can see in model
4, that although a higher % of college-educated individuals, voters for the democratic party,
136
Table 4.6: Robustness of main results II; external factors
(1)
Combined
(2)
Neighbor
(3)
Income group
External Factors
No.of Neighbors
% inc group signed
% pop group signed
-0.04
-0.18**
0.04***
(0.03)
(0.07)
(0.01)
-0.02
Internal Factors
p.c. income
% Bachelor&up
% Dem Vote
Population
Median Age
Unemployment
Industry dep
Poverty level
% loc gov empl
ICLEI member
-0.00
0.02*
0.02***
0.00
-0.03
-0.12**
0.02
0.03*
0.05
1.68***
(0.00)
(0.01)
(0.01)
(0.00)
(0.02)
(0.05)
(0.23)
(0.02)
(0.11)
(0.61)
-0.00
0.01
0.03***
0.00**
-0.04*
-0.11**
-0.16
0.03**
0.04
1.93***
(0.00)
(0.01)
(0.01)
(0.00)
(0.02)
(0.05)
(0.21)
(0.01)
(0.09)
(0.50)
-0.00
0.01
0.03***
0.00***
-0.04*
-0.11**
-0.15
0.03*
0.04
1.85***
External Factors II
Winter Meet.01-08
Winter Meet.01-07
Annual Meet.06-07
Annual Meet.06-06
1.89
1.62**
1.46
2.51***
(1.22)
(0.77)
(1.09)
(0.63)
2.37*
2.41***
1.88*
2.93***
(1.23)
(0.69)
(1.06)
(0.62)
Events
Mayoral campaign
Obama Elected
EECBG Feb 09
EECBG July 07
1.43***
-0.16
-0.42
-0.05
(0.25)
(0.42)
(0.41)
(0.29)
1.40***
0.05
-0.41
-0.05
t
t2
t3
Constant
0.04
0.00
-0.00**
-6.59***
(0.05)
(0.00)
(0.00)
(1.06)
0.06
-0.00
-0.00
-6.54***
Log lik.
Observations
-947.79
42484
Standard errors in parentheses
* p < 0.10, ** p < 0.05, *** p < 0.01
(0.03)
-0.16**
-989.71
43228
(4)
Pop group
(0.07)
0.04***
(0.01)
(0.00)
(0.01)
(0.01)
(0.00)
(0.02)
(0.05)
(0.22)
(0.01)
(0.10)
(0.49)
-0.00
0.02**
0.02***
0.00
-0.03
-0.10*
0.02
0.03*
0.04
1.69***
(0.00)
(0.01)
(0.01)
(0.00)
(0.02)
(0.05)
(0.23)
(0.02)
(0.11)
(0.61)
2.38*
2.17***
1.91*
2.86***
(1.24)
(0.70)
(1.06)
(0.62)
1.91
1.86**
1.41
2.62***
(1.22)
(0.77)
(1.10)
(0.62)
(0.26)
(0.40)
(0.41)
(0.28)
1.44***
-0.16
-0.42
0.01
(0.26)
(0.41)
(0.41)
(0.28)
1.46***
0.11
-0.37
-0.12
(0.26)
(0.40)
(0.41)
(0.29)
(0.05)
(0.00)
(0.00)
(1.01)
0.07
0.00
-0.00*
-6.53***
(0.05)
(0.00)
(0.00)
(1.03)
0.06
-0.00
-0.00
-6.95***
(0.05)
(0.00)
(0.00)
(1.07)
-968.66
42484
-951.79
42484
Testing Participation in Space and Time
137
Table 4.7: Robustness of main results III: networks and impute
(1)
Combined
(2)
w/o network
(3)
w/o impute
Internal Factors
p.c. income
% Bachelor&up
% Dem Vote
Population
Median Age
Unemployment
Industry dep
Poverty level
% loc gov empl
ICLEI member
-0.00
0.02*
0.02***
0.00
-0.03
-0.12**
0.02
0.03*
0.05
1.68***
(0.00)
(0.01)
(0.01)
(0.00)
(0.02)
(0.05)
(0.23)
(0.02)
(0.11)
(0.61)
-0.00
0.02*
0.02***
0.00
-0.04
-0.13**
0.00
0.03*
0.05
1.67***
(0.00)
(0.01)
(0.01)
(0.00)
(0.02)
(0.05)
(0.23)
(0.02)
(0.11)
(0.60)
-0.00
0.02**
0.03***
0.00
-0.04*
-0.12**
-0.02
0.02
0.06
1.63***
(0.00)
(0.01)
(0.01)
(0.00)
(0.03)
(0.06)
(0.24)
(0.02)
(0.12)
(0.63)
External Factors
No.of Neighbors
% inc group signed
% pop group signed
Winter Meet.01-08
Winter Meet.01-07
Annual Meet.06-07
Annual Meet.06-06
-0.04
-0.18**
0.04***
1.89
1.62**
1.46
2.51***
(0.03)
(0.07)
(0.01)
(1.22)
(0.77)
(1.09)
(0.63)
-0.04
-0.20***
0.04***
(0.03)
(0.07)
(0.01)
-0.04
-0.22***
0.04***
2.22*
1.67**
1.64
0.96
(0.03)
(0.08)
(0.01)
(1.26)
(0.80)
(1.10)
(1.08)
Events
Mayoral campaign
Obama Elected
EECBG Feb 09
EECBG July 07
1.43***
-0.16
-0.42
-0.05
(0.25)
(0.42)
(0.41)
(0.29)
1.40***
-0.16
-0.41
-0.12
(0.25)
(0.42)
(0.41)
(0.29)
1.46***
0.19
-0.39
0.17
(0.26)
(0.44)
(0.41)
(0.30)
t
t2
t3
Constant
0.04
0.00
-0.00**
-6.59***
(0.05)
(0.00)
(0.00)
(1.06)
0.06
0.00
-0.00*
-6.56***
(0.05)
(0.00)
(0.00)
(1.06)
0.04
0.00
-0.00**
-6.34***
(0.05)
(0.00)
(0.00)
(1.09)
Log lik.
Observations
-947.79
42484
Standard errors in parentheses
* p < 0.10, ** p < 0.05, *** p < 0.01
-956.81
42484
-858.85
41968
138
Table 4.8: Robustness of main results IV: reduced sample (cities >= 30.000)
(1)
Internal
Internal Factors
p.c. income
% Bachelor&up
% Dem Vote
Population
Median Age
Unemployment
Industry dep
Poverty level
Mayor-Council
% loc gov empl
ICLEI member
-0.00
0.03**
0.02**
0.00
0.04
-0.05
0.08
0.05*
0.39*
0.05
2.21***
(2)
External
(0.00)
(0.02)
(0.01)
(0.00)
(0.04)
(0.07)
(0.28)
(0.03)
(0.22)
(0.14)
(0.38)
External Factors
No.of Neighbors
% inc group signed
% pop group signed
Winter Meet.01-08
Winter Meet.01-07
Annual Meet.06-07
Annual Meet.06-06
-0.01
-0.17**
0.03***
2.07**
1.37*
1.49
2.20***
(0.02)
(0.08)
(0.01)
(1.05)
(0.82)
(1.10)
(0.57)
Events
Mayoral campaign
Obama Elected
EECBG Feb 09
EECBG July 07
t
t2
t3
Constant
0.03
0.00
-0.00
-8.70***
Log lik.
Observations
-582.87
11674
Standard errors in parentheses
* p < 0.10, ** p < 0.05, *** p < 0.01
(3)
Events
(0.05)
(0.00)
(0.00)
(1.45)
-0.01
0.00
-0.00
-4.76***
-580.32
11564
(0.05)
(0.00)
(0.00)
(0.43)
(4)
Combined
-0.00
0.04**
0.02***
0.00
0.05
-0.02
0.17
0.03
0.37
0.02
2.05***
(0.00)
(0.02)
(0.01)
(0.00)
(0.04)
(0.09)
(0.29)
(0.03)
(0.23)
(0.15)
(0.45)
-0.01
-0.11
0.01
1.93*
1.23
1.54
2.32***
(0.02)
(0.09)
(0.01)
(1.15)
(0.86)
(1.05)
(0.61)
1.10***
0.46
-0.85
-0.08
(0.29)
(0.49)
(0.57)
(0.35)
1.31***
0.13
-0.85
-0.09
(0.30)
(0.52)
(0.57)
(0.37)
0.01
0.00
-0.00
-5.10***
(0.05)
(0.00)
(0.00)
(0.40)
-0.01
0.00
-0.00*
-8.94***
(0.05)
(0.00)
(0.00)
(1.56)
-607.55
11872
-547.96
11369
Testing Participation in Space and Time
139
and former engagement in climate change significantly increase a city’s propensity to sign the
voluntary climate change agreement, most of the other coefficients are insignificant.
Predicted Probabilities
To get a grasp of the substantive effects of some of the indicators on the likelihood of MCPA
participation, I display (simulated) predicted probabilities in figure 4.9.25 The subfigures depict
the effects of two internal variables, % Democrats and unemployment, and the two external
variables that measure the peer effects of similar income and population group. In subfigure
4.9(a), we can see how the likelihood of MCPA participation steadily increases with a higher
percentage of democratic votes in the city. Going from the minimum value of % Democratic vote
to its maximum value increases the predicted probabilities only by about 0.4%. In subfigure
4.9(b), the probability of participation decreases the higher the unemployment rate in the city
is. Here, the confidence interval is discernibly narrower where the center of the data lies. The
negative effect on the predicted probability of signing the MCPA decreases only by around 0.3
% from having very low unemployment to very high values of unemployment. As expected in
the theoretical part, subfigure 4.9(c) shows the positive impact on participation by increases
in the percentage of cities in the same population group that participate in the MCPA. Until
about 50% of the cities in the population group have signed, the substantive effect is constantly
very small. However, if over 50% in the population group have signed, the predicted probability
increase steeply. So going from the mean value of the variable to its maximum value, the
predicted probability of joining the MCPA increases by around 6.4 %. A reversed threshold
effect can be observed for the % of cities of the same income group that have already joined (c.f.
subfigure 4.9(d)). Here, for low values of the % of cities in the same income group the negative
effect is very pronounced. However, this effect quickly approaches almost zero as soon as about
10% in the income group have signed. In predicted probability values, this means that the
negative effect of income is more pronounced when going from its minimum value to the mean
(-2.1 %) compared to going from its mean to the maximum values (-0.2 %). The main results
from these graphs are that, first, the predicted probabilities of MCPA participation from the
two contingency variables are larger in size compared to the ones from the city internal factors.
Second, there are threshold effects concerning the external variables. Recent research has found
similar threshold effects in the context of global environmental treaty ratification (Bernauer,
25
Simulations are based on the program CLARIFY by Tomz et al. (2003)
140
Kalbhenn, Koubi & Spilker 2010) as well as for the ratification of Long Range Transboundary
0
0
.002
.002
.004
.004
.006
.008
.006
Air Pollution (LRTAP) Agreements (Perrin & Bernauer 2010).
0
20
40
Democrats
99% confidence interval
60
80
0
expected signature behavior
5
10
15
Unemployment
99% confidence interval
25
expected signature behavior
.02
0
0
.05
.1
.04
.06
(b) Unemployment
.15
(a) % Democratic vote
20
0
20
40
60
% of population group that has signed
99% confidence interval
80
expected signature behavior
(c) % of same population group already signed
100
0
5
10
15
20
25
Inc_Group
99% confidence interval
expected signature behavior
(d) % of same income group already signed
Figure 4.9: Predicted Probabilities (Baseline model); other variables are kept at their mean;
binary variables at their median
From Table 4.9, we can see the predicted probabilities of joining the MCPA from the explanatory variables that take either the value 0 or 1. We can see that being a member of ICLEI
increases the probability to sign by 1.3 %. The strongest effect on the probability can be observed from having participated in the Annual meeting of the U.S. Conference of Mayors in June
2006. An attending cities’ probability to sign is over 3.1 % higher than that of comparable cities
that did not attend. To further illustrate substantive effects, I have simulated probabilities of
joining for five Midwestern cities based on selected covariates (while keeping all others at their
means) at the beginning of the observation period (in May 2005) in table 4.10. I have included
Testing Participation in Space and Time
141
Table 4.9: Simulated predicted probabilities
Min to max
ICLEI member
Annual Meet.06-06
Winter Meet.01-07
Mayoral campaign
.0126
(.0097)
.0315
(.0203)
.0131
(.0129)
.0072
(.0022)
the largest city in the sample, Chicago, as well as two other large cities, Cincinnati and Detroit.
These are joined by two mid-size cities, Duluth and Champaign. Whereas Duluth and Champaign are quite comparable on many of the variables shown, Duluth is a member of ICLEI while
Champaign is not. Furthermore, while Champaign has a higher % of educated people in the
city, Duluth had a higher percentage of democratic votes in their county in 2004. The predicted
probabilities from these constellations can be seen in the first column. Duluth, possibly due
to its membership in ICLEI, has a predicted probability of joining the MCPA that is about 6
times higher than that of Champaign. This higher simulated probability for Duluth matches the
real life developments. While Duluth joined the MCPA already in July 2005, Champaign has
still not signed on to the voluntary initiative. A similar observation can be made for the three
larger towns. Also being an ICLEI member, Chicago’s probability of signing is considerably
higher than that of the other towns. Indeed, Chicago joined in August 2005, whereas Cincinnati
signed up in June 2006 and Detroit – as I already mentioned above – only became a member in
February 2009.
142
City
Chicago
Detroit
Cincinnati
Duluth
Champaign
44.3
28.2
26.6
11.0
25.5
%bachelor&up
18671
18969
19964
14711
20168
p.c income
50.37
65.20
47.09
69.39
70.25
% Dem Vote
7.1
7.5
7.3
13.8
10.1
Unempl
0
0
16.667
16.667
16.667
% of pop group
0
1
0
0
1
ICLEI
67518
86918
331285
951270
2896016
Population
Table 4.10: Simulated Probabilities, City Examples (for May 2005)
Simulated probability*
0.050
(0.070)
0.002
(0.001)
0.002
(0.001)
0.012
(0.007)
0.002
(0.001)
Note: Robust standard errors in parentheses; *PR(signature = 1)
Testing Participation in Space and Time
143
4.5.2 Qualitative Evidence
This part of the analysis now looks into evidence gathered from the mail questionnaires. Table
4.11 lists the 23 cities that have returned the questionnaire (c.f. 4.7). Although this sounds like
a very low rate of response given it was sent to 150 cities, it has to be borne in mind that the
priority of the data collection effort was given to the date of signature.
The questionnaire, which is included in section 4.7 contains questions about the MCPA, asking
for example how the community had first heard about the initiative and what the main factors
were that influenced the mayors decision to sign the agreement. I also ask a battery of questions
about the MCPA, the city and some personal and demographic information about the mayor.
Table 4.11: Qualitative evidence is based on returned questionnaires from the following cities
State
Name
IL
IL
IL
IL
IL
IN
IN
IN
MI
MI
MN
MN
MN
MN
MN
MN
MN
MO
OH
OH
OH
OH
WI
Rolling Meadows
Wilmette
Evanston
Hoffman Estates
Schaumburg
Bloomington
Evansville
Michigan City
Marquette
Southgate
Austin
Bemidji
Hutchinson
Rosemount
Burnsville
Eagan
Roseville
Saint Louis
Alliance
Cincinnati
Parma
Warren
Madison
Mayor or
Representative
Kenneth A. Nelson
Christopher S. Canning
Elizabeth Tisdahl
William D. McLeod
Al Larson
Mark Kruzan
Jonathan Weinzapfel
Chuck Oberlie
John Kivela
Joseph G. Kuspa
Thomas A. Stiehm
Richard Lehmann
Steve Cook
Bill Droste
Elizabeth B. Kautz
Mike Maguire
Craig Klausing
Francis G. Slay
Toni E. Middleton
Larry Falkin
Dean E. Depiero
Michael J. O’Brien
Dave Cieslewicz
Pop >=30k
0
0
1
1
1
1
1
1
0
1
0
0
0
0
1
1
1
1
0
1
1
1
1
% Dem
70.25
70.25
70.25
70.25
70.25
53.43
40.72
49.56
53.60
69.39
60.99
50.13
36.45
48.48
48.48
48.48
63.04
80.29
50.59
47.09
66.57
61.65
66.02
The following brief overview of the mayors’ and city representatives’ answers fulfills two functions. First, qualitative in-depth evidence is given to corroborate the findings from the large-N
analysis and to illuminate these results with anecdotal evidence. In providing evidence, all three
determinants of voluntary action considered in the theoretical part – internal factors, external
144
factors as well as events – are considered. Second, beyond the explanation concerning the attribute, spatial, and event data from the above analysis, the answers of the questionnaires allow
to a certain extent an interpretation of the deeper motivational forces that led mayors to sign the
MCPA as well as their evaluation of the effects this signature had concerning their reputation.
On the level of mayoral characteristics, there are seven mayor who indicate that they consider themselves a Democrat, five consider themselves as Independent whereas four mayors see
themselves as Republicans.26 Asking what the impact of their decision to sign the agreement
was first on public sentiment towards them as a mayors and second on their chances for reelection, one can see that while only two out of four republicans indicate that their participation
had a positive effect on public sentiment, all indicate that it had no effect on their chances for
reelection. For the Democratic mayors, however, this looks different. While all except one of
the Democrats indicate a positive effect of their joining on their communities sentiment, four
even see a positive effect from participating in the agreement on their chances for reelection. I
now to the question concerning the factors that led the mayors to sign the MCPA.
Here again, one can see a slight difference between Democratic mayors and their Republican
counterparts in the sense that Democrats more often relate their reasons for signature to the
failure of the federal government whereas Republican mayors never name this as a reason. This
perceived failure is formulated by Mayor Kruzan of Bloomington, IN: ’Instead of joining 141
nations that ratified [the Kyoto Protocol], the US adopted wait and see attitude. Wait and see
is a choice. And it’s a choice that led this nation into energy dependence, a degraded economy,
and I would argue, war ’. In total, inaction on the federal level was given as a reason by three
Democrats and two of the Independent mayors.
Notably Democratic mayors not only give the inaction of the federal government as main
factor for signing the agreement, they also answer the question of why the initiative got so successful with this argument. With regard to the four Republican mayors, they indicate personal
concern (Alliance, OH), the city’s former engagement in environmental and sustainability issues
(Burnsville, MN) and quality of life and health issues (Rolling Meadows, IL and Bemidji, MN).
Although not backed up by data on the actual implementation of local climate change policies
in these cities, the fact that such a high percentage of Democratic mayors name federal inaction
as their main reason might suggest that political benefits were reaped due to action against the
federal government. Comparing the actions implemented in these communities with the ones
26
The remaining had not provided information on this question.
Testing Participation in Space and Time
145
implemented in the communities where the mayors indicated that they joined out of personal
concern for the environment would be an interesting extension to gain evidence on whether
policies are symbolic or substantive.
Overall, the two main factors that were named to have an positive effect on the decision to
sign the agreement were on the one hand environmental concerns (on a personal level (Alliance,
OH; Hutchinson, MN), due to general preferences of the village (Hoffman Estates), or from
environmental comissions (Evanston, IL)) and interlinked with that, on the other hand a desires
to increase quality of life in the city. The quality of life arguments ranges from arguments
of community design (’This Agreement dovetails perfectly into promoting healthier living by
walking, biking and enjoying our beautiful natural environment’ Mayor Kivela, Marquette, MI)
to the minimization of industrial effects (’The majority of our local electricity is generated by a
coal fired plant within our city limits and we are constantly reminded to the need to clean that
production within regulated utility’ Mayor Oberlie, Michigan City, MI). This confirms studies
such as Betsill (2001); Warden (2007); Kousky & Schneider (2003) on the importance of local
co-benefits from voluntary climate change policies on the local level.
Interestingly, one factor which was named several times as main factor for signing was the
future economic well-being of the city. As stated in the theoretical framework, this is a counterintuitive result in the realm of global climate change. However, some communities expect future
gains due to energy savings through green policies (Mayor Droste, Rosemount, MN: ’With a
lowdown in the economy we are looking at ways to save money. We were looking at efficient
lighting and heating in our older buildings. We were also making assessments regarding all of the
lawn mowing and watering we were doing’). As mentioned above, the notion that the initiation
of the EECBG program in 2007 was decisive for the participation of communities in the MCPA
– though not backed by the statistical analysis – was an incentive for at least two of the communities in the sample. Mayor Depiero of Parma, OH stated as his main reason for participation
that: ’HR2447 was designed to create the Energy and Efficiency Block Grant, which at the time
of passage of legislation, would have provided the City of Parma over $1.000.000 to use toward
energy efficiency projects for our businesses and residents. Signing the agreement was the first
step towards advocacy of the legislation’. Furthermore, Mayor Oberlie of Michigan City also
refers to the EECBG legislation as an incentive for participation.
Shifting focus now on the question how mayors initially heard about the MCPA, it becomes
obvious, that the network effects that came out of the statistical analysis as one of the main
146
finding, indeed play an important role for mayors. Nearly half of all respondents first came in
touch with the initiative through the Conference of Mayors (actual attendance of conferences or
through the newsletter), the National League of Cities or other local mayor’s networks. Local
and national media as well as direct contact to citizens or political entrepreneurs (Commissioners,
the Sierra Club) were also mentioned as bringing the MCPA to the mayor’s attention.
The mayors’ answers as to why the agreement spread so quickly are overwhelmingly based
on the opinion that the climate change topic recived increased attention in the public and the
media. This greater awareness is in the mayors’ opinion linked with the aforementioned inability
of the federal government to act and the threat of rising fossil fuel prices, which again lead to
more media attention and hence gave feedback to rising public concerns about the climate
change topic. Mayor Weinzapfel of Evansville, IN notes the predominant view in his answer
why the MCPA spread so quickly: ’National public focus on climate change and greenhouse gas
emissions. Rising concerns about air quality and public health. Rising fossil fuel prices and
international oil-related issues’. This shall only be completed by the statement of Schaumburg’s
Mayor Larson who sees one of the main factors in the public’s ’realization that global warming
is a real threat’.
Besides the mayors of Rosemount, MN, who experienced opposition against his signature of
the MCPA from ’a council member of a neighboring city’ and Alliance, OH, where opposition
came from individuals of the community, no other mayor experienced opposition against joining
the agreement.
Concerning the influence of other cities on the city in question, mayors usually state their
geographically closest neighbors alongside the next larger city in the region when asked about
with which cities they have the closest ties. However, Larry Falkin, the Director of the city’s office
of environmental quality in Cincinnati, OH27 , when asked about peers answered that Cincinnati
benchmarks itself against, Cleveland, Pittsburgh, Indianapolis, Lexington and Louisville, which
are all cities between 250.000 inhabitants and nearly 800.000 inhabitants located within about
250 miles. This supports the argument that cities see peers within their similar size group or
those larger.
27
Since Cincinnati was also part of the U.S. sample for which I conducted interview, they were asked whether
they want to fill out a questionnaire or conduct an interview and they decided for the interview. Therefore,
information on Cincinnati emanates from an interview asking all questions on the questionnaire and a few
more.
Testing Participation in Space and Time
147
The notion of policy entrepreneurship of cities and how it might pay off for the city can also
be illustrated with the city of Cincinnati in Ohio. Cincinnati has a long history of environmental
leadership and had its first office for environmental management from the early 1990s. It signed
the Mayors Climate Protection Agreement in June 2006. The answer to cooperations of Cincinnati in the realm of climate change policies is the following: ’We’ve been sort of contacted by
a variety of other cities who are interested in different pieces of what we are doing. Columbus,
Ohio, our capital city is one that we sort of coached quite a bit, because they are making some
great strides there, but they are a little bit newer to the game than we are, and being our capital
city they have a lot to do with how state politics happen so greening Columbus has some special
advantages for us in terms of the state policies that it helps produce’.
To conclude this brief overview of the qualitative evidence, it can be said that as expected,
one overall motivation for a cities’ participation does not exist, but we can see indeed that the
mayor as the decision-maker can also act as a policy entrepreneur and on its own agenda. This
has served as a first analysis to uncover the temporal and spatial pattern in signing the MCPA.
The following section provides suggestions for further research and concludes this paper.
4.6 Conclusion
In this paper I examined the determinants of cities’ willingness to join the MCPA in cities over
10,000 inhabitants in seven Midwestern states. I developed arguments on how various factors
influence local governments’ decision to voluntarily contribute to climate change mitigation
efforts by joining the Mayors Climate Protection Agreement. These factors were considered to
be city-specific internal factors, external factors, and time-specific events.
As far as internal factors are concerned, I have argued that there are specific determinants
leading to a higher demand for voluntary climate change policies within the city, e.g. a higher
percentage of voters supporting the Democratic Party, which then make governments more likely
to participate in voluntary efforts. I further assumed that the initiative to join the MCPA may
emanate from the supply side of local government. Here, the existence of a policy entrepreneur
and a mayor-council form of government were considered to increase the propensity to participate. Interdependencies between cities were conceptualized as emanating from geographic
linkages to other jurisdictions, peer group effects, as well as from participation in social networks. The consideration of certain time-specific events, such a mayoral election within the city,
148
complemented my theoretical framework. To test my claims, I have used a unique new data set
on monthly signing behavior of 749 Midwestern cities from February 2005 to June 2010.
The analysis has shown that to explain the participation of voluntary climate change initiatives, it does not suffice to look at city or local government characteristics only. While favorable
conditions, such as a more liberal and well-educated populace as well as the existence of a policy entrepreneur, measured by former engagement in climate change policies, do increase the
probability of an MCPA signature, other factors also have to be taken into account. I find that
a city’s participation in social networks as well as the MCPA participation of cities within the
same population group increases, whereas the joining of cities within the same income group
decreases a city’s likelihood to participate. Although internal characteristics, such as the % of
Democratic votes in the city or the human capital endowment, are also significantly related to a
higher propensity to sign, the substantive effects of these variables are much smaller. As far as
pivotal events are concerned, I find that the likelihood of participation significantly increases in
the months leading up to a mayoral election in the city. Insights gained from this large-N study
are further supported by qualitative evidence from the analysis of questionnaires.
I have argued that the cities’ decision to join the voluntary climate change effort would depend
on the participation of geographic neighbors, cities within the similar income or population
group, as well as on the participation in social networks. Whereas the impact of networks
that diffuse knowledge about the initiative and, thereby, increase participation by those cities,
which are part of them, is one of the strongest findings in both the quantitative as well as the
qualitative part of this analysis, the impact of neighboring cities is not found to be relevant for a
city’s membership in the MCPA. This contradicts some of the findings of studies by Vasi (2006)
and Krause (2011a), who both found a significant effect of neighbors on participation in voluntary
climate change initiatives. On the county level, Brody et al. (2008) and Schaffer (2010) have also
pointed to such interdependence based on geographical contiguity. A theoretically expected and
substantively large influence for a cities’ decision to join emanates from whether cities that have
a similar size have already joined. The finding from the statistical analysis that the population
group seems to matter for participation in the MCPA was further backed by qualitative evidence
suggesting that cities pay attention to what their peers (in terms of population size) do. In
conjecture with the non-finding for the influence of geographical neighbors, future research
might construct even more fine-grained geographic measures of influence that are contingent on
population sizes and may account for asymmetric influences. This would allow small cities to
Testing Participation in Space and Time
149
be influenced by geographically close small and large cities, whereas large cities would be only
influenced by similar sized or larger cities.
As highlighted above, a further notable finding that emerged out of the analysis is a persistently strong and significantly positive influence of a mayoral campaign in the city on MCPA
participation. An assumption that could be tested in future research is whether this effect is
conditional on demand. Given that there is demand for local climate change policies within
the community, one could imagine that even a local government or mayor that is usually not
an advocate of climate change policies might need to succumb to public demand and sign the
MCPA to signal cooperation on the issue in the face of an upcoming election.
Another extension deals with the possibility of structural breaks in the data that might have
changed the process. To this end, one could look into more detail whether there was a structural
difference between the process before and after the election of Barack Obama in November 2008.
Many potential differences in the reasons for participation in the two time periods are thinkable.
For example, whereas one might have expected more liberal and democratic localities to sign
before this election as a sign of dissatisfaction with the Bush Administration, one might expect
that with the new administration from 2009 onwards, local governments in areas with relatively
high Republican votes would decide to sign on to the agreement in preparation for possible
changes to national climate change policies.
A further question that emerges from the observation of the hazard rate as well as from descriptive statistics on the adoption rates of groups of cities of different population size deals
with the possible end of this process. The membership figures provided by the Mayor’s Climate
Protection Center have only change by two new members from July to November 2010 nationwide. It is not clear whether the trend will continue to decline or whether we might see another
rise in membership. One scenario could be that the national government continues to withhold
nationally binding legislation, but offers generous financial incentives to local government. But
how this impacts on the MCPA as a whole, i.e., whether it ceases to exist or is replaced by a
new agreement, remains to be seen.
Finally, future research should also consider the implementation of actual climate change policies as a consequence of this agreement in a large-N context. This would allow us to disentangle
purely symbolic participation of cities from cases, where cities have actually made progress on
the mitigation of GHG emission and have reduced their carbon footprint.
150
4.7 Appendix
4.7.1 Questionnaire
Dear Mayor XX,
My name is Lena Schaffer, I am a Research assistant and 3rd year PhD student in Political
Economy at the Swiss Federal Institute of Technology (ETH) in Zurich, Switzerland. I am
writing my PhD thesis on voluntary climate change commitments across U.S. cities and states
with a special focus on the U.S. Mayors Climate Protection Agreement (US-MCPA). According
to my records, your city is listed as being a participant in this agreement.
As far as my research is concerned, I am interested in how action and commitment by actors
at the local level spread across the United States, specifically in the Midwest, and how one can
explain the dynamics behind this process. To be able to test specific theories of the diffusion
of successful practices across time and space, I need the approximate date when you signed the
US-MCPA. This is the primary reason why I am contacting you, and an estimate as close as
possible to when the agreement was signed would be of great assistance to my research. Because
of the limited number of cities in my study, your participation is extremely valuable.
I am resending also the questionnaire on basic information about participation in the agreement, as well as a few questions about personal and city demographics. I know that you are
very busy, however if you could find the time to fill this out it would be greatly appreciated.
If you would be interested in knowing the results of my study, let me know and I would be
happy to send them to you in an email as soon as they are available.
If you have any further questions regarding this study, or would like additional information,
please contact me at (202-559-0677) or via email at climateresearch@ethz.ch.
Thank you very much in advance!
Best regards,
Lena Schaffer
Testing Participation in Space and Time
151
Center for Comparative and International
Studies (CIS)
Institute for Environmental Decisions (IED)
ETH Zürich
Weinbergstrasse 11, WEC C18
8092 Zürich
Switzerland
Lena Schaffer
Ph.D. Student
+41 44 632 0261
schaffer@ir.gess.ethz.ch
Questionnaire for the Research Project no. 20329 ETH-RDB “Voluntary Climate Change Initiatives in
the United States: Testing Spatial Dependence in Participation”
Lena Schaffer and Prof. Dr. Thomas Bernauer
When did you / your city join the Mayors Climate Protection Agreement (MCPA)?
Day
Month
Year
How did you first hear about this initiative?
Before signing, did you know of cities around you that had joined the initiative?
If yes, could you name them?
Comments:
Seite 1/6
152
What were the main factors that influenced your decision to sign the agreement?
Was your decision to sign the agreement influenced by any particular incentives offered
by the initiative?
If yes, could you identify them?
Comments:
Other than you personally, what people or groups (e.g. environmental NGO’s) within the
community were involved in getting the initiative started in your city?
Comments:
Seite 2/6
Testing Participation in Space and Time
153
Did you encounter opposition against your plans to sign the initiative? If yes, which
groups opposed this initiative? What were their main objections?
Group
Objections
Comments:
With which of the cities in your county / state do you have the closest ties?
What would be the main reason(s)?
Cities
Reasons
Comments
Have you implemented any new practices / city resolutions since joining the MCPA?
If yes, name and shortly describe 1-2 of the new practices / city resolutions.
New Practices / City Resolutions
Describe
Seite 3/6
154
Initiative as a whole:
In your opinion what were the most important factors that made this agreement spread
so quickly during its first years?
Are you a member in any other networks of mayors?
If yes, name the one or two most important ones.
Lastly
Do you think that your decision to sign the agreement has a positive, negative or no effect
on the public sentiment towards you presently?
Positive effect
No effect
Negative effect
Comments:
Do you think that your decision to sign the agreement might have a positive, negative or
no effect on your prospects for reelection?
Positive effect
No effect
Negative effect
Comments:
Seite 4/6
Testing Participation in Space and Time
155
Personal and Demographic Information
Age: How old are you?
What is your marital status?
Do you have children?
What is your educational background?
What is your profession?
How long have you been in politics?
How long / how many terms have you
been a mayor of this city?
In politics, do you consider yourself a Republican, a Democrat, an independent or of some
other party?
Republican
Democrat
Independent
Other
In economics, do you consider yourself a liberal or a conservative?
Liberal
Conservative
neither
Are you or have you been an active member within the U.S. conference of mayors, the
National League of Cities or any other national or state level organization? If yes in what
capacity?
Seite 5/6
156
Is your city vulnerable to natural disasters such as hurricanes/floods/droughts/tornadoes
or other natural disasters, and which?
If yes:
Has your city suffered from any major natural disaster during the last 5 years or since you
became the mayor of the city?
Seite 6/6
157
5 ’Where the Rubber Meets the Road’
Understanding the Mechanisms Leading to Cities’ Participation and
Non-Participation in the MCPA
Lena Maria Schaffer
5.1 Introduction
The previous chapters of this dissertation on local climate change policies have continuously
reduced their focus. Starting from a general overview of global climate change governance, I
have then concentrated on subnational units within the U.S. and have tried to explain countylevel participation patterns in the Mayors Climate Protection Agreement as well as city-level
diffusion effects of the initiative in a temporal setting. In these papers, mayors have been
implicitly assumed to be the ones that make the decision to participate in the MCPA, but have
not been explicitly asked for their reasons to participate or the reasons not to participate. How
did they hear about the initiative, and when? Did they have personal motivations to act on the
climate change issue or was there demand by citizens or industry to either participate in the
effort or not?
This chapter, therefore, provides three distinct contributions that relate to questions already
raised in the two chapters on the MCPA. First, I finally open the black box of the decision-maker
and ’let the mayors speak ’. I contrast the results obtained from my two quantitative large-N
analyses in chapters 3 and 4 with empirical findings from semi-structured interviews. To that
end, I have randomly selected 60 cities with more than 10,000 inhabitants, 30 of which are
signatories and 30 of which have not signed the MCPA, from the pool of 2700 cities throughout
the United States. Second, I focus on the unique circumstances of the initiative as a whole and
its potential impact on climate change policies within the U.S. Third and most importantly, I
further add to the literature on voluntary climate change policies by specifically asking about
reasons for non-participation. If we conceptualize the MCPA as a voluntary agreement that
158
inflicts virtually no cost upon the city who signs it, the question is: why have not even more
communities participated?
Evidence from the interviews with decision-makers confirm some of the findings of chapters
3 and 4 and add relevant new insights. Furthermore, talking to the mayors shed some light on
the question concerning the symbolic politics vs. genuine green activism debate discussed in
chapter 4. A result that can also be confirmed is that mayors, who regularly attend national
mayoral conferences, are more likely to sign voluntary agreements. One result that came out of
the previous chapters and that is strongly supported by the interviews is that decision-makers
in larger cities are more aware of the problem and are more likely to take action, whereas
smaller towns care more about distinctly local problems that are ’where the rubber meets the
road’ (Mayor Bush, Winter Spring, FL). A notable finding concerning the mayors, who have
not signed, is that some of them actually see the initiative as a declaration without real effects
and as nothing more than ’jumping on the climate change bandwagon as the road to political
prosperity’ (Representative from Californian city).
The chapter is organized as follows: I first introduce the research design of the study, highlighting how this study was conducted. In the analysis section, I then look at the assessments
from the interviews with a focus on reasons for non-participation and contrast them with those
for participation and with my earlier findings from the large-N studies. Finally, I briefly discuss
mayors’ evaluations of the MCPA initiative as a whole and the effects it has had on national
policy.
5.2 Research Design
The collection of qualitative interview evidence for this empirical section involved different steps.
The following paragraphs will deal with the sampling procedure, the preparation of the semistructured interviews and the actual interviews.
5.2.1 Sampling
The target population for this last dissertation chapter are cities with over 10.000 inhabitants.
From these 2.700, I have decided to draw a random sample of cities for which I would contact the
mayors to gain more information on their MCPA participation or non-participation. This avoids
bias on the side of the researcher in potentially selecting cases that are extreme or those known
to the researcher (Schnell et al. 2008). However, as we have seen from chapter 3 and chapter
’Where the Rubber Meets the Road’
159
4, MCPA non-participation is far more frequent than participation (23 %). Drawing a random
sample would probably overrepresent cities that have not signed. Schnell et al. (2008) suggest
the use of stratified random sampling in this case. I therefore added one strata (MCPA) along
which the elements from the random sample were drawn.1 Since I targeted to conduct around
15-20 interviews, I drew a sample of 60 cities, 30 of which had signed the MCPA. This higher
number of elements in the sample was needed to avoid resampling in case of non-response by the
units. The final sample therefore included 60 cities throughout the U.S. The map in figure 5.1
depicts these 60 sampled cities and their distribution throughout the U.S.. Cities in the sample
that have signed the MCPA are shown with red points, whereas those who are not members
of the MCPA are in violet. We can see that the sampling is somehow representative of actual
population concentrations with more cities having been sampled on the two coasts and in the
Midwest compared to the less populous western states. From table 5.1 it can be seen that not
each state was present in the sample. For example, Florida and California each had 8 cities in
the sample, Texas and Illinois had 7, while there was no city from Wyoming in the sample.
Figure 5.1: Random sample of 60 cities from 2700 cities according to their participatory status
5.2.2 Conduction of the Study
The main part of the study was conducted in April and May of 2010. All the 60 cities belonging
to the stratified random sample were then contacted via e-mail, informed about the study and
1
Stata 10’s sample command was used to execute the sampling.
160
Table 5.1: Distribution of states in the sample
STATE No.
%
AL
2
3.3%
CA
8
13.3%
CO
2
3.3%
FL
8
13.3%
ID
1
1.7%
IL
7
11.7%
IN
1
1.7%
KS
1
1.7%
MA
1
1.7%
MI
1
1.7%
MO
1
1.7%
NC
2
3.3%
NE
1
1.7%
NJ
2
3.3%
NY
1
1.7%
OH
3
5.0%
OR
1
1.7%
PA
3
5.0%
SC
1
1.7%
TN
1
1.7%
TX
7
11.7%
VA
1
1.7%
WA
1
1.7%
WI
3
5.0%
Total
60 100.0%
’Where the Rubber Meets the Road’
161
asked whether they would be willing to participate. A study information sheet (c.f. section
5.5.2) was sent along with the e-mail (c.f. section 5.5.1) and informed the mayors about the
duration of the interview and stated that, if they would give their consent, the interview would
be audio-recorded and used for an analysis. Furthermore, it also gave them the option to stay
anonymous.
After having sent out these first information sheet, we then contacted the cities via telephone
a few days later to ask whether they had seen the mail and were willing to participate in the
study.2 We managed to arrange 17 interviews with mayors and representatives, 7 of which had
not signed the agreement and 10 who were participants in the MCPA. We heard back from 12
other cities that their mayor or representative did not want to participate in our study or was
currently unavailable. The map in figure 5.2 now depicts those cities in which the mayor or a
city representative were interviewed with a star.
The questions for the interview firstly included personal characteristics of the mayors and
secondly some city-specific questions. The third bloc of question then dealt exclusively with
the MCPA. The interview sheet is presented in section 5.5.3. Finally, the interviews with the
mayors from the sample were done via Voice over IP (Skype). According to Diekmann (2004)
interviews over the phone generally do not differ from face-to-face interviews in terms of data
quality.
Figure 5.2: Interviewed cities from sample according to their participatory status
2
Research assistance by Gwen Tiernan is greatly acknowledged.
162
Table 5.2: List of Interviewees
State
City
Name
Ideology
Size
CA
CA
CA
CO
FL
FL
IL
MI
MO
NE
NJ
OH
PA
TN
TX
TX
WI
Redding
San Luis Obispo
Santa Rosa
Boulder
Gulfport
Winter Springs
Normal
Westland
Saint Peters
Fremont
Princeton
Cincinnati
Hanover Ê
Elizabethton
Harker Heights
McKinney
Wisconsin Rapids
Anonymous*
Dave Romero
Susan Gorin
Matt Appelbaum*
Mr. O’Reilley*
John F. Bush
Chris Koos
William R. Wild
Ron Darling*
Robert Hartwig*
Bernard P. Miller
Larry Falkin*
Ben Adams
Anonymous*
Ed Mullen
Brian Loughmiller
Mary Jo Carson
Independent
Republican
Democrat
n.a
Apolitical
Republican
Independent
Democrat
Democrat
n.a.
Democrat
n.a.
Republican
n.a.
Republican
Republican
Independent
80865
44174
147959
94673
12527
31666
45386
86601
51381
25174
14203
331285
14535
13372
17308
54369
18435
Note: * indicates if city official instead of mayor was interviewed
MCPA
0
1
1
1
1
0
1
1
1
0
1
1
0
0
0
1
0
’Where the Rubber Meets the Road’
163
5.3 Results
The following part provides an overview of the interviews and leads through the different issues
that mayors mentioned and that are especially relevant in view of the two previous dissertation
papers. Throughout this section, I deliberately give the statements a close reading and let the
mayors speak. Table 5.2.2 provides some information on the interview partners and the city.
5.3.1 Reasons for Non-Participation
The reasons for non-participation are an understudied area within local climate change policies.
Therefore, this research makes a substantial contribution to the existing literature. In the
following two paragraphs, I will first consider the two main reasons for non-participation in the
agreement that emerged out of the interview. First, mayors either do not see the need to do
something at this point in time and prioritize other issue (supply side), or second, they stress
that there simply is no demand for climate change policies within the community (demand side).
Some mayors that have not signed the agreement stress that although there is no opposition
against climate change policies and the MCPA in their communities, these policies were not
identified as having a high priority on the city’s agenda. Mayor Carson (Wisconsin Rapids,
WI) stated that there were no community factors due to the presence of which she would not
have signed the agreement. Instead, ’it was just the case of having too much on my plate with
economic development to get it done. [. . . ] So it was just about timing’. Mayor Bush (Winter
Springs, FL) emphasized the priorities of local government more metaphorically: ’Well, it’s just
not the priority of our commission. I mean, you gotta realize at the city level, we’re involved
where the rubber meets the road, the police department, the fire department, the water, the
roads, the sewers, that’s what we do’. He further emphasized that ’[p]eople at the local level
are interested in their taxes and their safety. And they don’t think it’s the city’s responsibility
to worry about climate change’.
Chapter 4 hypothesized that a city’s likelihood to sign the MCPA increases with a higher
demand for climate change policies within the city. Robert Hartwig, the city administrator of
Freemont, NE, when asked about the city’s reason for non-participation, confirms this in stating
that ’there hasn’t really been a decision to sign or not to sign, but there is no real pressure
from the community to sign’. One reason might be the lack of explicit demand. However, there
are also cities in which such a signature might be very unpopular. Mayor Ed Mullen of Harker
164
Heights, TX went so far as to say that: ’If I did that [signing the MCPA], they would run me
out of town’. However, he continued that this has only to do with the political issue of climate
change. He claimed that: ’I have one of the largest recycling efforts in the central Texas area
[. . . ], but I don’t do it necessarily for environmental reasons, I do it for economic reasons, because
recycling reduces the costs of waste management’. A representative of a large Californian city
mentioned that ’if he [the mayor] had signed it, it would have been in the newspaper and it
would have a negative effect, the fact that he did not sign means that it had no effect’. He
further clarified: ’So the idea of voluntarily joining an agreement like that is just not something
a politician that wants to get elected would consider here’. It therefore seems that mayors are
aware of their constituency’s position on the issue and act accordingly.
5.3.2 Reasons for Participation
At a first glance, it can be observed that many mayors, who signed the agreement, declare
that they want to improve the quality of life in their cities. Some mayors mentioned that their
decision to participate does not depend on whether climate change is real or not. Rather, the
goals and policy instruments that are suggested by the agreement have local co-benefits in the
area of, e.g., public health. Republican Mayor Romero (San Luis Obispo, CA) stated that ’I
believe many communities, whether or not they believe that global warming is caused primarily
by man’s actions, will recognize that the actions proposed are good governance’. In a similar
vein, the deputy director of the same city, Mrs. Murry argued that ’even if it [climate change]
is a big hoax, this is still the right thing to do in terms of using less energy, making a better
community, making it more walkable, all of those things, it’s just good practice’. Although many
studies (Betsill 2001; Engel & Orbach 2008) have found that local co-benefits are an important
issue to frame climate change policies on the local level, the question remains why communities
would sign local climate change agreements in the first place if their primary reason is co-benefits.
In chapter 4, I have argued that larger cities are more active in local climate change policies
since they have a higher administrative capacity. However, another aspect of the size of a city
was mentioned by Mayor Loughmiller (McKinney, TX), who relates an emerging awareness
of green reasoning to population growth. McKinney has been the fastest growing city in the
U.S. for three consecutive years and has grown from 20,000 in 1990 to around 55,000 in 2000.
Latest estimates from the 2010 Census now see it at 126,000 inhabitants. With this dramatic
development in mind, he provided the following example from his city: ’Obviously McKinney
’Where the Rubber Meets the Road’
165
at 126,000 is different than when I moved here and it was only 19,000. You know, McKinney
at 19,000, people really, I don t think you would have gotten a lot of support to participate
in initiatives that were more nationally known or globally, having a global impact, because at
that point you are a small town and everything is more focused just on your own little corner
of the world. But as you grow and you become part, you know, we became part more of the
Dallas Metropolitan area complex and we’re more involved in regional issues, like cities’ national
groups, I think that s when you start to gain that awareness, so I think that’s probably a large
part of it. I haven’t seen the list but I would guess that it probably started out with the
larger cities and worked its way down to the smaller communities that were suburbs of these
cities and so on’. This last observation by Mayor Loughmiller actually fits my large-N data on
the adoption rates of different population groups in chapter 4. Whether this effect of size on
participation is manifest also in this sample, one can take a look at table 5.2.2. We can see that
the range of population sizes of cities that have signed goes from 12.527 until 331.285, whereas
those for cities that have not signed goes from 13.372 until 80.865. This in itself does not seem
to differ, especially for smaller cities, however, if we then exclude the cities that have especially
favorable conditions, such as Princeton (highly educated population, much information about
climate change, see the discussion below) and Gulfport (waterfront community, directly affected;
see discussion below), we can see that the there is a tendency as also found within the two other
papers, that size seems to matter for participation. A similar observation concerning city size
came from Larry Falkin (Cincinnati, OH): ’The larger cities are the ones that are more likely to
have professional staff involvement in national networks and awareness of the larger issues’. This
statement pertains to the nexus between the administrative capacity of a city and its awareness
of climate change policies but also hints to the importance of networks to spread information
concerning new policies or solutions to given problems.
Similar to the pattern in the Midwest questionnaires in chapter 4, a divide was observable
along partisan lines. Democratic mayors tend to interpret their signature as a critique of the
inaction of the federal government during the Bush Administration. For example, Mr. Falkin
from Cincinnati, OH, not having indicated a party affiliation, stated: ’I think a lot of it was the
federal government’s failure to exercise leadership. That you know really a lot of people thought
climate protection should have been an issue where the main leadership should have come from
Washington. And because during the Bush Administration the leadership was not coming from
there, it was a good opportunity for people at the local level to sort of step up and in a way show
166
their disagreement with the national policies by establishing local policies’. Interestingly, the
same logic may work in reverse for the Obama Administration’s effect on MCPA participation.
A public official from the South justified the non-signing of the MCPA as resistance to ’leftwing international socialist pressure through our federal government’. In his view, this reflects
a view frequently expressed by southern mayors, who still doubt the reality of climate change,
seeing it more as a ’hullaballoo just to make some special interest groups or friends of some
liberal politician rich at our expense’. In a similar vein, Mayor Mullen from Harker Heights,
TX stated that: ’The agreement you’re speaking of, i.e., for the political piece, much of that is
considered in Texas to be simply politics and propaganda and not accepted as a proof of science
really’. Following from this, most of the non-signing mayors do not ascribe any real change in
policy output to the signing of the agreement; instead they consider it as merely a symbolic
act against the former government. I have already hinted at the discussion on the symbolic
vs. substantive interest in signing the MCPA in the theoretical part of chapter 4. Mayors can
use their participation in climate change initiatives to position themselves against the federal
government or other cities’ in the area. With the climate change issue as a relatively new topic
and a high coverage by national media, it is an attractive agenda issue for local governments
(Krause 2011a). Media coverage increases the visibility and popularity of politicians not only
among their own constituencies, but could also signal leadership and entrepreneurial capabilities
for other offices that might be future career options for mayors. A representative from a large
town in California stated on this topic that ’[signing] is absolutely a key to admission to move
to a higher level politically. [. . . ]. [T]hey all have aspirations to go to the legislature or to some
state commission like the California Energy Commission; and that’s part of their credentials to
show how environmentally responsible they are’. Although the particular situation in California
differs from the rest of country, we generally cannot dismiss the notion that there may be a lot
of cities that did sign only to give a signal against the Bush Administration which brought them
support from their constituencies, but did probably not result in direct policy output. I will
briefly return to this discussion again below when talking about the initiative as a whole.
While these were the main reasons that were mentioned by a majority of the mayors as
influential for their decision (not) to join the MPCA, there were additional issues that came up
in the interviews and that are relevant to this research. Therefore, the following section looks
at each of these in turn.
’Where the Rubber Meets the Road’
167
5.3.3 Further Issues
In the previous two dissertation papers, it was stipulated – and supported by the statistical
analyses – that the level of education matters for MCPA participation. This finding is also
confirmed by mayors from university cities. Council member Matt Appelbaum (Boulder, CO)
admitted that for enacting local climate change policies, ’it helps to have a community on board,
Boulder is a University town, it’s a pretty liberal place. People get it’. Mayor Miller (Princeton
Township, NJ) further pointed out the role that education and the position of a community
has with respect to the reception and use of information about global warming: ’I think you
probably have to make adjustments for the fact that Princeton is a university town, we have a
very high percentage of residents in town that have undergrad and grad degrees. The university
is engaged in a very, very strong sustainability program and we see that all the time, we work
closely with the university in that area. And we are also in between two major metropolitan
areas, New York and Philadelphia, we are about half way in between of them and there is a lot
of information in the press you know from the major cities, and national press. Probably the
most read newspaper here - other than local newspapers- is the New York Times’.
Some support for the Nature argument, which I introduced in chapter 3.1, was provided
by city manager O’Reilley (Gulfport, FL), who believes that the main reason for the mayor’s
signature was that ’we’re on a waterfront community, so we’re very environmentally conscious’.
The same holds true for Mayor Gorin (Santa Rosa, CA), who gave a similar reason for signing:
’Especially we have a number of roads, sea port, trains and housing developments that could be
underwater in the next 50 years, so we are very concerned about that’.
I now turn to evidence mentioned concerning the effects of networks and peers. Whereas many
mayors (and increasingly those who signed) say that they either heard from the MCPA through
the Conference of Mayors, or that they were active members, meaning that they also regularly
attend conferences, only Mayor Wild of Westland, MI explicitly stressed the importance of the
meetings for his participation. He stated that he had a background in recycling, but then ’when
we were in Miami for the conference [the U.S. Conference of Mayors annual meeting 2008 in
Miami] about two years ago, I really had a chance to see a lot of the initiatives that the city of
Miami had implemented and I guess I was hooked at that point. I kind of saw what Miami was
doing and realized that we could do that in our own city on a smaller level’. This confirms the
importance of these network meetings as such but further hints to the peer effects.
168
The possibility of federal grant money for cities through the Energy Efficiency and Conservation Block Grant (EECBG) program was seen as a major incentive for cities to join the
agreement in the previous paper. However, the effect of the EECBG introduction turned out to
be statistically insignificant. An example from my sample is McKinney in Texas, that received 1
million $ out of this program. Mayor Loughmiller (McKinney, TX) links the success of obtaining
federal money to participation in the MCPA. He stated: ’I think, really, the impact of being a
participant in that is that a lot of the things cities do in this area are tied to grant opportunities
as you’re developing your city, you know, you have more grant opportunities if you are doing
development that is tied to environmentally friendly initiatives, so I think really the benefit of
being a participant probably is more in the nature of - if you’re applying for a federal grant you
are part of that pool of cities that have come on from a national standpoint and agreed to do
things that fall within this initiatives’.
The opinion that green policy entrepreneurship has an significant influence on people’s attitudes towards environmental politics is confirmed by Mr. Darling (Saint Peters, MO). He
states that the local climate change initiatives were mostly initiated by Mayor Len Pagano,
’Our mayor is awful green. Meaning that he believes in green activities, for instance he drives a
little moped-like motorcycle, he is very environmentally conscious’. He stresses the importance
for local government to create positive incentives with regard to local climate change policies
such as giving out rewards for people who recycle instead of penalizing those that fail to recycle. Concerning the reaction of the community to Mayor Pagano’s Green policies, he gives the
following image: ’it s like feeding a hungry bear - you feed em more food and they want more
food. You energize the people on incentives to recycling and now you have to take incentives to
the higher level, and the higher level and the highest level in order to achieve the same amount,
it s pretty greedy in that regard’. This underscores the importance that also the supply side of
local government can have given there is a policy entrepreneur.
Also, there is some indication that local industry has an influence on either facilitating or
hindering the participation of cities. This relates to the hypothesis on abatement costs for
local industry due to the city’s engagement in voluntary climate change policies. Mayor Adams
(Hanover, PA), who has not yet signed the MCPA, admitts that ’I would probably say that
there is an interest [for climate change policies] in the community. I would have to ask the local
industry though’. Freemont in Nebraska, a city that has not signed the MCPA, has made a
different experience with industry’s attitude towards local climate change policies. Mr. Hartwig
’Where the Rubber Meets the Road’
169
(Freemont, NE) sees an emerging demand for green energy from the business side: ’A lot of
businesses that come to the city these days are asking what we have in the way of renewable
power, and can they participate in renewable energy credits and things like that. [This] is
basically why we’ve been exploring the wind power and the solar power. We don’t want to do
anything that would cause a business to say no to Freemont’. In a similar vein, Mayor Gorin
(Santa Rosa, CA) sees an economic competitive advantage for cities if they engage in green
policies: ’The business side, I emphasize it makes sense economically. We have positioned to our
community and our county as this green hub and we are winning awards all over the place for
innovative programs. So this is the way that we are going to solve our economic development
challenges, by making sure that we nurture and recruit green businesses to come and locate in
our community’.
This concludes the discussion about the city and its main reasons to join and experiences with
the MCPA. I now turn to the question about the success of the agreement as a whole and what
mayors thought were the main reasons for the development of the MCPA.
5.3.4 Success of the Agreement as a Whole
The question concerning explanations for the success of the MCPA as a whole is answered
uniformly with reference to increased public and media attention to the topic of climate change.
However, depending on whether a locality has signed or not, the answers concerning the effect
of this agreement vary. Mayor Carson (Wisconsin Rapids), who is not an MCPA signatory,
observed that ’a lot of cities signed on because it was the thing to do. It was the subject matter
of many coffee klatsches [sic!]and many public meetings’. Republican Mayor Adams (Hanover,
PA) points to the link between the public discourse and its effects: ’It’s an outcry that people
are finally starting to care. Most definitely there has been a shift in people’s mindset’. Democrat
Mayor Wild (Westland, MI) confirms this observation and states that ’global warming now is, it’s
on TV, it’s in the newspapers, I mean it has become common, you know, the average household
now understands global warming at least to some level, so I would think that recycling and the
green agenda, it has become more mainstream now’. This change in people’s mindset lead some
respondents to the assumption, that mayors sign the agreement to increase their probability of
re-election. A Californian representative states that politicians assume that ’jumping on the
climate change bandwagon is a path to political prosperity’. This, he mentioned, holds not only
for direct public sentiment but also for financial support for campaigns by the ’well financed
170
and organized special interest group, this environmental activist group, that thinks very highly
of these types of initiatives and plans and if you become associated with them you receive some
benefits’.
In contrast to this, a very consistent finding for mayors that signed the agreement is that
most cities actually were engaged in climate-friendly policies before they joined the MCPA. Mr.
Darling’s (Saint Peters, MO) comment is representative for many other mayors that stated that
signing the agreement was just logical given their environmental record: ’I think the mayor
signed that because he was saying, you know, I believe in that, we’re already doing these things,
that assist in saving CO2 in the atmosphere [. . . ] so signing the agreement is essentially saying
it’s not that we’re doing that, meaning that we will do it, but we’ve done that and it makes
sense in that regard’. Mayor Koos (Normal, IL) expresses the same opinion: ’It was just about
ratifying, those things that we were trying to do anyway and to join a larger group with a
larger voice I thought was pretty important’. This leads him to the conclusion that ’the climate
protection agreement is probably as much a declaration as anything. You know it is unifying
of people making a statement of common cause and so I think it helps build awareness and I
think it makes it a little easier for me to do in my community’. Therefore it seems that many
mayors who signed the agreement take the global warming topic very seriously manifesting in
already implemented policies they mention. For them, joining the MPCA is therefore seen as an
essential step in fighting global climate change by joining a larger coalition that might have the
potential to change things. However, they are aware of the fact that joining the MCPA without
implementing green policies is just a symbolic gesture .
Mayors who signed the agreement refute this view by showing a reflected and distinct evaluation about the range and the practices of the agreement. Council member Matt Appelbaum
(Boulder, CO) hints at the usefulness of networks to learn about best-practices and about the
notion of cities of cities as policy laboratories and mentions that ’we are a test bed for certain
policies and programs, we’ll see how they work, other cities will pick up on some other things
and we’ll learn some other things. This is what you have to do, you can be more innovative
at the city level, because we are smaller. [. . . ]. And so, all of some technologies can transfer,
the mechanisms and the programs don’t necessarily transfer, and that’s even true from state
to state. Cause different states have different regulations, cities have different authorities, and
states have public utility commissions that really separate structure’. He further emphasizes the
limitation of an all-encompassing policy solution for all communities. In the opinion of Mayor
’Where the Rubber Meets the Road’
171
Wild (Westland, MI) these different instruments and policies share the common denominator
of being small and flexible: ’T]he one thing about it is that most of our green initiatives that
we’ve been implemented, they’re not that hard. There are a lot of small things that you can do
that when you add them together make a difference’.
To conclude, in Matt Appelbaum’s (Boulder, CO) opinion, the motivation of the mayors to
join the agreement does not matter as long as they join and expand the movement. He observes
that ’some people may come out of the climate change or environmental issue, some perhaps
come out of money issues, some perhaps come out of global political issues, but I think all of
those together finally reached enough of a mass that people get it’. Through the participation
of many mayors, the initiative gained and sustained momentum of public attention and helped
to bring about a change in people’s mindset. The question about what impact it had on the
federal government is dealt with in the following section.
5.3.5 Impact on the Federal Government
Concerning the impact the initiative had on the federal level, the population of mayors who
signed the agreement is in consent that the signature of the agreement sends a signal to the
federal government. For example, Mrs. Murry (San Luis Obispo, CA) sees the signaling ’in
terms of showing kind of a sea change or a grassroots movement of peoples believe. Having that
come from a local level was kind of important to say ’You blowing it, federal government. You
need to do something and you need to participate and you need to be willing to commit and
we need levels of measures on the federal level, because the local level can only do so much,
and so much of this comes from the federal level, or the state level”. Also Mayor Gorin (Santa
Rosa, CA) sees the movement from a bottom-up perspective with the cities as driving forces,
acknowledging that ’it’s really interesting that President Obama and so many others and even
our own Governor, our leaders have acknowledged that these kind of actions start at the local
level. And it is the mayors who are really moving forward. They get it, they understand the
impacts for their communities and they understand how important it is, both from an economic
development perspective as well as preparing our communities for whatever the future may hold
for us, that we take these actions. [. . . ] It’s worldwide, think globally act locally and our county
and our community truly believes this’.
The assessment of cities that have signed the agreement is – on this issue – backed up by the
peers that did not sign. A representative of a Californian city finds that the MCPA ’probably
172
has some effect on states and the national government to say there is more consensus on climate
change so it is time to pass comprehensive legislation, that we try this at the national level too’.
5.4 Conclusion
This last chapter served as a complement to the full-fledged analysis papers presented in the
previous chapters. The qualitative paper delivered additional evidence that on the one hand can
be taken as a validation of the former analysis and on the other might add meaningful insights
into the different motivational forces of mayors in signing or not signing the MCPA. Further
support could be gathered for some of the most influential structural factors for joining the
MPCA: The average level of education in a city, the city size in terms of population, as well as
its vulnerability to natural disasters clearly makes a difference for the likelihood to join. Mayors
from either a university city, a big city or an especially vulnerable area name these factors as
decisive influences to engage in policies against global warming and to sign the MCPA. The
partisanship of mayors has played a role for Democrats during the Bush Administration and
for Republicans to some extent since the election of President Obama. This might lend to the
assumption that generally signing and not signing the agreement has the function of showing the
dissatisfaction with the sitting administration. A new aspect that could be added to the overall
framework through the analysis of the interviews was the differentiated look on the evaluation of
the MCPA. Many of the mayors that have not signed view the whole endeavour as mere symbolic
politics which possibly has an effect on the general public mood and the relevant political career
but produces no significant climate change policy output. On the other hand, mayors who signed
the agreement look at the topic from another angle: a predominant part of the mayors who signed
started implementing green policies already before joining the initiative. They thereby indicate
that they care about the threat of climate change or are generally interested in environmental
politics. Their motivation to join the MPCA is twofold. First, they partly affirm the symbolic
component critisized by the non-signer, presenting it in another light. They state the opinion
that symbolic politics raise public awareness for the topic, which in turn increases public demand
for these policies. Second, mayors join the MCPA because they are aware of the network effects
that emerge in terms of exchanging information and experience with certain policy instruments.
’Where the Rubber Meets the Road’
173
5.5 Appendix
5.5.1 Email
Dear Mayor XX,
My name is Lena Schaffer, I am a Research assistant and 3rd year PhD student in Political
Economy at the Swiss Federal Institute of Technology (ETH) in Zurich, Switzerland. I am
writing my PhD thesis on voluntary climate change commitments across U.S. cities and states
with a special focus on the U.S. Mayors Climate Protection Agreement. I am contacting you
because I am interested in conducting a brief informal interview with you.
In order to improve confidence in my quantitative analysis, I am doing semi-structured research interviews with mayors and representatives of cities with over 10,000 inhabitants. I have
randomly selected 30 out of all U.S. cities. Therefore the participation of every city, regardless
of its stance on the Mayors Climate Protection Agreement, is very valuable for an in-depth
understanding of voluntary climate change commitments. By taking part in a brief informal
interview, which would take place over the course of the next 4 weeks, you would be providing
us with valuable feedback and information.
As far as my research is concerned, I am interested in how action and commitment by actors
at the local level spread across the United States and how one can explain the dynamics behind
this process. The main questions that I will be addressing are what makes local governments
voluntarily implement certain policies, and which internal factors lead to the adoption of climate
change mitigation initiatives.
I have attached a letter of consent and further information about the procedure. If you would
be willing to be interviewed, please let me know a date and time when I or a member of the
research group can reach you, how you would like the interview to be conducted, and whether
or not you would like to remain anonymous. One of us will be contacting you shortly, however
if you have questions we can be reached by phone at 202-559-0677, or via e-mail (climateresearch@ethz.ch).
Thank you very much in advance!
Best regards,
Lena Schaffer
174
5.5.2 Study Information Sheet
’Where the Rubber Meets the Road’
175
Center for Comparative and International Studies
(CIS)
Institute for Environmental Decisions (IED)
ETH Zürich
Weinbergstrasse 11, WEC C18
8092 Zürich
Switzerland
To whom it may concern
Lena Schaffer
Ph.D. Student
+41 44 632 0261
schaffer@ir.gess.ethz.ch
Information Sheet for the Research Project no. 20329 ETH-RDB !"#$%&'()*+ ,$-.('/+
Change Initiatives in the United States: Testing Spatial Dependence in Participation!
Lena Schaffer and Thomas Bernauer
You are being asked to participate in a research study about if and how cities and other sub
national units are involved in local policy measures with respect to global climate change. The
specific focus of the study is the U.S. Mayors Climate Protection Agreement. The goal is to
better understand the factors leading to participation or non-participation within this policy
initiative.
Participation in this study is voluntary. It will involve an interview of approximately 30-45 minutes
in length and will be conducted over the phone. With your permission, the interview will be audio
recorded to facilitate collection of information, and later transcribed for analysis. You will be
asked questions regarding three broad areas: personal and demographic questions, questions
about your city, and questions regarding your city's participation or non-participation in the
MCPA.
The only foreseeable discomfort associated with the study is the invasion of your privacy. You
will not be paid for participation in this research. There are no direct benefits from participation in
the study. However, this study may help explain innovations in local policy measures to address
global warming and how these measures are adopted and implemented at the municipal level.
You may refuse to participate or discontinue your involvement by advising the researcher at any
time without penalty. You may choose to skip a question as well as decline to be audio taped.
We will take notes or record your responses simply to ensure accuracy in our analyses. If you do
not wish to be recorded, or would like your answers to be anonymous, please inform the
researcher at the beginning of the interview. If you choose to be recorded, the tapes will be
erased at the completion of the study.
If you are a public official or have played a publicly visible role in your organization, your identity
may be discernible in the reporting of this research. If you would like details that identify you to
be omitted from reports developed from this interview data, please tell the researcher.
All research data collected will be stored securely and confidentially in a locked office at ETH
Zürich. Furthermore all material will be destroyed 12 months after the completion of the study. If
you have any questions regarding this study, or would like additional information to assist you in
reaching a decision about participation, please contact me at (202-559-0677) or via email at
climateresearch@ethz.ch or !"#$%&'($))"*+,*%-"&&%".(/%'(% You can also contact my supervisor,
Professor Thomas Bernauer at +41 44 632 64 66.
Seite 1/1
176
5.5.3 Interview Template
’Where the Rubber Meets the Road’
Questions for Interviews with Mayors/ City representatives
Lena Schaffer
1) Personal information
o
o
o
o
o
o
o
o
o
o
o
Age: How old are you?
What is your marital status?
Do you have children?
What is your educational background?
What is your profession?
In politics, do you consider yourself a Republican, a Democrat, an independent or of
some other party? (if other party, ask him/her to identify it)
In economics, do you consider yourself a liberal or a conservative?
How long have you been in politics?
How long / how many terms have you been a mayor of this city?
Are you or have you been an active member within the U.S. conference of mayors or
the National League of Cities or any other national or state level organization? If yes
in what capacity?
Do you have any aspirations for a future political career beyond your present one at
the county, state and/or national levels?
2) Information regarding your city
(Name of city) has around (take population from dataset) inhabitants, what is the
main sector of employment in your city?
o What is the main source of revenue in your city?
o
o Is there a university / college in your town? What type of university / college is it?
o Is your city vulnerable to natural disasters such as
hurricanes/floods/droughts/tornadoes or other natural disasters?
o if yes: Has your city suffered from any major natural disaster during the last 5 years or
since you became the mayor of the city?
o What are the three cities nearest to your city?
o With which of the cities that you just mentioned do you have the closest ties? why?
3) MCPA Climate Information
As you know, there exists an initiative called the Mayors Climate Protection Agreement
(MCPA). Is (/ or has) your city (been) a member of this initiative?
If more information is needed…
(Seattle Mayor Greg Nickels launched this initiative in early 2005 to advance the goals of the Kyoto Protocol
through leadership and action by American cities. After 141 cities had signed on to the agreement, it became an
initiative under the umbrella of the US conference of Mayors. And since then, more than 1000 cities now have
joined this agreement.
“Strive to meet or beat the Kyoto Protocol targets in their own communities, through actions ranging from antisprawl land-use policies to urban forest restoration projects to public information campaigns”)
177
178
If Yes:
How did you first hear about the initiative?
When did you join the MCPA (month / year)
If No:
How did you first hear about the initiative?
What were the main factors that influenced your
decision to sign the agreement?
PROBE: Any other reasons?
What were the main factors that influenced your
decision to not sign the agreement?
PROBE: Any other reasons?
Was your decision to sign the agreement
influenced/affected by any particular/specific
incentives offered by the initiative?
If yes, could you identify them?
Have you implemented any new practices / city
resolutions since joining the MCPA?
If yes, name 1-2 of the new practices / city
resolutions.
Do you know mayors around you that have joined
the initiative?
Other than you personally, what people or groups
within the community were involved in getting
the initiative started in your city?
Do you know mayors around you that have
joined the initiative?
Initiative as a whole:
In your opinion what were the most important
factors that made this agreement spread so quickly
during its first years?
In your opinion what were the most important
factors that made this agreement spread so
quickly during its first years?
In your opinion has the MCPA been successful in
effectively addressing the climate change issue at
the local level?
In your opinion has the MCPA been successful
in effectively addressing the climate change
issue at the local level?
In your opinion has (or could) the MCPA been
successful in affecting the federal government's
position on the climate change issue?
In your opinion has (or could) the MCPA been
successful in affecting the federal government's
position on the climate change issue?
Are you a member in any other networks of
mayors?
Lastly
Do you think that your decision to sign the
agreement has a positive, negative or no effect on
the public sentiment towards you presently?
Are you a member in other networks of mayors?
Do you think that your decision to sign the
agreement might have a positive, negative or no
effect on your prospects for reelection?
!
Do you think that your decision to not sign the
agreement has a positive, negative or no effect
on the public sentiment towards you
presently?
Do you think that your decision to not sign the
agreement might have a positive, negative or no
effect on your prospects for reelection?
179
Bibliography
Adger, W. N. (2010), ‘Climate Change, Human Well-Being and Insecurity’, New Political Economy 15(2), 2–16.
Aldy, J. (2005), ‘An Environmental Kuznets Curve Analysis of U.S. State-Level Carbon Dioxide
Emissions’, The Journal of Environment & Development 14(1), 48–72.
Allen, M. (2003), ‘Liability for Climate Change’, Nature 421(6926), 891–892.
Allen, M., Frame, D., Frieler, K., Hare, W., Huntingford, C., Jones, C., Knutti, R., Lowe, J.,
Meinshausen, M. & Meinshausen, N. (2009), ‘The Exit Strategy’, Nature Reports Climate
Change pp. 56–58.
Anselin, L. (1988), Spatial Econometrics Methods and Models, Kluver Academic Publishers,
Dordrecht etc.
Anselin, L. (2004), Advances in Spatial Econometrics Methodology. Tools and Applications,
Springer, Berlin.
Baettig, M. B. & Bernauer, T. (2009), ‘National Institutions and Global Public Goods. Are
Democracies More Cooperative in Climate Change Policy?’, International Organization 63(2).
Baettig, M. B., Wild, M. & Imboden, D. (2007), ‘A Climate Change Index: Where Climate
Change may be most prominent in the 21st Century’, Geophysical Research Letters 34, 1–6.
Barrett, S. (2003), Environment and Statecraft. The Strategy of Environmental Treaty-making,
Oxford University Press, Oxford.
Baumol, W. J. & Oates, W. E. (1988), The Theory of Environmental Policy, Cambridge University Press.
Baybeck, B. & Huckfeldt, R. (2002), ‘Urban Contexts, Spatially Dispersed Networks, and the
Diffusion of Political Information’, Political Geography 21(2), 195–220.
180
Beck, N., Gleditsch, K. S. & Beardsley, K. (2006), ‘Space is More than Geography: Using Spatial
Econometrics in the Study of Political Economy’, International Studies Quarterly 50(1), 27–
44.
Beck, N., Katz, J. & Tucker, R. (1998), ‘Taking Time seriously: Time-series-cross-section Analysis with a binary Dependent Variable’, American Journal of Political Science 42(4), 1260–
1288.
Bernauer, T., Kalbhenn, A., Koubi, V. & Spilker, G. (2010), ‘A Comparison of International
and Domestic Sources of Global Governance Dynamics’, British Journal of Political Science
40(03), 509–538.
Bernauer, T. & Koubi, V. (2009), ‘Effects of Political Institutions on Air Quality’, Ecological
Economics 68(5), 1355–1365.
Bernauer, T., Koubi, V., Kalbhenn, A. & Ruoff, G. (2010), Climate Change, Economic Growth,
and Conflict, in ‘Climate Change and Security’, Trondheim, Norway.
Bernauer, T. & Schaffer, L. M. (2010), ‘Climate Change Governance’, CIS Working Paper
60, 1–29.
Beron, K. J., Murdoch, J. C. & Vijverberg, W. P. M. (2003), ‘Why Cooperate? Public Goods,
Economic Power, and the Montreal Protocol’, Review of Economics and Statistics 85(2), 286–
297.
Berry, F. & Berry, W. (1999), Innovation and Diffusion Models in Policy Research, in P. A.
Sabatier, ed., ‘Theories of the Policy Process’, Westview Press, pp. 169–200.
Berry, F. S. & Berry, W. D. (1990), ‘State Lottery Adoptions as Policy Innovations: An Event
History Analysis’, The American Political Science Review 84(2), 395–415.
Berry, W. D. & Baybeck, B. (2005), ‘Using Geographic Information Systems to Study Interstate
Competition’, American Political Science Review 99(04), 505–519.
Betsill, M. (2000), ‘Localizing Global Climate Change: Controlling Greenhouse Gas Emissions
in U.S. Cities’, Belfer Center for Science and International Affairs (BCSIA) Discussion Paper
2000-20, 1–52.
181
Betsill, M. & Bulkeley, H. (2006), ‘Cities and the Multilevel Governance of Global Climate
Change’, Global Governance 12(2), 141–159.
Betsill, M. M. (2001), ‘Mitigating Climate Change in US Cities: Opportunities and Obstacles’,
Local Environment 6(4), 393–406.
Betsill, M. M. & Bulkeley, H. (2004), ‘Transnational Networks and Global Environmental
Governance: The Cities for Climate Protection Program’, International Studies Quarterly
48(2), 471–493.
Boehmke, F. J. & Witmer, R. (2004), ‘Disentangling Diffusion: The Effects of Social Learning
and Economic Competition on State Policy Innovation and Expansion’, Political Research
Quarterly 57(1), 39–51.
Boehringer, C. & Helm, C. (2008), ‘On the Fair Division of Greenhouse Gas Abatement Cost’,
Resource and Energy Economics 30(2), 260–276.
Boehringer, C. & Jochem, P. (2007), ‘Measuring the immeasurable. A Survey of sustainability
Indices’, Ecological Economics .
Box-Steffensmeier, J. & Jones, B. (2004), Event History Modeling: A Guide for Social Scientists,
Cambridge University Press.
Boykoff, M. (2007), ‘Flogging a Dead Norm? Newspaper Coverage of anthropogenic Climate
Change in the United States and United Kingdom from 2003 to 2006’, Area 39(4), 470–481.
Braun, D. & Gilardi, F. (2006), ‘Taking Galton’s Problem Seriously. Towards a Theory of Policy
Diffusion’, Journal of Theoretical Politics 18(3), 298.
Brody, S., Zahran, S., Grove, H. & Arnold, V. (2008), ‘A Spatial Analysis of Local Climate
Change Policy in the United States: Risk, Stress and Opportunity’, Landscape and Urban
Planning 87, 33–41.
Brueckner, J. K. (2003), ‘Strategic Interaction Among Governments: An Overview of Empirical
Studies’, International Regional Science Review 26(2), 175–188.
Buchinsky, M. (1998), ‘Recent Advances in Quantile Regression Models. A Practical Guideline
for Empirical Research’, Journal of Human Resources 33(1), 88–126.
182
Buhaug, H., Gleditsch, N. P. & Theisen, O. M. (2008), Implications of Climate Change for
Armed Conflict, in ‘Social Dimensions of Climate Change’, The World Bank, Washington,
D.C., pp. 5–6. World Bank Workshop on the Social Dimensions of Climate Change, The
World Bank, Washington.
Burke, M. B., Miguel, E., Satyanath, S., Dykema, J. A. & Lobell, D. B. (2009), ‘Warming
increases the Risk of Civil War in Africa’, Proceedings of the National Academy of Sciences
106(49), 20670–74.
Buttel, F. H. (1979), ‘Age and Environmental Concern: A Multivariate Analysis’, Youth and
Society 10(3), 237–56.
Carter, D. & Signorino, C. (2010), ‘Back to the future: Modeling Time Dependence in Binary
Data’, Political Analysis .
Climate Analysis Indicators Tool (CAIT US) version 4.0. (2010).
Cline, W. R. (1999), Discounting for the very Long Term, in P. R. Portney & J. P. Weyant, eds,
‘Discounting and intergenerational equity’, Resources for the Future, Washington, pp. 131–
140.
Collier, P. (2010), The Plundered Planet: Why We Must and How We Can Manage Nature for
Global Prosperity, Oxford University Press, Oxford.
Congleton, R. D. (1992), ‘Political Institutions and Pollution Control’, The Review of Economics
and Statistics 74(3), 412–421.
Cornes, R. & Sandler, T. (1996), The Theory of Externalities, Public Goods, and Club Goods,
second edn, Cambridge University Press, Cambridge.
Daley, D. M. & Garand, J. C. (2005), ‘Horizontal Diffusion, Vertical Diffusion, and Internal Pressure in State Environmental Policymaking, 1989-1998’, American Politics Research
33(5), 615–644.
Dasgupta, S., Laplante, B., Wang, H. & Wheeler, D. (2002), ‘Confronting the Environmental
Kuznets Curve’, Journal of Economic Perspectives 16(1), 147–168.
Davis, M. L. & Hayes, K. (1993), ‘The Demand for Good Government’, The Review of Economics
and Statistics 75(1), 148–152.
183
de Bruyn, S. M. & Heintz, R. J. (1998), The Environmental Kuznets Curve Hypothesis, in
‘Handbook of Environmental Economics’, Blackwell Publishing, Oxford.
Dell, M., Jones, B. F. & Olken, B. A. (2008), ‘Climate Shocks and Economic Growth: Evidence
from the last half Century’, NBER Working Paper 14132.
Dellink, R., Elzen, M., Aiking, H., Bergsma, E., Berkhout, F., Dekker, T. & Gupta, J. (2009),
‘Sharing the Burden of Financing Adaptation to Climate Change’, Global Environmental
Change 19(2009), 411–21.
Den Elzen, M. G. J., Schaeffer, M. & Lucas, P. L. (2005), ‘Differentiating Future Commitments
on the Basis of Countries’ relative historical Responsibility for Climate Change. Uncertainties
in the ’Brazilian Proposal’ in the Context of a Policy Implementation’, Climatic Change
71(3), 277–301.
Diekmann, A. (2004), Empirische Sozialforschung, Vol. 11, Rowohlt.
Dinda, S. (2004), ‘Environmental Kuznets Curve Hypothesis. A Survey’, Ecological Economics
49(4), 431–455.
Dunlap, R. E. & Mertig, A. G. (1992), American Environmentalism. The US Environmental
Movement, 1970-1990, Crane Russak & Co.
Economist (2010), ‘Let it be - The Democrats abondon their Efforts to limit Emissions through
legislation’, Economist .
Engel, K. (2009), ‘Whither Subnational Climate Change Initiatives in the Wake of Federal
Climate Legislation?’, Publius: The Journal of Federalism 39(3), 432.
Engel, K. & Orbach, B. Y. (2008), ‘Micro-Motives and State and Local Climate Change Initiatives’, Harvard Law & Policy Review 2, 119–137.
Feiock, R. C., Jeong, M. G. & Kim, J. (2003), ‘Credible commitment and Council-Manager
Government. Implications for Policy Instrument Choices’, Public Administration Review
63(5), 616–625.
Fisher, D. R. (2006), ‘Bringing the Material back in. Understanding the U.S. Position on Climate
Change’, Sociological Forum 21(3), 467–494.
184
Franzese, R. J. & Hays, J. C. (2006), ‘Strategic Interaction among EU Governments in Active Labor Market Policy-making: Subsidiarity and Policy Coordination under the European
Employment Strategy’, European Union Politics 7(2), 167–189.
Franzese, R. J. & Hays, J. C. (2007), ‘Spatial Econometric Models of Cross-Sectional Interdependence in Political Science Panel and Time-Series-Cross-Section Data’, Political Analysis
15(2), 140–164.
Franzese, R. J. & Hays, J. C. (2008a), Empirical Models of Spatial Interdependence, in ‘Oxford
Handbook of Political Methodology,’, Oxford University Press, Oxford.
Franzese, R. J. & Hays, J. C. (2008b), ‘Interdependence in Comparative Politics: Substance,
Theory, Empirics, Substance’, Comparative Political Studies 41(4-5), 742–780.
Fredriksson, P. G. & Millimet, D. L. (2002), ‘Strategic Interaction and the Determination of
Environmental Policy across US States’, Journal of Urban Economics 51(1), 101–122.
Fredriksson, P. G. & Millimet, D. L. (2004a), ‘Comparative Politics and Environmental Taxation’, Journal of Environmental Economics and Management 48(1), 705–722.
Fredriksson, P. G. & Millimet, D. L. (2004b), ‘Electoral Rules and Environmental Policy’, Economics Letters 84(2), 237–244.
Freudenberg, M. (2003), ‘Composite Indicators of Country Performance: A Critical Assessment’,
OECD Science, Technology and Industry Working Papers 16, 1–34.
Fuessel, H. M. (2010), ‘Modeling Impacts and Adaptation in Global IAMs’, Wiley Interdisciplinary Reviews: Climate Change 1(March/April), 288–303.
Galeotti, M., Lanza, A. & Pauli, F. (2006), ‘Reassessing the Environmental Kuznets Curve for
CO2 Emissions: A Robustness Exercise’, Ecological Economics 57(1), 152–163.
Germanwatch (2010), ‘Klimaschutzindex 2010’, accessed at www.germanwatch.org/ksi .
Gilardi, F. (2010), ‘Who Learns from What in Policy Diffusion Processes?’, American Journal
of Political Science 54(3), 650–666.
Glaeser, E. & Kahn, M. (2010), ‘The Greenness of Cities: Carbon Dioxide Emissions and Urban
Development’, Journal of Urban Economics 67(3), 404–418.
185
Gleditsch, K. S. (2002), All International Politics is Local. The Diffusion of Conflict, Integration,
and Democratization, University of Michigan Press, Ann Arbor.
Gleditsch, K. S. & Ward, M. D. (2001), ‘Measuring Space: A Minimum-Distance Database and
Applications to International Studies’, Journal of Peace Research 38(6), 739–758.
Gore, C. (2010), ‘The Limits and Opportunities of Networks: Municipalities and Canadian
Climate Change Policy’, Review of Policy Research 27(1), 27–46.
Gray, V. (1973), ‘Innovation in the States: A Diffusion Study’, American Political Science
Review 67(4), 1174–1185.
Greene, W. H. (2003), Econometric Analysis, fifth edn, Prentice Hall, Upper Saddle River.
Grossman, G. M. & Krueger, A. B. (1995), ‘Economic Growth and the Environment’, The
Quarterly Journal of Economics .
Gurney, K. R., Mendoza, D. L., Zhou, Y., Fischer, M. L., Miller, C. C., Geethakumar, S. &
de la Rue du Can, S. (2009), ‘High Resolution Fossil Fuel Combustion CO2 Emission Fluxes
for the United States’, Environmental Science & Technology 43(14), 5535–5541.
Hainmueller, J. & Hiscox, M. (2006), ‘Learning to Love Globalization: Education and Individual
Attitudes toward International Trade’, International Organization 60(02), 469–498.
Hao, L. & Naiman, D. Q. (2007), Quantile Regression, Sage Publications, Inc.
Hardin, G. (1968), ‘The Tragedy of the Commons’, Science 162, 1243–1248.
Hayhoe, K., VanDorn, J., Wuebbles, D., Cherkauer, K., Schlegal, N. & Vavrus, S. (2009),
‘Regional Climate Change Projections for Chicago and the Great Lakes’, Journal of Great
Lakes Research .
Hays, J. C. (2009), Globalization and the New Politics of Embedded Liberalism, Oxford University
Press, Oxford.
Hill, R. J. & Magnani, E. (2002), ‘An Exploration of the Conceptual and Empirical Basis of the
Environmental Kuznets Curve’, Australian Economic Papers 41(2), 239–254.
Holtz-Eakin, D. & Selden, T. (1995), ‘Stoking the Fires? CO2 Emissions and Economic Growth’,
Journal of Public Economics 57(1), 85–101.
186
Hornback, K. E. (1974), ‘Orbits of Opinion: The Role of Age in the Environmental Movement’s
Attentive Public’, Unpublished PhD dissertation, Department of Sociology, Michigan State .
Huang, W., Lee, G. & Wu, C. (2008), ‘GHG Emissions, GDP Growth and the Kyoto Protocol:
A Revisit of Environmental Kuznets Curve Hypothesis’, Energy Policy 36(1), 239–247.
ICPSR (1984), ‘Higher education general information survey (hegis) xviii: Institutional characteristics of colleges and universities,1983-1984’.
ICPSR (2008), ‘County characteristics, 2000-2007 [united states]’.
Institute, Hazards & Vulnerability Research (2009), ‘The spatial hazard events and losses
database for the united states, version 7.0 [online database].’.
IPCC (2007a), Climate Change 2007: The Physical Science Basis; Summary for Policymakers,
IPCC, Geneva.
IPCC (2007b), Summary for policymakers, in M. Parry, O. Canziani, J. Palutikof, P. van der
Linden & C. Hanson, eds, ‘Climate Change 2007: Impacts, Adaption and Vulnerability. Contribution of Working Group II to the Forth Assessment Report of the Intergovernemental
Panel on Climate Change’, Cambridge University Press, Cambridge, pp. 7–22.
Jacobson, M. Z. (2010), ‘Enhancement of Local Air Pollution by Urban CO2 Domes’, Environmental Science & Technology 44(7), 2497–2502.
Jaenicke, M. (2005), ‘Trend Setters in Environmental Policy: The Character and Role of Pioneer
Countries’, European Environment 15(2), 129–142.
Jahn, D. (2006), ‘Globalization as the Missing Link in the Analysis of Diffusion Patterns in
Welfare State Development’, International Organization 60(02), 401–431.
Johnson, S. (2009), ‘Climate Change and Global Justice: Crafting Fair Solutions for Nations
and Peoples’, Harvard Environmental Law Review 33, 297–301.
Kafalenos, R. S., Leonard, K. J., Beagan, D. M., Burkett, V. R., Keim, B. D., Meyers, A., Hunt,
D. T., Hyman, R. C., Maynard, M. K. & Fritsche, B. (2008), What are the Implications of
Climate Change and Variability for Gulf Coast Transportation, in M. Savonis, V. Burkett &
J. Potter, eds, ‘Impacts of Climate Change and Variability on Transportation Systems and
187
Infrastructure: Gulf Coast Study, Phase I’, U.S. Department of Transportation, Washington,
pp. 4–1–4F–27 (104).
Kahn, M. (2006), Green Cities: Urban Growth and the Environment, Cambridge University
Press.
Kamal-Chaoui, L. & Robert, A. (2009), ‘Competitive Cities and Climate Change’, OECD Regional Development Working Papers .
Karl, T. R., Melillo, J. M. & Peterson, T. C. (2009), Global Climate Change Impacts in the
United States, Cambridge University Press, Cambridge.
Kellstedt, P., Zahran, S. & Vedlitz, A. (2008), ‘Personal Efficacy, the Information Environment, and Attitudes toward Global Warming and Climate Change in the United States’, Risk
Analysis 28(1), 113–126.
Kern, K. & Gotelind, A. (2009), Governing Climate Change in Cities: Modes of Urban Climate
Governance in Multi-level Systems, in ‘ed.) OECD, OECD Conference Proceedings, Paris,
pp. 171’.
Klinsky, S. & Dowlatabadi, H. (2009), ‘Conceptualizations of Justice in Climate Policy’, Climate
Policy 9(1), 88–108.
Kluger, J. (2006), ‘Global Warming Heats Up’, Time Magazine .
Knuth, S., Nagle, B., Steuer, C. & Yarnal, B. (2007), ‘Universities and Climate Change Mitigation: Advancing Grassroots Climate Policy in the US’, Local Environment 12(5), 485–504.
Koenker, R. & Bassett, G. (1982), ‘Robust Tests for Heteroscedasticity Based on Regression
Quantiles’, Econometrica 50(1).
Kolstad, C. (1996), ‘Learning and Stock Effects in Environmental Regulation: The Case of
Greenhouse Gas Emissions’, Journal of Environmental Economics and Management 31(1), 1–
18.
Konisky, D. M. (2007), ‘Regulatory Competition and Environmental Enforcement: Is There a
Race to the Bottom?’, American Journal of Political Science 51(4), 853–872.
Kousky, C. & Schneider, S. H. (2003), ‘Global Climate Policy: Will Cities Lead the Way?’,
Climate Policy 3(4), 359–372.
188
Krause, R. (2011a), ‘Policy Innovation, Intergovernmental Relations, and the Adoption of Climate Protection Initiatives by US Cities’, Journal of Urban Affairs 30(1), 45–60.
Krause, R. (2011b), ‘Symbolic or Substantive Policy? Measuring the Extent of Local Commitment to Climate Protection’, Environment and Planning C: Government and Policy 29(1), 46–
62.
Kwon, M., Berry, F. & Feiock, R. (2009), ‘Understanding the Adoption and Timing of Economic
Development Strategies in US Cities using Innovation and Institutional Analysis’, Journal of
Public Administration Research and Theory .
Lee, T. (2009a), Global City and Climate Change Networks, in ‘Annual Meeting of the International Studies Association (ISA)’, New York.
Lee, T. (2009b), ‘Why Do Cities Participate (or not) in the Global Climate Change Network(s)?’,
Paper presented at the ISA’s 50th Annual Convention, New York .
Leitner, H. & Sheppard, E. (2002), ‘”The City is Dead, Long Live the Net”: Harnessing European
Interurban Networks for a Neoliberal Agenda’, Antipode 34(3), 495–518.
Lempert, R., Scheffran, J. & Sprinz, D. (2009), ‘Methods for Long-Term Environmental Policy
Challenges PolicyChallenges’, Global Environmental Politics 9(3), 106–133.
Levinson, A. (2003), ‘Environmental Regulatory Competition: A Status Report and Some New
Evident’, National Tax Journal 56, 91–106.
Levy, J. T. (2007), ‘Federalism, Liberalism, and the Separation of Loyalties’, American Political
Science Review 101(03), 459–477.
Li, Q. & Reuveny, R. (2006), ‘Democracy and Environmental Degradation’, International Studies
Quarterly 50, 935–956.
Lindseth, G. (2004), ‘The Cities for Climate Protection Campaign (CCPC) and the Framing of
Local Climate Policy’, Local Environment 9(4), 325–336.
List, J. A., Millimet, D. L., Fredriksson, P. G. & McHone, W. W. (2003), ‘Effects of Environmental Regulations on Manufacturing Plant Births: Evidence from a Propensity Score Matching
Estimator’, Review of Economics and Statistics 85(4), 944–952.
189
Long, J. S. (1997), Regression Models for Categorical and Limited Dependent Variables, Sage
Publications, Inc.
Lubell, M., Feiock, R. & De La Cruz, E. (2009), ‘Local Institutions and the Politics of Urban
Growth’, American Journal of Political Science 53(3), 649–665.
Lutsey, N. & Sperling, D. (2008), ‘America’s Bottom-up Climate Change Mitigation Policy’,
Energy Policy 36(2), 673–685.
Maddala, G. S. (2001), Introduction to Econometrics, 3. edn, Wiley, Chichester.
Manski, C. F. (1993), ‘Identification of Endogenous Social Effects: The Reflection Problem’,
The Review of Economic Studies 60(3), 531–542.
Mayors
Climate
Protection
Center
(2007),
‘Block
Grant
Legislative
Updates’,
http://www.usmayors.org/climateprotection/EEBG.htm .
Mendelsohn, R. (2001), Global Warming and the American Economy: A Regional Assessment
of Climate Change Impacts, Edward Elgar, Cheltenham, UK.
Meyer, S. M. (1995), ‘The Economic Impact of Environmental Regulation’, Journal of Environmental Law and Practise 3, 4–15.
Meyer, S. M. & Konisky, D. M. (2007), ‘Adopting Local Environmental Institutions: Environmental Need and Economic Constraints’, Political Research Quarterly 60(1), 3–16.
Midlarsky, M. I. (1998), ‘Democracy and the Environment: An Empirical Assessment’, Journal
of Peace Research 35(3), 341–361.
Miguel, E., Satyanath, S. & Sergenti, E. (2004), ‘Economic Shocks and Civil Conflict: An
Instrumental Variables Approach’, Journal of Political Economy 112(4), 725–753.
Millimet, D. L. (2003), ‘Assessing the Empirical Impact of Environmental Federalism’, Journal
of Regional Science 43(4), 711–733.
Millimet, D. L., Fredriksson, P. G. & List, J. A. (2002), Strategic Competition in Environmental and Fiscal Policies: Theory and Evidence from the United States, Kluwer Academic
Publishers.
190
Mintrom, M. (1997), ‘Policy Entrepreneurs and the Diffusion of Innovation’, American Journal
of Political Science 41(3), 738–770.
Mitchell, R. B. (2006), ‘Problem Structure, Institutional Design, and the Relative Effectiveness
of International Environmental Agreements’, Global Environmental Politics 6(3), 72–89.
Molotch, H. (1976), ‘The City as a Growth Machine: Toward a Political Economy of Place’,
American Journal of Sociology 82(2), 309.
Mooney, C. & Lee, M. (1995), ‘Legislative Morality in the American States: The Case of Pre-Roe
Abortion Regulation Reform’, American Journal of Political Science 39(3), 599–627.
Mueller, D. C. (2003), Public Choice III, Cambridge University Press, Cambridge.
National Atlas of the United States (2004), ‘Cities and towns of the united states: National
atlas of the united states’.
Neumayer, E. (2002a), ‘Can Natural Factors Explain any Cross-Country Differences in Carbon
Dioxide Emissions?’, Energy Policy 30(1), 7–12.
Neumayer, E. (2002b), ‘Do Democracies Exhibit Stronger International Environmental Commitment? A Cross-country Analysis’, Journal of Peace Research 39(2), 139–164.
Nicholson-Crotty, S. (2009), ‘The Politics of Diffusion: Public Policy in the American States’,
The Journal of Politics 71(01), 192–205.
Nordhaus, W. (2010), ‘Economic Aspects of Global Warming in a Post-Copenhagen Environment’, Proceedings of the National Academy of Sciences 107(26), 11721.
Oates, W. E. (1998), The Economics of Fiscal Federalism and Local Finance, Elgar, Cheltenham,
UK.
Oates, W. E. (2002), A Reconsideration of Environmental Federalism, in M. Kunce, S. Gerking,
W. Morgan, J. A. List & A. de Zeeuw, eds, ‘Recent Advances in Environmental Economics’,
Edward Elgar, Cheltenham, pp. 1–32.
Oates, W. E. & Schwab, R. M. (1988), ‘Economic Competition among Jurisdictions: Efficiency
Enhancing or Distortion Inducing?’, Journal of Public Economics 35, 333–354.
191
O’Brien, R. M. (2007), ‘A Caution Regarding Rules of Thumb for Variance Inflation Factors’,
Quality and Quantity 41(5), 673–690.
O’Connor, R. E., Bord, R. J., Yarnal, B. & Wiefek, N. (2002), ‘Who Wants to Reduce Greenhouse Gas Emissions?’, Social Science Quarterly 83(1), 1–17.
Olsen, M. (1971), Logic of Collective Action: Public Goods and the Theory of Groups, Harvard
University Press.
Olson, M. (1965), The Logic of Collective Action, Cambridge University Press.
Ostrom, E., Gardner, R. & Walker, J. (1994), Rules, Games, and Common-Pool Resources,
University of Michigan Press, Ann Arbor.
Oye, K. & Maxwell, J. (1995), ‘Self-Interest and Environmental Management’, Journal of Theoretical Politics 6(4), 593.
Pagoulatos, A., Goetz, S. J., Debertin, D. L. & Johannson, T. (2004), ‘Interactions Between Economic Growth and Environmental Quality in U.S. Counties’, Growth and Change 35(1), 90–
108.
Paterson, M. (2001), Principles of Justice in the Context of Global Climate Change, in U. Luterbacher & D. F. Sprinz, eds, ‘International Relations and Global Climate Change’, MIT Press,
Cambridge, MA, pp. 119–126.
Perrin, S. & Bernauer, T. (2010), ‘International Regime Formation Revisited: Explaining Ratification Behaviour with Respect to Long-Range Transboundary Air Pollution Agreements in
Europe’, European Union Politics 11(3), 405.
Pew
Center
(2010),
http://www.pewclimate.org/what s being done/in the states/
state action maps.cfm.
Pluemper, T. & Neumayer, E. (2010), ‘Model Specification in the Analysis of Spatial Dependence’, European Journal of Political Research 49(3), 418–442.
Posner, E. A. & Sunstein, C. R. (2007), ‘Climate Change Justice’, Georgetown Law Journal
96, 1565–1612.
Potoski, M. & Woods, N. D. (2002), ‘Air Pollution Regulation in the United States’, Policy
Studies Journal 30(2), 208–226.
192
Rabe, B. (2004), Statehouse and Greenhouse: The Emerging Politics of American Climate
Change Policy, Brookings Institution Press.
Rabe, B. (2008), ‘States on Steroids: The Intergovernmental Odyssey of American Climate
Policy’, Review of Policy Research 25(2), 105–128.
Rappaport, J. & Sachs, J. D. (2003), ‘The United States as a Coastal Nation’, Journal of
Economic Growth 8(1), 5–46.
Rincke, J. (2005), Policy Innovation in Local Jurisdictions: Testing the Neighborhood Influence
Against the Free Riding Hypothesis, Zentrum fuer Europ. Wirtschaftsforschung.
Ringquist, E. J. (1994), ‘Policy Influence and Policy Responsiveness in State Pollution Control’,
Policy Studies Journal 22(1), 25–43.
Rogelj, J., Nabel, J., Chen, C., Hare, W., Markmann, K., Meinshausen, M., Schaeffer,
M., Macey, K. & Haehne, N. (2010), ‘Copenhagen Accord Pledges are Paltry’, Nature
464(7292), 1126–1128.
Rogers, E. (1995), Diffusion of Innovations, Free Press, New York.
Rouphail, N., Coelho, M. & Farias, T. (2006), ‘Effect of Roundabout Operations on Pollutant
Emissions’.
Sampei, Y. & Aoyagi-Usui, M. (2009), ‘Mass-media Coverage, its Influence on Public Awareness of Climate-change Issues, and Implications for Japan’s National Campaign to Reduce
Greenhouse Gas’, Global Environmental Change .
Sandler, T. (1997), Global Challenges an Approach to Environmental, Political, and Economic
Problems, Cambridge University Press, Cambridge etc.
Sattler, T. & Bernauer, T. (2011), ‘Gravitation or Discrimination? Determinants of Litigation
in the World Trade Organisation’, European Journal of Political Research 50(2), 143–167.
Schaffer, L. M. (2010), ‘Nature, Nurture, Or Neighbors? Testing Participation Patterns in
Voluntary Initiatives in U.S. Counties’, Paper presented at the Annual Meeting of the Swiss
Political Science Association, Geneva .
Schnell, R., Hill, P. & Esser, E. (2008), Methoden der empirischen Sozialforschung, Oldenbourg
Wissenschaftsverlag.
193
Schreurs, M. A. (2008), ‘From the Bottom Up: Local and Subnational Climate Change Politics’,
The Journal of Environment & Development 17(4), 343.
Schreurs, M. A. & Epstein, M. (2007), ‘Environmental Federalism in the United States and the
Swing of the Pendulum from the Federal to the State Level’, AICGS Policy Report 31, 33–52.
Schwartz, P. & Randall, D. (2003), ‘An Abrupt Climate Change Scenario and its Implications
for United States National Security’, Prepared for the Department of Defense by the Global
Business Network .
Selin, H. & VanDeveer, S. (2007), ‘Political Science and Prediction: What’s Next for U.S.
Climate Change Policy?’, Review of Policy Research 24(1), 1–27.
Shipan, C. R. & Volden, C. (2006), ‘Bottom-Up Federalism: The Diffusion of Antismoking
Policies from US Cities to States’, American Journal of Political Science 50(4), 825–843.
Shipan, C. R. & Volden, C. (2008), ‘The Mechanisms of Policy Diffusion’, American Journal of
Political Science 52(4), 840–857.
Simmons, B. A. & Elkins, Z. (2004), ‘The Globalization of Liberalization: Policy Diffusion in
the International Political Economy’, American Political Science Review 98(01), 171–189.
Simmons, B., Dobbin, F. & Garrett, G. (2008), The Global Diffusion of Markets and Democracy,
Cambridge University Press.
Singh, R., Murty, H., Gupta, S. & Dikshit, A. (2009), ‘An Overview of Sustainability Assessment
Methodologies’, Ecological indicators .
Sprinz, D. & Vaahtoranta, T. (1994), ‘The Interest-based Explanation of International Environmental Policy’, International Organization 48(01), 77–105.
Stavins, R. N. (2003), Experience with Market-based Environmental Policy Instruments, in K.G. Maeler & J. R. Vincent, eds, ‘Handbook of Environmental Economics’, Vol. 1, Elsevier,
Amsterdam, pp. 355–435.
Stern, N. (2008), ‘The Economics of Climate Change’, American Economic Review 98(2), 1–37.
Stern, N., Peters, S., Bakhshi, V., Bowen, A., Cameron, C., Catovsky, S., Crane, D., Cruickshank, S. & Dietz, S. (2006), ‘Stern Review: The Economics of Climate Change’.
194
Tang, Z., Brody, S. D., Quinn, C., Chang, L. & Wei, T. (2010), ‘Moving from Agenda to Action:
Evaluating Local Climate Change Action Plans’, Journal of Environmental Planning and
Management 53(1), 41–62.
Theisen, O., Holtermann, H. & Buhaug, H. (2010), ‘Drought, Political Exclusion, and Civil
War’, Paper presented at International Studies Association Annual Convention, New Orleans
.
Theodori, G. L. & Luloff, A. E. (2002), ‘Position on Environmental Issues and Engagement in
Proenvironmental Behaviors’, Society & Natural Resources 15(6), 471–482.
Tobler, W. R. (1970), ‘A Computer Movie Simulating Urban Growth in the Detroit Region’,
Economic geography 46, 234–240.
Tol, R. S. J. & Yohe, G. W. (2006), ‘A Review of the Stern Review’, World Economics 7(4), 233–
250.
Tomz, M., Wittenberg, J. & King, G. (2003), ‘CLARIFY: Software for interpreting and presenting statistical results’.
Urpelainen, J. (2009), ‘Explaining the Schwarzenegger Phenomenon: Local Frontrunners in
Climate Policy’, Global Environmental Politics 9(3), 1–24.
Vanderheiden, S. (2008), Atmospheric Justice: A Political Theory of Climate Change, Oxford
University Press, Oxford.
Vasi, I. (2006), ‘Organizational Environments, Framing Processes, and the Diffusion of the
Program to Address Global Climate Change Among Local Governments in the United States’,
Sociological Forum 21, 439–466.
Volden, C. (2006), ‘States as Policy Laboratories: Emulating Success in the Childrens Health
Insurance Program’, American Journal of Political Science 50, 294–312.
Volden, C. (2007), ‘Intergovernmental Grants: A Formal Model of Interrelated National and
Subnational Political Decisions’, Publius: The Journal of Federalism 37(2), 209.
Von Stein, J. (2008), ‘The International Law and Politics of Climate Change: Ratification
of the United Nations Framework Convention and the Kyoto Protocol’, Journal of Conflict
Resolution 52(2), 243.
195
Walker, J. L. (1969), ‘The Diffusion of Innovations among the American States’, American
Political Science Review 63(3), 880–899.
Walsh, B. (2008), ‘Wind Shift’, Time Magazine .
Ward, H. (2008), ‘Liberal Democracy and Sustainability’, Environmental Politics 17(3), 386–
409.
Ward, M. & Gleditsch, K. S. (2008), Spatial Regression Models, Sage Publications, Los Angeles.
Warden, T. (2007), The Engagement of U.S. Cities and the Global Warming Issue, 2005-2007,
PhD thesis, University of California.
Wasserman, S. & Faust, K. (1994), Social Network Analysis: Methods and Applications, Cambridge University Press.
Watson, R. (2009), The Interlinkage of Global Climate Change, Economy and Poverty, Technical
report.
White, H. (1980), ‘A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct
Test for Heteroskedasticity’, Econometrica: Journal of the Econometric Society 48(4), 817–
838.
Winkelmann, R. (2008), Econometric Analysis of Count Data, Springer Verlag.
Wooldridge, J. M. (2009), Introductory Econometrics: A Modern Approach, South Western
Cengage Learning.
Young, H. (2009), ‘Innovation Diffusion in Heterogeneous Populations: Contagion, Social Influence, and Social Learning’, The American Economic Review 99(5), 1899–1924.
Zahran, S., Brody, S., Vedlitz, A. & Grover, H. (2008), ‘Vulnerability and Capacity: Explaining
Local Commitment to Climate-Change Policy’, Environment and Planning C: Government
and Policy 26, 544–562.
Zahran, S., Grover, H., Brody, S. & Vedlitz, A. (2008), ‘Risk, Stress, and Capacity: Explaining
Metropolitan Commitment to Climate Protection’, Urban Affairs Review 43(4), 447–474.