krause - climate change initiatives - us cities

16
POLICY INNOVATION, INTERGOVERNMENTAL RELATIONS, AND THE ADOPTION OF CLIMATE PROTECTION INITIATIVES BY U.S. CITIES RACHEL M. KRAUSE Indiana University ABSTRACT: In the absence of federal requirements, how do state- and municipal-level character- istics impact the probability of local policy innovation? This article provides insight by examining the adoption of sub-national climate change mitigation initiatives in the United States. Drawing from literature on policy innovation, a multilevel model is developed to examine the factors influ- encing over 900 U.S. cities to eschew free-rider tendencies and formally commit to greenhouse gas reduction. Multilevel analysis recognizes the nested structure of cities within states and accounts for the shared economic, political, and policy environments experienced by cities within the same state. The level of initiative state governments have taken toward climate protection varies considerably, and the influence of different state policies on related local decisions is empirically examined. Re- sults are consistent with hypotheses derived from the innovation literature and suggest local-level characteristics are the dominant drivers of cities’ decisions to commit to climate protection. Metropolitan areas are key drivers of climate change while also being uniquely vulnerable to its effects. An estimated 30 to 40% 1 of global anthropogenic greenhouse gas (GHG) emissions emanate from within cities’ boundaries (Intergovernmental Panel on Climate Change, 2008; Satterthwaite, 2008) and due to their higher population densities and extensive land cover modifi- cation, a majority of cities are expected to experience more dramatic climate-induced changes than surrounding areas (Grimmond, 2007). As with most global challenges, however, climate change is generally considered an issue best addressed at national and international levels. Carbon dioxide dissipates globally so the environmental benefits of its reduced atmospheric concentration are the same no matter where in the world emissions are reduced. Unlike many traditional air pollutants, where the costs and benefits of abatement are felt locally, the reduction of GHG emissions results in nonexcludable global benefits. Each abating entity receives only minimal direct benefits from its efforts, yet bears the full burden of the cost. As such, in a noncoercive environment, the incentive to free-ride would theoretically block the involvement of subnational governments in efforts to produce public goods such as increased climate protection (Olson, 1965). In the face of federal inaction and in apparent defiance of this logic, a significant number of U.S. state and local governments have taken initiative to reduce the GHG emissions coming from Direct correspondence to: Rachel M. Krause, Indiana University, School of Public and Environmental Affairs, 1315 E 10th St, Bloomington, IN 47405. E-mail: [email protected]. JOURNAL OF URBAN AFFAIRS, Volume 33, Number 1, pages 45–60. Copyright C 2010 Urban Affairs Association All rights of reproduction in any form reserved. ISSN: 0735-2166. DOI: 10.1111/j.1467-9906.2010.00510.x

Upload: mytemp-accnt

Post on 31-Mar-2016

214 views

Category:

Documents


0 download

DESCRIPTION

M etropolitanareasarekeydriversofclimatechangewhilealsobeinguniquelyvulnerableto JOURNALOFURBANAFFAIRS,Volume33,Number1,pages45–60. Copyright C 2010UrbanAffairsAssociation Allrightsofreproductioninanyformreserved. ISSN:0735-2166. DOI:10.1111/j.1467-9906.2010.00510.x Directcorrespondenceto:RachelM.Krause,IndianaUniversity,SchoolofPublicandEnvironmentalAffairs,1315 E10thSt,Bloomington,IN47405.E-mail:[email protected]. 46 II JOURNALOFURBANAFFAIRS II Vol.33/No.1/2011

TRANSCRIPT

POLICY INNOVATION, INTERGOVERNMENTALRELATIONS, AND THE ADOPTION OF CLIMATE

PROTECTION INITIATIVES BY U.S. CITIES

RACHEL M. KRAUSEIndiana University

ABSTRACT: In the absence of federal requirements, how do state- and municipal-level character-istics impact the probability of local policy innovation? This article provides insight by examiningthe adoption of sub-national climate change mitigation initiatives in the United States. Drawingfrom literature on policy innovation, a multilevel model is developed to examine the factors influ-encing over 900 U.S. cities to eschew free-rider tendencies and formally commit to greenhouse gasreduction. Multilevel analysis recognizes the nested structure of cities within states and accounts forthe shared economic, political, and policy environments experienced by cities within the same state.The level of initiative state governments have taken toward climate protection varies considerably,and the influence of different state policies on related local decisions is empirically examined. Re-sults are consistent with hypotheses derived from the innovation literature and suggest local-levelcharacteristics are the dominant drivers of cities’ decisions to commit to climate protection.

Metropolitan areas are key drivers of climate change while also being uniquely vulnerable toits effects. An estimated 30 to 40%1 of global anthropogenic greenhouse gas (GHG) emissionsemanate from within cities’ boundaries (Intergovernmental Panel on Climate Change, 2008;Satterthwaite, 2008) and due to their higher population densities and extensive land cover modifi-cation, a majority of cities are expected to experience more dramatic climate-induced changes thansurrounding areas (Grimmond, 2007). As with most global challenges, however, climate change isgenerally considered an issue best addressed at national and international levels. Carbon dioxidedissipates globally so the environmental benefits of its reduced atmospheric concentration are thesame no matter where in the world emissions are reduced. Unlike many traditional air pollutants,where the costs and benefits of abatement are felt locally, the reduction of GHG emissions resultsin nonexcludable global benefits. Each abating entity receives only minimal direct benefits fromits efforts, yet bears the full burden of the cost. As such, in a noncoercive environment, theincentive to free-ride would theoretically block the involvement of subnational governments inefforts to produce public goods such as increased climate protection (Olson, 1965).

In the face of federal inaction and in apparent defiance of this logic, a significant number ofU.S. state and local governments have taken initiative to reduce the GHG emissions coming from

Direct correspondence to: Rachel M. Krause, Indiana University, School of Public and Environmental Affairs, 1315E 10th St, Bloomington, IN 47405. E-mail: [email protected].

JOURNAL OF URBAN AFFAIRS, Volume 33, Number 1, pages 45–60.Copyright C© 2010 Urban Affairs AssociationAll rights of reproduction in any form reserved.ISSN: 0735-2166. DOI: 10.1111/j.1467-9906.2010.00510.x

46 II JOURNAL OF URBAN AFFAIRS II Vol. 33/No. 1/2011

their jurisdictions. Even if the federal government increases its involvement in climate protection,state and local governments will remain important units of analysis, in part because they haveauthority over key climate-relevant policy areas often including transportation, land-use, buildingcodes, electricity production and transmission, and waste management (Coenen & Menkveld,2002). Further, subnational governments provide a great deal of variation against which theoriesof policy innovation and governmental decision making can be tested. This article investigatesthe factors that influence cities’ decisions to engage in climate policy innovation, while explicitlyrecognizing the nested political, economic, and policy structure of cities within states.

BACKGROUND ON SUBNATIONAL CLIMATE PROTECTION INITIATIVES

Scholars frequently point to state governments as the climate protection leaders in the UnitedStates (Lustey & Sperling, 2008; Pew Center, 2009; Rabe, 2004). Either on their own or as partof regional groups, a majority of state governments have implemented policies aimed at green-ing energy production. These frequently include requirements for GHG emissions reporting andregional portfolio standards, which require utilities to produce a minimum amount of renewableenergy. Climate action plans, which identify state-specific opportunities to decrease GHG emis-sions, have been developed by 36 states (Pew Center, 2009). More recently, states have begun toadopt GHG targets, most of which are nonbinding, but set formal reduction goals for state efforts.

Organized local-level involvement in the United States on climate change can be traced backto 1991, when six municipalities began working with ICLEI (then known as the InternationalCouncil for Local Environmental Initiatives) to develop comprehensive strategies to reduce theirGHG emissions2 (Betsill, 2001). Membership in ICLEI’s climate protection program increasedslowly for its first decade and a half, until experiencing a notable acceleration in 2005 followingthe launch of the U.S. Mayors’ Climate Protection Agreement (MCPA).3 Sponsored by the U.S.Conference of Mayors, the MCPA reflects the commitment of signatory cities to reduce theirGHG emissions by 7% below their 1990 levels, that is, the Kyoto Protocol’s goal amount for theUnited States. Most local climate change activity falls under the umbrella of these two affiliatedorganizations. As of July 2009, 960 municipalities had signed the MCPA and nearly 550 hadjoined ICLEI. This equates to approximately 5% of all U.S. municipalities, covering nearly 30%of the population, with some type of formal involvement in GHG mitigation. Although networkmembership and the adoption of reduction goals clearly do not equate to effective climate policy,they are an explicit acknowledgment of local ability and responsibility to help mitigate climatechange.

Much of the research that has been conducted on municipal climate protection efforts thus faris qualitative in nature. Initial work utilized case study methodologies to examine the motivationsand approaches of select cities that had adopted explicit climate protection goals (Bulkeley &Betsill, 2003), or to observe the functioning of a specific climate protection network, such asICLEI (Betsill, 2001; Bulkeley & Betsill, 2003; Lindseth, 2004). A second generation of qualita-tive studies evaluates, via interviews and document analysis, the ways in which small samples of(typically) best practice cities implement their climate protection plans (Aall, Groven, & Lind-seth, 2007; Wheeler, 2008). Conclusions emanating from these lines of research suggest thatmunicipalities often frame climate protection initiatives so to emphasize the localized cobenefitsof efforts, such as economic development and local environmental improvements (Bulkeley &Betsill, 2003; Lindseth, 2004). The presence of a committed policy “champion” or “entrepreneur”within the local government has been found as key to both the adoption and implementation ofclimate protection initiatives (Bulkeley & Betsill, 2003). Finally, findings suggest that munici-palities often limit their implementation of climate protection initiatives to activities that target“low hanging fruits” and are generally reluctant to invest their own funds into policy changes

II Policy Innovation, Intergovernmental Relations, and the Adoption of Climate Protection II 47

(Aall et al., 2007; Wheeler, 2008). These studies provide valuable insight to the climate protec-tion dynamic that exists within specific municipalities; however, as with all case study and smallsample research, the ability to generalize findings is limited.

In a series of quantitative papers, Zahran and colleagues use an environmental stress, civiccapacity, and vulnerability lens to address the question of why local governments join climateprotection networks. They find that metropolitan areas characterized by GHG intensive opera-tions/behaviors are significantly less likely to engage in voluntary climate protection activities(Zahran, Grover, Brody, & Vedlitz, 2008). On the other hand, cities with greater levels of humancapital, as measured by demographic statistics such as income, education, and environmentalgroup activity, are significantly more likely to participate in such networks (Zahran, Brody,Vedlitz, Grover, & Miller, 2008; Zahran, Grover et al., 2008). Their findings are less clear regard-ing local vulnerability to climate change-related risk, which is expected to differ considerablyacross the country. In some cases, cities at greater risk of experiencing negative impacts aresignificantly more likely to increase participation in climate protection efforts (Zahran, Brodyet al., 2008). In other cases, this relationship appears to be insignificant (Zahran, Grover et al.,2008).

A fundamental inconsistency exists within the vulnerability causal story. Because carbondioxide dissipates globally, a city can eliminate 100% of its emissions and have virtually noimpact on its climate-related risks. Thus, the idea that higher vulnerability leads to an increasedlikelihood of engagement in GHG mitigation is flawed, unless it is premised upon an assumedmisunderstanding by the public and/or decision makers regarding how climate change works.Although this is not an altogether unreasonable assumption, the vulnerability causal story fitsbetter with adaptation rather than mitigation efforts.

This article takes a new angle and examines local climate protection decisions through a policyinnovation and intergovernmental relations lens. Municipal climate protection initiatives are oftenviewed as a result of local factors and independent from state-level actions (Selin & VanDeveer,2007). However, the rate of local participation in the MCPA varies significantly between states,indicating that state-level policies and/or characteristics might influence the propensity of thecities within their borders to join. The sample of local governments considered by this studyis limited to incorporated places with populations greater than 25,000. Of these, approximately40% are members of the MCPA. However, in a “typical” state, the probability of each citybeing a member is 0.582.4 Ninety percent of states have probabilities of local participation thatfall between 0.185 and 0.894. This distribution illustrates that the probability of local MCPAmembership is uneven across states. The reason for such state-to-state variation is either because(1) the internal characteristics of cities relevant to adoption tend to differ by state, or (2) somethingabout the state economic, political, policy, or social environment influences cities’ propensity toadopt.

POLICY INNOVATION AND RESEARCH HYPOTHESES

Policy innovation, as distinct from policy invention, is described in the literature as the adoptionof a policy or program by a government entity that had never before utilized it; that is, it isnew to the government adopting it, but is not necessarily an altogether new idea (Berry &Berry, 1999). Studies of policy innovation often seek to explain why some governments adopta particular policy while others do not (Berry & Berry, 1990; Feiock & West, 1993; Glick &Hays, 1991) or to explain why some entities are generally more innovative than others (Savage,1978; Walker, 1969). The majority of empirical research on policy innovation has been conductedon the state level. However, a growing number of studies have considered the local adoption ofregulatory policies, including gun control laws (Godwin & Schroedel, 2000) and smoking bans

48 II JOURNAL OF URBAN AFFAIRS II Vol. 33/No. 1/2011

(Skeer, George, Hamilton, Cheng, & Siegel, 2004), and the provision of new services, such ascurb-side recycling (Feiock & West, 1993).

The process by which policy innovation occurs can be characterized as being either “acute” or“incubated” (Deyle, Meo, & James, 1994; Polsby, 1984). Acute innovation occurs over a shortperiod of time at the behest of a small group of decision-makers, often in response to a crisis.Incubated innovation results in policy change developed as part of a longer process addressinga chronic problem or environmental condition. It is more likely to be influenced by scientificand technical information and stakeholder negotiations. The decision-making process associatedwith municipal GHG mitigation commitments can be characterized as incubated innovation.Innovations of this type often face barriers to securing a position on the policy-making agendaand policy entrepreneurs therefore play a considerable strategic role in pushing the issue forward(Deyle et al., 1994). This is consistent with findings from qualitative research on local climateprotection, which often single out the influence of policy entrepreneurs and networks of technicalexperts (such as those within ICLEI), who often frame GHG mitigation policy in ways that arerelevant to local concerns (Bulkeley & Betsill, 2003; Lindseth, 2004).

In a seminal piece on innovation in organizations, Mohr (1969) finds evidence supporting hishypothesis that innovation is “directly related to the motivation to innovate, inversely related tothe strength of obstacles to innovation, and directly related to the resources available for over-coming such obstacles” (p. 114). Berry and Berry (1990, 1999) identify two distinct perspectivesfrom which policy innovation is typically studied, which they stress are not mutually exclusiveunder Mohr’s theory. The first perspective considers the internal determinants of the adoptinggovernment. According to this view, the factors encouraging or restricting innovation are thepolitical, economic, and/or social characteristics of the potentially adopting government. Thesecond perspective, tested via diffusion models, holds that innovation is the result of the emula-tion of policies previously adopted by other governmental units, which have been communicatedvia intergovernmental channels. Although traditionally considered separately, Berry and Berry(1999) posit that more realistic empirical models consider both simultaneously, and recommendevent history analysis as a means of estimation.

This article adds to the existing literature in two primary ways. First, both the unit of analysisand the type of policy it considers are relatively underexamined in innovation studies. The articleconsiders the adoption, by local governments, of an entrepreneurial policy, that is, one whosebenefits are dispersed but whose costs are concentrated locally (Wilson, 1980). Second, it utilizesmultilevel modeling, a methodology not commonly seen in adoption studies, but one that iscalled for when the adopting units of government are embedded within larger ones, such as citieswithin states. This model examines how the characteristics internal to cities and states impactthe likelihood of local innovation. It also allows us to see if previously enacted state policiesinfluence the likelihood that cities within their borders will adopt related policies. This can beseen as a form of vertical diffusion.

The first set of hypotheses presented in this article builds from Shipan and Volden (2006) whosework examines the vertical diffusion of smoking regulations from local to state governments.They posit that initiatives taken at one level of government may exert either a complementary orweakening influence to similar initiatives taken at another level. Expanding from their work, thisarticle proposes two opposing hypotheses regarding the potential influence of state-level climatechange policies on the local adoption of related ones. The first suggests that state-level climatepolicies will enable and/or encourage the adoption of related local policies (i.e. they will createa “snowball effect,” as termed by Shipan and Volden). On the other hand, the second positsthat state-level policies will reduce the probability of local initiatives because they lead to theperception that sufficient action is already being taken and additional local action is therefore notnecessary (i.e., a “pressure valve effect”) (Shipan & Volden, 2006).

II Policy Innovation, Intergovernmental Relations, and the Adoption of Climate Protection II 49

The article’s second set of hypotheses considers the impact of cities’ and states’ internal charac-teristics and mirror those suggested by Mohr (1969): namely, the likelihood of a city’s membershipin the MCPA is hypothesized to be determined by the relative strengths of its motivations andobstacles to environmental action, modified by the resources it has available to overcome thoseobstacles. The next section describes the social, economic, and political independent variablesused to model cities’ decisions to commit to climate protection and categorizes them accordingto Mohr’s hypothesis.

DATA AND METHODS

Sample and Variables

The sample of level-1 observations considered in this article draws from the 1,078 incorporatedplaces in the United States with populations greater than 25,000, per the 2000 County and CityDatabook. Of these, 1,026 or approximately 95%, have sufficient data to be included in theanalysis. The 50 U.S. states make up the level-2 observations.

Of interest is the probability that a city will become involved in climate change protectionthrough the adoption of local initiatives. A dichotomous dummy variable representing participa-tion of the U.S. Mayor’s Climate Protection Agreement is used as this article’s dependent variable(1 if participant; 0 if not). Signatories to the Agreement commit to taking actions that will reducetheir cities’ GHG emissions by 7% below their 1990 levels, that is, the amount specified for theUnited States in the Kyoto Protocol.

Independent variables are introduced at both the state and local levels (See Table 1 for thedescription and source of each and Table 2 for summary statistics). Following Mohr’s hypothesis,the independent variables can be viewed as resources, motivations, or obstacles either supportingor inhibiting innovation. The level-1 independent variables that are indicative of the resourcesa city has available to pursue innovative policies include its population size and per capitageneral revenue. A longstanding consensus in the literature suggests that large entities withgreater resources are more likely to adopt policy innovations (Walker, 1969). It follows that largecities are more likely to have the, sometimes considerable, administrative capacity necessary tocoordinate climate protection programs (Betsill, 2001). Ownership of a municipal utility mayalso indicate higher city income and the increased ability to adopt and follow-through on GHGreduction pledges. The experiences of similar or nearby entities with a new policy can serve as aninformational resource and facilitate adoption. Horizontal diffusion studies, often conducted atthe state level, suggest that the adoption of policies by neighboring states increases the likelihoodof a given state to similarly adopt (Berry & Berry, 1999). The number of neighboring cities eachcity has that are members of the MCPA is also included as an independent “resource” variable.

Certain demographic statistics such as a city’s median household income and average edu-cational attainment can be viewed as motivations to innovation, as both environmental concernand civic engagement are correlated with these attributes (Rothenberg, 2002; Verba, Schlozman,Brady, & Nie, 1993). Climate change initiatives in the United States, particularly those that involveemission caps, have been characterized by partisanship. Democrats are generally more in favorof such efforts, whereas Republicans have more often been opposed. As such, an independentvariable indicating local political leanings estimates the level of political support or oppositionthat may accompany the adoption of a local climate change initiative. The presence of a Mayor-Council form of local government is another possible feature motivating the adoption of climatechange initiatives, as this political form has the effect of making local government more overtlypolitical. Along these lines, Clingermayer (1990) finds that the adoption of symbolic policiesand credit-claiming is more common in municipalities with Mayor-Council government types.

50 II JOURNAL OF URBAN AFFAIRS II Vol. 33/No. 1/2011

TABLE 1

Variables Description and Source

Dependent VariableMayors’ Climate Protection

AgreementA dummy variable indicating whether or not each city’s mayor had signedthe Mayors’ Climate Protection Agreement by May 2008, committing thecity to achieve what would have been the U.S. GHG reduction goal underthe Kyoto Protocol.

Source: U.S. Conference of Mayors, Mayors Climate Protection Center.Level 1 (Local) Variables

Population Logged population of each city in 1999.Source: County and City Data Book: 2000 CD-Rom.

Education Percentage of population over the age of 25 with a BA or higher.Source: U.S. Census Bureau 2000, SF-3.

Income Median household income in 1999 in $1,000s.Source: U.S. Census Bureau 2000, SF-3.

Percent Democrat The percentage of each county’s total votes that supported the Democraticcandidate in the 2000 presidential election.

Source: CQ Voting and Elections Collection.Per capita general revenue Per capita general revenue for each city in 1997 in $100s.

Source: County and City Data Book 2000, Department of Housing andUrban Development’s State of Cities data source.

Type of city government A dummy variable indicating if a city has a mayor-council form of govern-ment(1) or a different form(0).

Source: The Municipal Year Book 2000, International City/CouncilManagement Association.

Number of unhealthy air days The number of days in 2000 that each city’s county had “unhealthy” airquality as determined by the EPA’s Air Quality Index.

Source: Environmental Protection Agency, Air Quality Standard.Municipal utilities A dummy variable indicating the municipal ownership of an electricity pro-

ducing or distributing utility.Source: U.S. Energy Information Administration.

Participating “neighbors” The number of “neighbor cities” within a 50-mile radius of each city that ishas joined the Mayor’s Agreement.

Source: Constructed with GIS using data from the U.S. Conference ofMayors and the U.S. Census Bureau.

Value added by manufacturing Value (in $100,000s) added by the manufacturing sector to each city’scounty or metro/micropolitan statistical area in 2002.

Source: U.S. Census Bureau, 2002 Economic Census, NAICS 31-33.Level 2 (State) Variables

Climate action plan A dummy variable indicating whether or not a state had completed aclimate action plan specifying ways GHG emissions can be reduced state-wide prior to the 2005 start of the Mayors’ Climate Change Agreement.

Source: Pew Center on Global Climate Change, Knigge, and Bausch(2006).

GHG target A dummy variable indicating whether or not a state adopted a greenhousegas reduction target (either binding or nonbinding prior to the 2005 startof the Mayors’ Climate Change Agreement.

Source: Environmental Protection Agency.Government ideology A measure of the ideological position of state governments in 2005. Higher

scores are more liberal, lower are conservative.Source: Revised 1960–2006 citizen ideology series, updated Berry,Ringquist, Fording, and Hansen 1998.

Manufacturing Percentage of each state’s domestic product that came from manufactur-ing in 2005.

Source: U.S. Bureau of Economic Analysis.

II Policy Innovation, Intergovernmental Relations, and the Adoption of Climate Protection II 51

TABLE 2

Descriptive Statistics

Mean Std. Dev N

Level 1: City CharacteristicsMayor’s Agreement Signatory 0.40 0.49 1026

Resources Population (logged) 11.11 0.78 1026Gen rev (per cap, 100s) 10.37 6.45 1026Municipal utility 0.12 0.32 1026Participating neighbors 6.67 7.79 1026

Motivations Income (1,000’s) 43.62 15.25 1026Education 26.50 13.55 1026Government type 0.33 0.47 1026Unhealthy air quality days 15.46 27.31 1026Percent vote Democrat 49.82 11.79 1026

Obstacle Manufacturing value added ($100,000’s) 86.29 137.26 1026

Level 2: State CharacteristicsUnclear Greenhouse gas target 0.16 0.37 50

Climate action plan 0.58 0.50 50Obstacle Pcnt. GSP manufacturing 0.12 0.06 50Motivation Liberal government ideology 49.94 26.74 50

Studies frequently show that municipalities emphasize the local cobenefits of climate protection,including the public health and visibility improvements associated with decreased local air pol-lution (Betsill, 2001; Bulkeley & Betsill, 2003). In certain cities, leadership in climate protectionand alternative energy has been directly tied to community air quality concerns (Portney, 2004).As such, having poor urban air quality may provide cities with additional motivation to adoptclimate change initiatives.

The decline of manufacturing in the United States has been pointed to as a factor enablinglocal governments to become engaged in planning efforts that explicitly include sustainability asan objective (Portney, 2003; Zahran, Grover et al., 2008). Despite a potential conflict with thepreviously described assumption that poor air quality may motivate climate protection initiatives,cities where manufacturing remains economically important may be less inclined to adopt GHGreducing initiatives than those that have a service or technology-based economy. The amount ofeconomic value added by manufacturing is considered the best measure of the sector’s relativeimportance to specific geographic areas and is included in the model as an independent variablehypothesized to act as an obstacle to adoption (U.S. Census Bureau, 2002)

Like local variables, state-level variables can provide motivation, resources, and obstacles tolocal policy adoption. For example, Feiock and West (1993) find that most state-level efforts toencourage municipalities to provide recycling services significantly increase the probability thatthis will occur. Under the climate change scenario, the state-level encouragement is less direct.For example, as of the time of this study, no states mandate that cities reduce their GHGs, whilesome do mandate recycling. Still, the presence of state climate policies, such as climate actionplans (CAPs), may serve as both a resource and a motivation for cities to adopt their own relatedinitiatives. CAPs describe sources of GHG emissions within a state and how they can be reduced.This information may prove helpful to cities in the development of their own plans and theadoption of a CAP may send a signal that the state considers climate change protection importantand is likely to pursue additional policies in that area. On the other hand, the presence of a CAPmight provide a “pressure valve effect” and reduce local motivation (Shipan & Volden, 2006).

52 II JOURNAL OF URBAN AFFAIRS II Vol. 33/No. 1/2011

State GHG reduction targets, whether binding or not, may similarly influence local propensity toadopt. Other potentially relevant state-level characteristics include the amount that manufacturingcontributes to a state’s economy (potential obstacle) and the ideology of the state government ona liberal–conservative continuum (potential obstacle, as it moves right).

Method of Analysis

Multiple cities are clustered within each state and are thus subject to the same state-level politi-cal, policy, and economic conditions. As a result of this nested structure, the assumption that eachcity represents an independent observation is likely incorrect and statistical methods that assumeindependently and identically distributed errors are thus inappropriate (Primo, Jacobsmeier, &Milyo, 2007). Standard statistical methods, including logit and probit models, are likely to yieldresults that underestimate the standard errors associated with coefficient estimates, particularlygroup-level coefficients (Snijders & Bosker, 1999). As such, this article uses multilevel modeling,which allows independent variables at different levels to be analyzed simultaneously and accountsfor the likelihood that observations of cities within the same state are not completely independentof each other.5

The idea underlying multilevel modeling is that the dependent value, Y , is influenced bothby individual level-1 (local) and group level-2 (state) variables (Snijders & Bosker, 1999). Thegeneral structural form of a random intercept equation is

Yij = β0j + β1jX1ij + rij

where

β0j = γ00 + γ01Z1ij + u0j

β1j = γ10,

where i represents individuals, j represents groups and β0J is the intercept that varies according tothe value of level-2 independent variable Z1ij and the error term u0j . Although multilevel modelscan also allow for the possibility that group-level forces influence the value of the coefficientsof level-1 variables (β1j ), in random intercept models they are held constant (note the absenceof an error term and level-2 variables in β1j = γ10). The question of what factors influence localclimate initiative adoption is explored via a random intercept logistic multilevel model and isestimated with the statistical program HLM6.

RESULTS

The multilevel model in this article is developed following the process described by Hox(1995). Notably, only city-level variables are statistically significant. The final Bernoulli modelpredicting the likelihood of cities participating in the U.S. MCPA is represented by the followingstructural equations:

Level-1: Structural Equation

η = β0 + β1(logpop) + β2(income) + β3(education) + β4(democrat)

+β5(govtype) + β6(gen rev) + β7(municipal util) + β8(air quality)

+β9(neighbor) + β10(man val added)

II Policy Innovation, Intergovernmental Relations, and the Adoption of Climate Protection II 53

Level-2: Structural Equation

β0 = γ00 + u0

β1 = γ10

β2 = γ20

β3 = γ30

β4 = γ40

β5 = γ50

β6 = γ60

β7 = γ70

β8 = γ80

β9 = γ90

β10 = γ10

As determined by deviance tests, the final model is not statistically different from the full modelthat has state-level variables added to the intercept term (see Table 3).

Findings at the State Level

None of the state-level characteristics were found to significantly influence the propensityof local governments to adopt climate protection policies. Thus, neither the “snowball” nor“pressure valve” hypotheses described above receive empirical support: the presence of stateGHG reduction targets and/or climate action plans fail to impact local adoption in either direction.This is different from previous findings that show support for the snowball-effect hypothesis, thatis, that policies adopted by one level of government increase the likelihood that other levels willadopt similar policies (Feiock & West, 1993; Shipan & Volden, 2006). A key policy implicationof this finding is that if state climate policy goals include having local governments follow suit byadopting complementary initiatives, more explicit encouragement is necessary. States “leadingby example” is not effective in this context.

When compared to Feiock and West’s (1993) study, which found that supportive state-levelpolicies significantly increase the probability that local governments will adopt curb-side re-cycling programs, two factors stand out that may explain the different outcomes. First, severalof the state policies they examine encourage local policy adoption directly. Mandates for theprovision of some kind of local recycling (although not necessarily curb-side service) and incen-tive programs indeed may have had a greater effect on local actions than do the more indirectclimate action plans and GHG reduction goals. A second explanation that may account for thedifference in outcome significance concerns the statistical methods employed. Feiock and Westuse a standard probit analysis. The use of standard models, which assume observations are fullyindependent of each other, can result in standard errors for level-2 variables being underestimatedwhen observations are clustered into groups. The statistical significance of state-level policies onthe adoption of complimentary local ones may be less than is generally thought because standardanalysis leads to their overstatement.

Findings at the Local Level

All but one of the city-level variables included in the analysis are significant, and most havethe effect predicted by Mohr’s motivation-resource-obstacle hypothesis. As expected, having

54 II JOURNAL OF URBAN AFFAIRS II Vol. 33/No. 1/2011

TABLE 3

Impact of State and Local Characteristics on City’s Decision to Commit to Greenhouse Gas EmissionReduction by Signing the Mayor’s Climate Protection Agreement

Model 2:Model 1: Local Local + State

Variable Characteristics Characteristics

Intercept −0.531∗∗∗ (0.152) −0.669∗∗∗ (0.247)City-Level VariablesFixed Effects

Population (logged) 0.932∗∗∗ (0.113) 0.939∗∗∗ (0.114)Income −0.038∗∗∗ (0.008) −0.038∗∗∗ (0.008)Education 0.067∗∗∗ (0.008) 0.068∗∗∗ (0.008)Percent Democrat 0.031∗∗∗ (0.009) 0.031∗∗∗ (0.009)Government type 0.353∗ (0.189) 0.347∗ (0.194)General revenue 0.046∗∗∗ (0.015) 0.045∗∗∗ (0.015)Municipal utility −0.490∗∗ (0.241) −0.498∗∗ (0.241)Unhealthy air days −0.002 (0.004) −0.003 (0.005)Participating neighbors 0.040∗∗ (0.015) 0.038∗∗ (0.016)Manufacturing value added −0.002∗∗∗ (0.001) −0.002∗∗∗ (0.001)

State-Level VariableEffects on the Intercept

GHG reduction target 0.207 (0.383)Climate action plan 0.142 (0.313)Government ideology −0.003 (0.005)GSP from manufacturing 0.992 (2.379)

Model CharacteristicsReliability 0.385 0.385Final variance component 0.267 0.268Deviance (parameters) 2985.981 (12) 2985.106 (16)

∗p < 0.1, ∗∗p < 0.05, ∗∗∗p < 0.01, Standard errors in parenthesis.Level-1 n = 1026, Level-2 n = 50.Note: Continuous variables are grand mean centered, dichotomous variables are uncentered; population average modelsoutcome are shown; full maximum likelihood used.

a mayor-council government type and citizens with higher levels of education and democraticpolitical leanings appear to be significant motivations for local climate protection innovation.Larger populations and higher levels of per capita general revenue are significant enablingresources. Evidence of horizontal diffusion is also present, with results suggesting that citieswith larger numbers of participating neighbors are themselves more likely to commit to climateprotection by joining the MCPA. This may be a result of communication and information sharingthat occurs between officials in neighboring cities or of increased public pressure resulting fromheightened regional awareness about the initiative. In terms of substantive significance, a city’ssize, the level of its residents’ educational attainment, and the number of its neighbors that areMCPA participants have the largest impact on the likelihood of climate protection commitment(see Table 4).

Results also support the stated hypothesis that a higher reliance on manufacturing in a localeconomy decreases the probability that a city will commit to climate protection. This is consis-tent with the concept of incubated innovation, in which the adoption of new policy is heavilyinfluenced by the ability to achieve consensus among stakeholders (Deyle et al., 1994). A strongmanufacturing interest likely decreases the ability to form consensus around local climate pro-tection. These results also reflect findings from Zahran, Grover et al., (2008), who conclude that

II Policy Innovation, Intergovernmental Relations, and the Adoption of Climate Protection II 55

TABLE 4

Substantive Effect Variables on Joining the Mayor’s Climate Protection Agreement

Independent Variable Change in the Likelihood of Joining

Logged population (1 std. deviation increase) 0.180∗∗∗Income (1 std. dev. increase) −0.127∗∗∗Education (1 std. dev. increase) 0.223∗∗∗Democratic voters (1 std. dev. increase) 0.091∗∗∗Mayor Council Gov. (0 to 1) 0.086∗Per cap general rev. (1 std. dev. increase) 0.073∗∗∗Municipal utility (0 to 1) −0.119∗∗Unhealthy air (1 std. dev. increase) −0.06Participating neighbors (1 std. dev. increase) 0.151∗∗Manufacturing value added (1 std. dev. increase) −0.006∗∗∗

∗p < 0.1, ∗∗p < 0.05, ∗∗∗p < 0.01.

the amount of stress a locale exerts on the environment, in terms of high levels of engagementin carbon-intensive activities, is negatively associated with joining a climate protection network.The real and perceived costs of GHG reduction, both in terms of financial investment and lifestyledisruption, act as a statistically significant barrier. However, and perhaps surprisingly, the mag-nitude of this barrier effect is quite small. Cities’ likelihood of MPCA participation decreases byless than 1% as the total amount of money manufacturing adds to a local economy increases bya standard deviation.

Unexpected results include the negative signs associated with median household income andcity ownership of a municipal electric utility. Referring back to Mohr’s hypothesis, having ahigher median household income was originally classified as a factor motivating innovation incity governance. Environmental protection is typically seen as a normal good, the demand forwhich increases along with prosperity and economic well-being (Rothenberg, 2002). This makesthe resulting negative sign for income appear counterintuitive. However, the significant negativevalue for income is present only when controlling for education. When the variable for educationis removed from the model, the impact of income loses all significance.

A Closer Look at the Influence of Municipally Owned Utilities

The ownership of a municipal electric utility was originally considered a resource able toprovide support to local GHG emission reduction efforts. However, the results of the analysissuggest that it instead acts as an obstacle. All else equal, cities with a municipally owned utility arealmost 12% less likely to join the MCPA than those without one. This runs counter to observationsmade in several prominent cases where municipal utilities have assumed leadership roles in thepromotion of local GHG reduction by providing staff and funding for climate protection as wellas making changes to their own energy production activities.6

Municipal utilities are nonprofit organizations, free from shareholder demands. This organiza-tional structure could conceivably make them more responsive to local climate protection initia-tives. On the other hand, many municipal utilities act only as energy distributors and those thatdo generate their own electricity are on average more carbon-intensive than investor-owned ones(Wilson, Plummer, Fischlein, & Smith, 2008). Further, because they are often small, municipallyowned utilities are regularly exempt from state and federal regulations, including requirementsfor demand-side management (DSM) programs (Wilson et al., 2008).

56 II JOURNAL OF URBAN AFFAIRS II Vol. 33/No. 1/2011

TABLE 5

Impact of Municipal Utility Characteristics on Joining MCAP

Parameter Estimate Standard Error

Income −0.074∗∗ 0.031Education 0.101∗∗∗ 0.027Percent Democrat 0.022 0.234Government type 0.477 0.518General revenue −0.009 0.042Unhealthy air days 0.021 0.014Participating neighbors 0.130∗∗ 0.059Manufacturing value added −0.001∗ 0.000Sales (TWh)a 0.211∗∗ 0.086Nameplate capacity (MW)b 0.003∗∗∗ 0.001Constant −2.860∗∗ 1.563

Logit analysis. ∗p < 0.10, ∗∗p < 0.05, ∗∗∗p < 0.01.x2 45.39 Prob x2 0.000 n = 123.aUtilities’ total electricity sales in terawatt hours, Energy Information Administration, 2005.bNameplate capacity in megawatts; a generator’s maximum output. Fifty-eight of the utilities have a nameplate capacity ofzero, meaning they produce no electricity inhouse, but distribute power purchased from other generators.

One hundred twenty-three cities from the original sample own electric utilities. A subsequentanalysis of this subsample suggests that utility size and generator characteristics are significantdeterminants of a city’s MCPA membership. Small, distribution-only utilities act as the maininhibitors. Cities that own larger utilities, both in terms of total sales and generating capac-ity, are more likely to make reduction commitments than their small-utility counterparts (seeTable 5). This may be because larger operations are subject to more regulations and thus mayalready be cleaner. Also, when compared to small utilities, larger ones likely have higher capacity,both in terms of their ability to change their own power-generating operations and to allocate staffand financial resources to follow through on GHG reduction commitments. Finally, efficiencyinnovations undertaken by utilities, such as DSM, often result in short-term revenue loss andyield savings only if they defer the construction of new facilities (Hirst, 1994). Such savings areunlikely to be realized by distribution-only utilities and by small municipal generators not facingincreasing demand.

DISCUSSION AND POLICY IMPLICATIONS

This article’s first major finding is that state-level characteristics and climate-related policieshave an insignificant effect on the likelihood that their cities are active in climate protection. Ofthe state climate plans and GHG reduction targets that have been adopted, few have any regulatoryteeth. California is the only state with legislation that enables enforcement for failure to adhereto emission caps (Pew Center, 2009). Information and vague goals emanating from state officesare insufficient to motivate municipal decision making. State government’s “leading by example”fails to yield results in this case. If this is an objective of state climate policy, more explicit positiveincentives (i.e., intergovernmental grants) or negative incentives (i.e., enforcement threats) appearnecessary.

Whereas state-level variables fail to have significant impacts, local-level characteristics largelydo. Demographic characteristics, including city size, education rate, median income, and politicalleaning, are significant determinants of city membership in the MCPA, as are the form ofmunicipal government, the amount of per capita of general revenue collected, and the importance

II Policy Innovation, Intergovernmental Relations, and the Adoption of Climate Protection II 57

of manufacturing in the local economy. While these findings are interesting in and of themselves,their implications for policy are minimal as little can be done to change them for the sake ofincreasing attention to climate protection. On the other hand, the findings regarding the presenceof horizontal diffusion and the impact of municipally owned utilities may be of practical use topolicymakers seeking to increase local involvement in climate protection.

As previously discussed, this article’s results show that, all else equal, cities that are surroundedby higher numbers of MCPA members are themselves more likely to join. The diffusion of localpolicy to reduce GHGs is no doubt influenced by the activities of climate protection networks(Bulkeley & Betsill, 2003). The regionally based diffusion patterns may reflect areas of focus innetwork activity or may reflect intergovernmental communication channels, which are strongestbetween nearby cities. Either way, the observed patterns of horizontal diffusion could be leveragedby policymakers with the creation and/or support of formal geographic networks designed asforums of communication about innovative local policy.

The finding that municipally owned utilities, particularly small ones, tend to act as obstacles toclimate protection initiatives is similarly important. First, it tempers assumptions that municipalutilities are showing widespread leadership on this issue. Second, the U.S. Environmental Pro-tection Agency, the Department of Energy, and many state regulators are actively interested inremoving the many contradictory incentives that exist, which limit utilities’ engagement in energyconservation/efficiency measures (U.S. EPA & U.S. DOE, 2006). In the case of municipal utilitiesand climate protection, it appears that obstacles extend beyond the utilities’ direct operations andcontinue to influence the decision-making of the cities that own them. It is likely that over thenext several years regulations for utility operations will be revised considerably and the proposedchanges would benefit from additional study exploring this dynamic.

LIMITATIONS AND FUTURE RESEARCH

Two limitations to this study deserve note and offer direction for future research. First, aswith all adoption models that use dichotomous dependent variables, joining the MCPA does notdistinguish between “deep” and “superficial” commitment (Berry & Berry, 1999; Glick & Hays,1991). Two cities may adopt the same policy label or goal, but put different amounts of effortinto achieving it. The Mayor’s Climate Protection Agreement has no enforcement mechanismor monitoring system, making it possible for some cities to sign on and take minimal follow-through action, while others make significant policy changes to mitigate their GHG emissions.Currently, there is not a single comparable measure that adequately indicates the extent of localGHG reduction efforts taking places within a large number of municipalities, and a lack ofreliable GHG emissions data for units below the state level makes generalizable evaluation oflocal policy impact virtually impossible. To make progress in this area, researchers must acquirecomparable data on the specific efforts that a large number of local governments are undertakingto reduce their GHG emissions (Krause, 2009), arrive at a consistent methodology by whichto estimate local GHG emissions (Dodman, 2009), and develop a set of criteria to evaluate theimplementation of climate change programs, perhaps by building off assessments conducted atthe state level (Feldman & Wilt, 1996; Lustey & Sperling, 2008).

Second, the model employed in the study is static, but is looking at a process that is somewhatdynamic. Panel data and event history analysis are thus often recommended as the most appropriatemethodology for adoption models (Berry & Berry, 1999). However, the time frame in this studyis three years, much shorter than most, so the use of models that account for changes over time isless necessary. As the relevant time frame grows, dynamic models may offer important additionalinsights. Examining when a city joins, as opposed to simply if it does, may draw a clearer pictureabout the factors that lead to this decision. Dynamic extensions of the model presented in this

58 II JOURNAL OF URBAN AFFAIRS II Vol. 33/No. 1/2011

article may prove particularly useful in examining potential ripple effects in policy across differentlevels of government, once a comprehensive federal climate protection policy is adopted.

CONCLUSION

Subnational governments have become the de facto leaders of the U.S. climate protectionefforts. Direct municipal involvement in the issue has accelerated since the 2005 formation of theMayor’s Climate Protection Agreement. Mayoral signatories to the Agreement commit to takeactions to reduce GHG emissions in their cities. Significant state-by-state differences exist in therate of city participation, which raises the question about whether this is influenced by differencesin the state-level policy or economic environment. The results of the multilevel model run heresuggest that this is not the case. Rather, the primary determinants influencing the propensity ofmunicipalities to make GHG reduction commitments are local. Several demographic, economic,and city government characteristics emerge as significant determinants of local involvementin climate protection and generally support Mohr’s motivation-resources-obstacles hypothesis.Findings of horizontal diffusion and the barriers posed by municipally owned electric utilitieshave potential to assist the efforts of policymakers interested in encouraging local involvementin climate protection.

ENDNOTES

1 Considerable debate exists over the proper accounting framework to use when assessing local GHG emissions,including how to determine appropriate urban boundaries and whether to use a production- or consumption-based emissions methodology (Dodman, 2009; Larsen, 2009). Satterthwaite (2008) utilizes IPCC data and aproduction-based methodology and likely produces a lower-end estimate.

2 The six U.S. participants included: Chula Vista, CA; Dade County, FL; Denver, CO; Minneapolis, MN; Portland,OR; and St. Paul, MN.

3 Membership in ICLEI increased an average of less than 10 U.S. cities a year for its first decade (Betsill, 2001;ICLEI, 1997). Membership in 2008 is increasing by up to four cities a week (Susan Ode, Outreach director,ICLEI, personal communication, July 2008).

4 This is calculated by running an unconditional two-level model, that is, one with no explanatory variables ateither level. The resulting log odds of participation is −0.3312, making the corresponding probability 0.582.Raudenbush and Bryk (2002) recommend this procedure as part of examining the magnitude of level-2 variationin models with binary dependent variables.

5 The number of level-1 observations (cities) varies widely from group to group. The range extends from onesample city in Vermont to 217 cities in California. The result is less efficient estimates than would otherwiseresult.

6 Examples of municipally owned utilities that are particularly active in climate protection include Austin Energy,Los Angeles Water and Power, and Seattle City Light.

REFERENCES

Aall, C., Groven, K., & Lindseth, G. (2007). The scope of action for local climate policy: The case of Norway.Global Environmental Politics, 7(2), 83–101.

Berry, F. S., & Berry, W. D. (1990). State lottery adoptions as policy innovations: An event history analysis.American Political Science Review, 84, 395–415.

Berry, F. S., & Berry, W. D. (1999). Innovation and diffusion models in policy research. In P. A. Sabatier (Ed.),Theories of the policy process (pp. 169–200). Boulder, CO: Westview Press.

II Policy Innovation, Intergovernmental Relations, and the Adoption of Climate Protection II 59

Betsill, M. M. (2001). Mitigating climate change in U.S. cities: Opportunities and obstacles. Local Environment,6(4), 393–406.

Bulkeley, H., & Betsill, M. (2003). Cities and climate change: Urban sustainability and global environmentalgovernance. New York: Routledge.

Clingermayer, J. C. (1990). The adoption of economic development models by large cities: A test of economic,interest group, and institutional explanations. Policy Studies Journal, 18(3), 539–552.

Coenen, F., & Menkveld, M. (2002). The role of local authorities in a transition towards a carbon-neutral society.In M. T. J. Kok, W. J. V. Vermeulen, A. P. C. Faaij, & D. de Jager (Eds.), Global warming and socialinnovation: The challenge of a climate neutral society (pp. 107–125). London: Earthscan Publications.

Deyle, R. E., Meo, M., & James, T. E. (1994). State policy innovation and climate change: A coastal erosion analog.In D. L. Feldman (Ed.), Global climate change and public policy (pp. 39–66). Chicago: Nelson-Hall.

Dodman, D. (2009). Blaming cities for climate change? An analysis of urban greenhouse gas emissions inventories.Environment and Urbanization, 21, 185–201.

Feiock, R. C., & West, J. (1993). Testing competing explanations for policy adoption: Municipal solid wasterecycling programs. Political Research Quarterly, 46(2), 399–419.

Feldman, F. L., & Wilt, C. A. (1996). Evaluating the implementation of state-level global climate change programs.The Journal of Environment and Development, 5(1), 46–72.

Glick, H. R., & Hays, S. P. (1991). Innovation and reinvention in state policymaking: Theory and the evolution ofliving will laws. Journal of Politics, 53, 835–850.

Godwin, M. L., & Schroedel, J. R. (2000). Policy diffusion and strategies for promoting policy change: Evidencefrom California local gun control ordinances. Policy Studies Journal, 28(4), 760–776.

Grimmond, S. (2007). Urbanization and global environmental change: Local effects of urban warming. TheGeographical Journal, 173(1), 83–88.

Hirst, E. (1994). Electric-utility DSM programs in a competitive market, Oak Ridge, TN: Office of EnergyEfficiency and Renewable Energy, Oak Ridge National Laboratory.

Hox, J. J. (1995). Applied multilevel analysis. Amsterdam: TT-Publikaties.Intergovernmental Panel on Climate Change (2008). Climate change 2007: Synthesis report. Geneva, Switzerland:

Author.International Council for Local Environmental Initiatives (ICLEI) (1997). Local government implementation of

climate protection: A report to the United Nations. Toronto: ICLEI.Krause, R. (2009, November). Symbolic or substantive policy? Measuring the extent of local commitment to

climate protection. Paper presented at the Association for Public Policy Analysis and Management meeting,Washington: DC.

Larsen, H. N. (2009). The case for consumption-based accounting of GHG emissions to promote local climateaction. Environmental Science and Policy, 12(7), 791–798.

Lindseth, G. (2004). The cities for climate protection campaign (CCPC) and the framing of local climate policy.Local Environment, 9(4), 325–336.

Lustey, N., & Sperling, D. (2008). America’s bottom-up climate change mitigation policy. Energy Policy, 36,673–685.

Mohr, L. (1969). Determinants of innovation in organizations. American Political Science Review, 75, 963–974.Olson, M. (1965). The logic of collective action. Cambridge, MA: Harvard University Press.Pew Center on Global Climate Change (2009). Climate change 101: State action. Retrieved from www.

pewclimate.org/docUploads/Climate101-State-Jan09_1.pdf on July 7, 2009.Polsby, N. A. (1984). Political innovation in America: The politics of policy innovation. Berkeley: University of

California Press.Portney, K. E. (2003). Taking sustainable cities seriously. Cambridge, MA: MIT Press.Portney, K. E. (2004). Civic engagement and sustainable cities in the U.S. Public Administration Review, 65(5),

579–592.Primo, D. M., Jacobsmeier, M. L., & Milyo, J. (2007). Estimating the impact of state policies and institutions with

mixed-level data. State Politics and Policy Quarterly, 7(4), 446–459.Rabe, B. (2004). Statehouse and greenhouse. Washington, DC: Brookings Institute Press.Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods.

Thousand Oaks, CA: Sage Publications.

60 II JOURNAL OF URBAN AFFAIRS II Vol. 33/No. 1/2011

Rothenberg, L. S. (2002). Environmental choices: Policy responses to green demands. Washington, DC: CQ Press.Satterthwaite, D. (2008). Cities’ contribution to global warming: Notes on the allocation of greenhouse gas

emissions. Environment and Urbanization, 20, 539–549.Savage, R. L. (1978). Policy innovativeness as a trait of American states. Journal of Politics, 40, 212–224.Selin, H., & VanDeveer, S. D. (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 U.S. cities

to states. American Journal of Political Science, 50(4), 825–843.Skeer, M., George, S., Hamilton, W. L., Cheng, D. M., & Siegel, M. (2004). Town-level characteristics and

smoking policy adoption in Massachusetts: Are local restaurant smoking regulation fostering disparities inhealth protection? American Journal of Public Health, 94(2), 286–292.

Snijders, T., & Bosker, R. (1999). Multilevel analysis: An introduction to basic and advanced multilevel modeling.Thousand Oak, CA: Sage Publications.

United States Census Bureau (2002). 2002 Economic Census, Manufacturing Geographic Area Series. Report No.EC02-31S-G1. Retrieved from www.census.gov/econ/census02/guide/EC02_31.HTM on March 10, 2009.

United States Environmental Protection Agency & Department of Energy (2006). National Action Plan for EnergyEfficiency. Retrieved from www.epa.gov/cleanrgy/documents/napee/napee_report.pdf on April 15, 2009.

Verba, S., Schlozman, K. L., Brady, H., & Nie, N. H. (1993). Race, ethnicity and political resources: Participationin the United States. British Journal of Political Science, 23, 453–497.

Walker, J. L. (1969). The diffusion of innovations among the American states. American Political Science Review,63, 880–899.

Wheeler, S. M. (2008). State and municipal climate plans. Journal of the American Planning Association, 74(7),481–497.

Wilson, E. J., Plummer, J., Fischlein, M., & Smith, T. M. (2008). Implementing energy efficiency: Challenges andopportunities for rural electric co-operatives and small municipal utilities. Energy Policy, 36, 3383–3397.

Wilson, J. Q. (1980) The politics of regulation. New York: Basic Books.Zahran, S., Brody, S. D., Vedlitz, A., Grover, H., & Miller, C. (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. D., & Vedlitz, A. (2008). Risk, stress and capacity: Explaining metropolitancommitment to climate protection. Urban Affairs Review, 43, 447–474.