Unpacking the intensity of policy conflict: a study of Colorado s oil and gas subsystem

Transcription

1 DOI /s RESEARCH NOTE Unpacking the intensity of policy conflict: a study of Colorado s oil and gas subsystem Tanya Heikkila 1 Christopher M. Weible 1 Springer Science+Business Media New York 2017 Abstract This article applies the Policy Conflict Framework (PCF) to describe and explain the characteristics of policy conflict within the oil and gas subsystem in Colorado. We use data from a survey of policy actors to assess three cognitive characteristics of policy conflict: divergence in policy positions, perceived threats from opponents positions, and an unwillingness to compromise. Aggregating these indicators across policy actors in the subsystem, we find a moderately high level of policy conflict intensity, but we also find substantial variation in the characteristics of policy conflict across policy actors. To help explain this variation, we examine how interpersonal and intrapersonal attributes of policy actors relate to the characteristics of policy conflict. In particular, we find that insular policy actor networks, interest group affiliations, and rigidity of risk and benefit perceptions associate more consistently with conflict characteristics than political views, education, or experience. We conclude with a discussion of the strengths and limitations of this first application of the PCF and reiterate the need for theoretically and empirically rigorous measures of policy conflict. Keywords Policy conflict Oil and gas development Hydraulic fracturing Policy process Policy theory Introduction Policy processes are rife with conflicts. Yet, policy process theories typically treat conflict as a background concept. While policy scholars attend to related phenomena, such as belief differences, polarized networks, and political mobilizations, they pay limited attention to & Tanya Heikkila tanya.heikkila@ucdenver.edu Christopher M. Weible chris.weible@ucdenver.edu 1 School of Public Affairs, University of Colorado Denver, 1380 Lawrence Street, Suite 500, Denver, CO 80207, USA

2 the conceptual development and theoretical importance of conflict. As a result, a theoretical and empirical void exists with respect to its causes, definition, and consequences. Moreover, researchers often assume conflict exists in a policy setting without direct measurement of its underlying characteristics. This article explores policy conflict using the Policy Conflict Framework (PCF): a framework designed to advance generalized and localized knowledge about policy conflicts, including its sources and effects, as well as its variations over time (Weible and Heikkila 2017). We apply the PCF to study a conflict within a specific policy subsystem: oil and gas development in Colorado. Debates over oil and gas development appear pervasive across much of the USA, particularly with the growth in hydraulic fracturing and horizontal drilling techniques (Rabe and Borick 2013; Boudet et al. 2014; Heikkila and Weible 2016). Research on policy conflicts over oil and gas development has explored the nature of the positions of political proponents and opponents, such as their perceptions of the environmental risks or the benefits to the economy (Smith and Ferguson 2013; Cotton et al. 2014; Cook 2014; Jaspal and Nerlich 2013; Heikkila et al. 2014; Heikkila and Weible 2016). Scholars have also explored the strategies and interactions of policy actors across various countries as they try to influence policy outcomes related to oil and gas (Ingold et al. 2016). Despite attention to policy conflicts on this issue, none of the literature has explored empirically what characterizes these policy conflicts across policy actors within a policy subsystem. This article therefore asks (1) What are the characteristics of policy conflict occurring in the oil and gas subsystem in Colorado; and (2) how do attributes of policy actors help explain policy conflict characteristics within the oil and gas subsystem? To answer these research questions, this article uses survey data from policy actors involved in the oil and gas subsystem in Colorado in Policy actors are people actively involved in influencing policy processes on a given issue. They may be affiliated with governments, nonprofits, the private sector, consulting and academia, citizen-based organizations, and the news media. While various methodological approaches would allow us to analyze policy conflicts, we start with observations of policy actors because of their expertise and sustained involvement in attempting to influence the policy processes that ultimately shape societal outcomes (Sabatier 1991). At the same time, studying the nature of policy conflict among policy actors within a specific subsystem provides a way to aggregate data to understand the intensity of subsystem-level conflict. This article begins with an overview of the PCF and then develops hypotheses to help explain the characteristics of conflict in the context of oil and gas policy. It then describes the oil and gas subsystem in Colorado and the methods of data collection. The results and discussion sections demonstrate the applicability of the PCF, as well as the need for building and testing contextually relevant hypotheses from the PCF. We conclude with the strengths and limitations of this application and reiterate the need for theoretically and empirically rigorous measures of policy conflict. Policy Conflict Framework The scope of the PCF is to understand and explain the cognitive and behavioral characteristics of episodes of policy conflict, the attributes of the policy setting that are associated with policy conflict episodes, and the feedback effects of outcomes and outputs on the policy setting (Weible and Heikkila 2017). PCF makes three key assumptions. First, policy conflict episodes occur across three interacting levels of action: political system, policy

3 subsystem, and policy action situations. Second, the model of the individual builds on other policy theories and cognitive psychology. It argues that individuals face cognitive constraints through selective attention, biased assimilation, and emotional conditioning. That is, individuals will selectively attend to information that comports with their beliefs, will remember losses more than gains, and will use emotions and heuristics to guide decisions (Kahneman and Tversky 2013; Jenkins-Smith et al. 2014). Individuals tend to identify positively with members of their in-groups and negatively with members of out-groups (Tajfel et al. 1971; Billig and Tajfel 1973). Third, depending on the intensity and effects, policy conflicts can have both healthy and unhealthy consequences at any level of action. This article applies a subset of the PCF. It focuses on the relationship between the attributes of policy actors within the policy setting (in this case, the oil and gas subsystem in Colorado) and the cognitive characteristics of policy conflict. Other aspects of the policy setting (i.e., events, attributes of different levels of action), as well as the behavioral characteristics and feedbacks, will be explored in future research. (See the PCF flow diagram below, where the subset of features of interest to this study is in bold.) As shown in Fig. 1, the PCF pivots around episodes of policy conflict. Episodes are divided into cognitive and behavioral components. The cognitive characteristics of policy conflicts include three dimensions: divergence in policy positions, threats from opponents policy positions, and an unwillingness to compromise on policy positions. 1. Divergence in Policy Positions The first dimension is defined as differences between policy actors on what government should or should not do in relation to a societal issue (Tilly and Tarrow 2007; Jenkins-Smith et al. 2014). Most policy conflicts feature interacting layers of divergences in policy positions. 2. Threats from Policy Positions This second dimension relates to the extent that policy actors perceive threats to themselves or society from the policy positions of their opponents (Tilly and Tarrow 2007; Cobb and Elder 1972). Underlying this dimension is the argument that policy conflicts are unlikely to exist or persist if policy actors do not feel threatened by their opponents position. This dimension of policy conflict builds on the assumption that policy actors respond more to losses than gains (Kahneman and Tversky 2013; Jenkins-Smith et al. 2014), which leads policy actors to perceive the policy positions of opponents as threatening. 3. Unwillingness to Compromise on Policy Positions The third dimension is unwillingness to compromise by policy actors on policy positions. Policy conflicts would not exist if policy actors were willing to compromise on their policy positions. Some theories emphasize that divergence in policy positions can be mitigated through belief change (Jenkins-Smith et al. 2014). PCF suggests that in the absence of belief change, Fig. 1 PCF flow diagram with key variables for this study in bold (adapted from Weible and Heikkila 2017)

4 a willingness to cooperate with opponents in reaching a negotiated solution can overcome divergence in policy positions (e.g., see Susskind et al. 1999). The cognitive characteristics of policy conflict exist at the individual level in the minds of policy actors. Yet, individual-level characteristics can be aggregated to help understand conflict characteristics and their intensity across political systems, policy subsystems, or policy action situations. Intensity, however, cannot be assessed by looking at the average perceptions across a population of policy actors within a given level of action. Instead, the distributions of the perceptions of policy actors should be examined to uncover the variation in the level of intensity of policy conflicts. Some policy actors involved in policy conflicts may perceive high levels of the conflict attributes as described above. Other policy actors may perceive low levels on these dimensions. A more intense policy conflict exists when a majority of policy actors expresses divergence in policy positions, perceives threats from opponents positions, and are unwilling to compromise. 1 Variation in policy conflict characteristics, and thus the intensity of conflict, will condition and be conditioned by the policy setting. The attributes of the policy setting identified in the PCF include (1) attributes of the level of action (e.g., institutional, socioeconomic, physical conditions), (2) interpersonal and intrapersonal policy actor attributes, (3) policy issue attributes, and (4) event attributes. This study seeks to understand the variation in cognitive conflict characteristics based on attributes of policy actors involved in the subsystem. We do not analyze attributes of the policy issue, as we focus on a single issue domain and, therefore, control for issue variance. Second, we do not include events in the analysis, because we zoom into a particular snapshot in time. Below we stipulate several hypotheses that relate intrapersonal and interpersonal attributes of policy actors to the cognitive characteristics of policy conflict. Hypotheses The PCF posits that intra- and interpersonal policy actor attributes shape policy conflicts. From the PCF, the intrapersonal attributes examined in this article include deep core beliefs, policy-relevant knowledge, and perceptions of risks and benefits. Not all attributes are expected to be important in all policy conflicts and sometimes their direct effect on policy conflict cannot be asserted a priori. In defining the first intrapersonal attribute, deep core beliefs, we rely on the Advocacy Coalition Framework (Jenkins-Smith et al. 2014). Deep core beliefs are fundamental normative or ontological orientations and include, but are not limited to, ideology, identities, fundamental interests, and cultural orientations. We conceptualize deep core beliefs for this study via a traditional right left measure of political ideology. For deep core beliefs, we do not have a hypothesis about how liberal or conservative ideology shapes perceptions of policy conflict on this issue. We simply expect that political ideology will condition the characteristics of policy conflict. Policy-relevant knowledge is the second intrapersonal policy actor attribute. This attribute captures policy actors experience, expertise, and training in relation to the policy issue. Policy-relevant knowledge also represents the potential for cognitive biases associated with expertise and education. For this research, policy-relevant knowledge includes 1 Additionally, the dimensions of policy conflict may or may not be correlated. For example, a policy actor could be divergent in their policy positions but may perceive low threats from their opponents and be willing to compromise.

5 the diversity of experience with oil and gas issues and formal education. We expect that a diversity of experience will be associated with less divergent policy positions on oil and gas development, fewer perceived threats, and more willingness to compromise. We hypothesize that the lower the diversity in experience of policy actors on issues related to oil and gas development, the more policy actors will exhibit policy conflict characteristics. We do not posit a relationship for level of formal education. Perceptions of risks and benefits are the third intrapersonal policy actor attribute. Policy issues, whether they involve choices about the provision of public goods and services, protection or use of natural resources, or public health and safety, entail various degrees of societal benefits or risks. Perceptions of risks and benefits associated with a particular policy action or issue can affect conflict characteristics of policy actors. For instance, when policy actors perceive levels of risks or benefits that diverge from other policy actors, they may take more divergent positions or feel more threatened and unwilling to compromise. This article conceptualizes risks and benefits in two ways: (1) perceived relative balance of risks and benefits, and (2) perceived rigidity of risk and benefit perceptions. We hypothesize that policy actors who perceive more risks relative to benefits, and the more rigid they are in those perceptions, the more policy actors will exhibit policy conflict characteristics. In addition to intrapersonal attributes, we explore how interpersonal attributes of policy actors relate to policy conflict. A key interpersonal attribute of policy actors identified in the PCF is their networks. Past studies have demonstrated that policy actors become involved in advocacy coalitions to influence public policies and achieve their goals in policy conflicts (Jenkins-Smith et al. 2014). However, these advocacy coalitions emerge from a broader network base of relations among policy actors that are informal or formal, not necessarily political, possibly overlapping or part of other policy subsystems, and developed over time, sometimes through past policy conflicts (McAdam et al. 1996; Tilly and Tarrow 2007). Policy actors with more isolated network relations are less likely to be exposed to diverse information sources and different perspectives. Combined with mental processes of biased assimilation, these policy actors are more likely to take divergent positions on issues, more likely to feel threatened by those with opposing positions and less likely to compromise on their positions. We hypothesize that policy actors with more insulated policy actor networks area more likely to exhibit policy conflict characteristics. PCF identifies organizational affiliations as another key interpersonal attribute. Organizations attract certain types of policy actors and provide direction for both conceptual and behavior characteristics of conflicts (Jenkins-Smith et al. 1991). We do not posit a general hypothesis that is applicable across all policy conflicts for the association between organizational affiliations and the characteristics of policy conflict. Instead, in the context of oil and gas development, we hypothesize that individuals affiliated with organizations that have taken advocacy positions (e.g., environmental groups or industry associations) will exhibit higher levels of conflict characteristics than non-advocacy-oriented organizational affiliations (e.g., government and academia). Study context The level of action for this application of the PCF is the policy subsystem related to oil and gas development using hydraulic fracturing in Colorado. Oil and gas development using hydraulic fracturing began to expand rapidly in Colorado around Between 2007 and 2013, crude oil production in Colorado rose 146% and natural gas production rose 38%,

6 making Colorado one of the largest producers of energy from oil and natural gas in the USA (US Energy Information Administration 2014a). Most of the gas and oil development since 2007 has occurred in the Denver-Julesberg Basin in northeast Colorado (US Energy Information Administration 2014b), which is close to Colorado s major metropolitan areas. The expansion of oil and gas development has resulted in a few attempts by local governments to ban hydraulic fracturing, particularly in urban communities near shale deposits. Yet, most instances of policy change have occurred at the state level. The state Oil and Gas Conservation Commission made several major changes to the state s oil and gas regulations between 2008 and 2014 in response to the growth of the industry. Further policy debates, especially related to the authority of local governments in governing hydraulic fracturing, have arisen in the state legislature, and discussed through a Governor s Task Force in 2014 and early While Colorado had 65 percent fewer new wells drilled in early 2016 than in early 2015, production from existing wells has continued to climb and political debates are ongoing (Hood 2016). For example, a lawsuit against two local communities that attempted to either ban hydraulic fracturing or impose a moratorium was brought to the Colorado Supreme Court in December of 2015, and in 2016 the Court ruled that these actions violated the preemption clause of the state constitution. In early 2016, several citizen initiatives were proposed for the November ballot in Colorado, including one that would have given local governments more control over oil and gas development, which industry countered with media campaigns. However, none of the initiatives gained enough signatures to reach the ballot. In early 2017, Colorado s Attorney General sued Boulder County, which has had a moratorium on fracking since 2012, arguing that the moratorium violated state law. Also, a number of local advocacy groups have formed in recent years to protest industry proposals to drill new oil and gas wells within their communities. The policy conflict, which remains pervasive in Colorado, raises questions about its underlying characteristics. Data and methods In studying the subsystem-level conflict around oil and gas development in Colorado, we conducted a survey in 2015 of policy actors. We first conducted nine interviews of policy actors from diverse backgrounds to inform the design of the survey and provide background information on the research context. The survey was administered by through Qualtrics, an online survey platform. The survey population included 453 individuals actively involved or knowledgeable about oil and gas development in Colorado. These individuals were identified using a purposive sampling approach based on evidence in media reports, online reports, public hearings and testimony, and recommendations from interviewees. 2 Two hundred thirteen people responded to the survey yielding a 47% response rate. 3 Not all respondents chose to answer every question, so response rates vary 2 The initial target list of respondents was 630 individuals. After eliminating bounced s from the list and individuals who were not actively involved in the issue, the final population was The response rates by organizational affiliation are local government (60 of 127 = 47%), industry (51 of = 41%), environmental nonprofits (31 of 61 = 51%), state government (21 of 28 = 75%), legal professionals (17 of 34 = 50%), organized citizen groups (9 of 18 = 50%), university/consultants (12 of 30 = 40%), industry nonprofits (6 of 12 = 50%), other nonprofits (5 of 9 = 56%), media (1 of 6 = 17% response rate), federal government (0 of 4 = 0%), and other (0 of 1 = 0%).

7 by question. Consistent with PCF, the survey questions were designed to capture indicators of the characteristics of policy conflicts, attributes of the policy setting, and effects of conflicts. 4 Operationalization of cognitive characteristics of policy conflict We used the survey data to develop and explore indicators that represent the three cognitive characteristics of policy conflict identified in PCF: divergent policy positions, perceived threats, and unwillingness to compromise. To measure divergent policy positions, we used our survey question that asked respondents to report on a scale of 1 5 what comes closest to their current policy position on oil and gas development that uses hydraulic fracturing: stop, limit, continue at the current rate, expand moderately, or expand extensively (1 = stop; 5 = expand extensively). To calculate the measure, we subtracted the mean score of this scale (2.96) from each respondent s position score and took the absolute value. Higher values, thus, equate with beliefs divergent (or extreme) from the average. The second cognitive dimension is perceived threats from opponents policy positions. The indicator for perceived threats came from a survey question that asked: To what extent do the views and actions of those you disagree with on oil and gas development that uses hydraulic fracturing (a) threaten you personally or professionally; and (b) threaten the state of Colorado? Both response categories were on a scale of 1 5 (1 = not at all; 5 = a great deal). To represent the threats concept, we add the values of these two response categories for a combined threats score for each individual (ranging from 2 to 10). The indicator for the third dimension of policy conflict characteristics, unwillingness to compromise, was derived from a battery of questions on the survey that asked respondents (on a scale of 1 5; 1 = strongly disagree; 5 = strongly agree) if they would support government decisions that would significantly limit hydraulic fracturing under several hypothetical conditions, or if they would support government decisions that would significantly limit hydraulic fracturing in Colorado under several conditions. For the expand questions, the battery of hypothetical conditions included six items, such as scientific evidence shows it is completely safe to the environment or public health or Colorado regulators passed and enforced stricter regulations. For the limit question, the battery of hypothetical conditions also included six items, such as convincing scientific evidence shows it is a significant threat to the environment or public health or a catastrophic disaster or emergency occurred from oil and gas development using hydraulic fracturing. Response categories were on a 5-point scale, from strongly disagree to strongly agree. We reversed the scale so 5 equals strongly disagree and 1 equals agree for the unwillingness measure. To create a single value for each respondent s unwillingness to compromise based on this battery, we selected respondents who preferred to stop or limit fracking (from the previous survey question) and calculated each respondent s mean response to the expand battery. We then selected respondents who preferred to expand moderately or expand extensively and calculated each of these respondents mean value from the limit battery. For respondents who reported a policy preference for continuing at the current rate, we averaged the Likert score values from both batteries (limit and expand). The resulting score for unwillingness to compromise therefore ranges 4 Access to the full survey instrument is available from the authors.

8 from 1 to 5 (5 = high unwillingness), and is tied to an individual s initial policy preference. As is depicted in the PCF, we expect these three policy conflict characteristics to interrelate, whereby the combinations of higher levels of each will be associated with higher conflict intensity. Therefore, we created a conflict composite index to represent the combination of these intersecting dimensions. Because the operational measures are not on the same scale, and we have no expectation that one variable should carry more weight than the other, we rescaled the variables before combing them. We created a common scale of 1 3 for each variable (1 = low, 2 = medium, and 3 = high levels of the conflict characteristic). Values that were 1 standard deviation below the mean and lower were equal to 1. Values one standard deviation above the mean and higher were equal to 3, and those between one standard deviation below and one standard deviation above were equal to 2. We then multiplied the scores for each respondent across the dimensions to represent a conflict composite index. 5 The distribution of this composite index for all individuals in the subsystem was assessed to explore the aggregate intensity of conflict across the subsystem. Operationalization of policy actor attributes associated with conflict The second goal of this article is to analyze how the attributes of policy actors help explain the variance in policy conflict among the actors in the policy subsystem. Therefore, we examine the relationships between the cognitive characteristics of policy conflict and the intrapersonal and interpersonal attributes of the policy actors. Under the intrapersonal attributes category, we measured political ideology by asking respondents to report whether they identify as extremely liberal, liberal, moderate, conservative, or extremely conservative (on a scale of -2 = extremely liberal;?2 = extremely conservative). Intrapersonal attributes also include two measures for policy-relevant knowledge. To create a variable for respondents diversity of experience with oil and gas issues, we used a survey question that asked respondents to rank their level of experience with eight different fields or areas of expertise (on a scale of 1 4: 1 = No experience; 4 = A lot of experience). The eight areas included a variety of types of expertise such as researching technical issues, analyzing financial issues, planning or working on oil and gas operations, regulating, owning mineral rights, living near oil and gas operations, and engaging in political activity to influence government on oil and gas. The final variable for the analysis was an additive score of each of the different types of experiences across the eight areas. For level of education, we asked respondents to identify their highest degree on a 5-point scale, from high school degree, some college, bachelor s degree, master s degree, to PhD/ MD/JD. We have two indicators for the third intrapersonal attribute, risk and benefit perceptions. The first indicator is the level of perceived risks relative to perceived benefits. For this indicator, we used a battery of questions that asked respondents to rank on a 5-point scale (1 = strongly disagree; 5 = strongly agree) the extent to which they agree or disagree that 5 These three dimensions form a conflict composite index that combines individual respondents scores on the three dimensions of the cognitive characteristics of conflict along a concord conflict spectrum. Unlike typical measures of latent variables where correlated items in the scale are expected to load onto an abstract concept, we conceptualize the concord conflict spectrum as comprised of three dimensions that might or might not be correlated. For example, perceptions of threats might or more not be correlated with divergent policy positions.

9 eight different items are potential problems (i.e., insufficient capacity for regulation, degradation of air quality) related to oil and gas development that uses hydraulic fracturing. We asked a companion question about the extent to which respondents agree or disagree that seven different items are potential benefits (i.e., national energy security, job creation). We then took the average ranking of the risks for each respondent and subtracted the average ranking of benefits as the measure of perceived risks relative to benefits. The second indicator is rigidity in perceptions of risks and benefits. For this question, we used survey questions that asked respondents whether they have become less concerned, stayed the same, or more concerned, about the risks and then about the benefits of oil and gas development (-1 = less concerned; 0 = same;?1 = more concerned). Then, we created a single variable that is tied to the position variable by assigning a respondent who wants fracking stopped or limited the value for their rigidity measure on the risks, and respondents who want fracking expanded the value of their score on the benefits. For respondents who had reported wanting oil and gas development to continue at the current rate, we averaged the score of their position on risk and benefit rigidity questions. Under the interpersonal attributes category, we have indicators for network insularity and organizational affiliation. Our measure of network insularity came from our survey question that asked respondents to rank the importance of their interactions with various types of groups in achieving their personal or professional goals related to oil and gas development (on a scale of 1 to 5; 1 = not at all important and 5 = very important). To create the insularity variable, we took the combined degree of importance of interactions with the two industry groups on the battery, subtracted the combined score for the interactions with environmental groups and citizen groups, and then took the absolute value of this measure. So a low score is interpreted as balanced interactions and a high score can be interpreted as more insular interactions. To measure organizational affiliation, we coded the organization types associated with respondents contact information collected during the sampling process. For the purposes of this analysis, we coded respondents in three categories: (1) affiliated with private industry or industry association; (2) affiliated with environmental organizations or citizen groups; and (3) affiliated with government organizations or others. In the models that follow, we include industry affiliations and environmental/citizen group affiliations as variables in the model. Government/other affiliations are the baseline group. Results In analyzing the data, we first describe the cognitive characteristics of policy conflict. Prior to diagnosing how the policy actor attributes are associated with these conflict characteristics, we provide the descriptive statistics on the cognitive characteristics and analyze the composite index of these characteristics. (Descriptive statistics for all of the variables are provided in the Appendix.) Cognitive characteristics of policy conflict Table 1 presents the descriptive statistics for the cognitive characteristics of policy conflict, which include divergent policy positions, perceived threats, and unwillingness to compromise. The mean scores for each of these characteristics lie at the midpoint of the measurement scale. However, the standard deviation for each of these scores indicates substantial variation

10 Table 1 Descriptive statistics for cognitive characteristics of policy conflict in Colorado s oil and gas policy subsystem Cognitive characteristics Valid cases Minimum Maximum Mean SD Divergent policy position Perceived threats Unwilling to compromise within the sample with respect to these characteristics with some policy actors perceiving high levels of conflict and others perceiving low levels of conflict. 6 As previously noted, in creating the combined variable, we rescaled each of the individual characteristic variables to a 1 3 scale (1 = low; 2 = moderate; 3 = high) for each of the three indicators of conflict characteristics. We then multiplied those rescaled variables to represent the conflict composite index as a measure of conflict intensity for each individual policy actor, which ranges from 1 to 27 in value (median = 8). In looking at the distribution of the composite index across the policy actors in the oil and gas subsystem in Colorado (see Fig. 2), we find variation in the level of conflict intensity. Given a value of 8 is the multiplicative moderate value for the three dimensions, the distribution in Fig. 2 indicates that 43% of respondents fall on the lower end of the spectrum (\8), 21% of respondents view conflict of moderate intensity (=8), and 36% of respondents fall on the higher end of the spectrum ([8). However, an examination of extreme scores on both ends of the spectrum shows a slightly different portrayal of the data. That is, a sizeable proportion of respondents fall on the far high end of the conflict composite index with scores of 12, 18, and 27 (34%). This proportion exceeds the proportion of respondents who fall on the far low end of the conflict composite index (or the high concord end) with scores of 1, 2, and 3 (12%). As a result, the distribution is positively skewed toward higher levels of conflict intensity (skewness = 1.21; standard error of skewness = 0.178). Explaining variation in conflict characteristics with policy actor attributes To explore variation across the individual actors involved in the policy conflict, we examine how the intra- and interpersonal attributes of policy actors relate to the cognitive conflict characteristics in the case of oil and gas development in Colorado. We ran four separate ordered logistic regression models, regressing each of the five dependent variables on the measures of the policy actor attributes (see Table 2). Since the dependent variables are not continuous, as they are based on Likert scale survey responses, ordered logit was deemed most appropriate for the data. 7 The odd ratios with standard errors in the 6 The three cognitive characteristics of policy conflict also interrelate. Spearman s Rho correlations show significant correlations (p \ 0.000) with correlations between 0.36 and This suggests that policy actors are more likely to feel threatened and unwilling to compromise when they have divergent policy positions. It remains an empirical question whether such correlations will also be found in other studies or the implications of such correlations on the dynamics of policy conflicts. 7 Tests for collinearity and outliers were also done using the ordinary least square models. Mean variance inflation factor (VIF) values for each of the models were under 2. Tests for influential cases included assessing the Cook s distance score and DFBETA scores for each of the variables. No cases on the five models exhibited Cook s distance values greater than 1, or absolute value scores of the DFBETA s greater than 1. As more conservative measures, cases with absolute values of the DFBETA scores exceeding 2/sqrt(n), or Cook s distance scores greater than 4/n were evaluated (n = 168). Several cases in each model marginally exceeded the conservative thresholds. For the purposes of this research, we chose not to drop

11 Fig. 2 Distribution of policy actors conflict composite index Percent of Respondents 25% 20% 15% 10% 5% 0% Conflict Composite Index Table 2 Ordered logistic regression models associating policy actor attributes with conflict characteristics Divergent policy position Odds ratio (SE) Perceived threats Odds ratio (SE) Unwilling to compromise Odds ratio (SE) Conflict composite index Odds ratio (SE) Intrapersonal attributes Political ideology 0.75 (0.18) 1.01 (0.23) 0.66* (0.14) 0.65* (0.14) Diversity of experience 1.02 (0.03) 1.04 (0.03) 0.97 (0.03) 1.02 (0.03) Level of education 0.68** (0.12) 0.07 (0.16) 0.89 (0.13) 0.90 (0.14) Perceived risks relative to benefits 0.85 (0.10) 1.15 (0.11) 0.79** (0.08) 1.05 (0.11) Rigidity of risk benefit perceptions Interpersonal attributes a 5.56*** (1.87) 2.02** (0.29) 1.57 (0.44) 3.18*** (0.93) Insular networks 1.22*** (0.09) 1.20** 1.36*** (0.10) 1.35*** (0.10) (0.09) Industry affiliation 2.22* (0.97) 4.97*** 1.71 (0.64) 3.50*** (1.46) (0.20) Env. or citizen group 2.04* (0.87) 2.09* (0.92) 2.68** (1.18) 2.26* (0.97) affiliation Number of obs Log likelihood LR v 2 (10) 66.80*** 50.61*** 39.1*** 72.28*** Pseudo R a Baseline organizational affiliation is government or other ; *** p \.01; ** p \.05; * p \.10 parentheses are reported. The ordered logit models were compared with ordinary least square regression models with similar results in terms of significant variables and model robustness. Footnote 7 continued cases in the models, as the data arguably are within the range of the design of the survey instrument that we used to collect the data. All models were also run with and without the cases with Cook s D scores that exceeded 0.5 (1 case in the level of government model; 1 case in the unwillingness to compromise model; and 2 cases in conflict composite index model) and removal of these cases did not change any of the significant variables identified in the results, or the direction of signs.

12 Among the intrapersonal attributes, more conservative political ideology decreases the odds for conflict characteristics across three of the four the models, and is significant in the unwillingness to compromise and conflict composite index. Specifically, the odds of a respondent being unwilling to comprise, or of having a higher composite index score, are 0.66 lower and 0.65 lower respectively for unit increases toward a more conservative political ideology. Conversely, respondents who report more liberal ideology are more likely to be unwilling to compromise and have higher composite index scores. 8 The two variables that represent policy-relevant knowledge generally are not significant in the models. First, we do not find that a diversity of experience around oil and gas development issues is a significant explanatory factor for the cognitive characteristics of conflict, which is counter to our hypothesis. For the level of formal education, we find that it is significant in only one model. That is, a one-unit increase in higher education levels reduces the odds of policy actors taking divergent policy positions by Also under the intrapersonal attributes, the variable representing perceived risks relative to benefits shows significant results in only one of the models. Specifically, an increase in the level of risk perception decreases the odds (0.79) of expressing more unwillingness to compromise, which we did not expect. However, the variable for rigidity of risks and benefit perception is significant, increasing odds for the dependent variables in three of the four models. These show that increased rigidity amplifies the odds for higher levels of divergent policy positions (5.56), perceived threats (2.02), and for the conflict composite index (3.18), as hypothesized. In looking at the variables representing the interpersonal characteristics, we find strong support for our hypotheses. First, insular networks are associated with higher odds for divergent policy positions (1.22), perceived threats (1.20), unwillingness to compromise (1.36), and the conflict composite index (1.35). Second, industry affiliation increases the odds of higher levels of divergent policy positions (2.22), perceived threats (4.97), and conflict composite index (3.50), relative to those respondents with government and other affiliations. Industry affiliation is not significant, however, in increasing the odds for unwillingness to compromise. Being affiliated with an environmental organization is associated with increasing the odds of higher levels of divergent policy positions (2.04) relative to government/other affiliations, as well as increasing the odds of higher levels of perceived threats (2.09), unwillingness to compromise (2.68), and conflict composite index (2.26). Conclusion This article utilized the PCF to explore the cognitive characteristics of policy conflict in the oil and gas subsystem in Colorado, and to explain examine how policy actor attributes associate with cognitive conflict characteristics expressed by the policy actors in this subsystem. In the analysis of policy actor responses, the data demonstrate that the conflict over oil and gas development in Colorado cannot be represented monolithically, or simply as 8 We also ran models with a variable for extreme political views in lieu of political views, which was created by taking the absolute value of the political view variable. While the extreme political view variable was significant and positive in the combined model, it was not significant in any of the individual models and the robustness of the models was not significantly improved. The coefficients were all positive, however, across all models.

13 highly contentious. The cognitive conflict characteristics of policy actors divergent policy positions, perceptions of threats, and unwilling to compromise vary across a subsystem. We found more policy actors with higher levels of these conflict characteristics than very low levels (or high concord), while a substantial proportion of policy actors have moderate levels of these conflict characteristics. We also found general support for the hypotheses we developed, which relate several policy actor attributes identified in the PCF to conflict characteristics. Under the category of intrapersonal attributes, rigidity of risks and benefits were important explanatory variables for conflict characteristics, but simply having higher risk relative to benefit perceptions was not associated with higher levels of conflict characteristics. However, the ways in which specific variables play out in the analysis may be contingent on the policy setting. For instance, higher perceptions of risks relative to benefits decreased the odds for an unwillingness to compromise, which was not expected. That is, in the case of oil and gas policy conflict, people who perceive more risks may be more willing to compromise. Yet, we also found that policy actors with more conservative ideology also are more likely to be willing to compromise. Interpersonal characteristics namely insular networks and organizational affiliations appear to be more consistently and more strongly associated with conflict characteristics than intrapersonal characteristics. As researchers have indicated that in-group and out-group affiliations can play a more important role than ideology in political polarization (Iyengar et al. 2012), these results, therefore, are compatible with this larger body of the literature. In addition to the insights about policy conflicts related to oil and gas issues, this article contributes to the policy process literature by being the first to apply the PCF empirically. The purpose of the PCF is to guide the analysis of policy conflicts and identify general patterns of relationships. We do not expect the specific patterns of relationships identified in our data on oil and gas conflicts in Colorado to be representative of other conflicts. Policy conflicts are contingent on the policy setting. Policy conflicts also involve a temporal dimension as they evolve. Over time and with enough case studies, we might be able to establish a formal set of hypotheses that are applicable in some contexts but not others. Deriving this formal set of hypotheses is beyond the scope of this article, though the hypotheses offered in this article represent an inaugural move in that direction. In this effort, the next step is to conduct further analyses of policy conflicts based on this approach across different topics or over time on the same topic. It is equally imperative to analyze patterns of policy conflicts across different types of policy settings, such as in policy action situations within a subsystem, or to compare how subsystem-level conflicts relate to conflicts across broader political systems. We recognize that this article is not without limitations. In addition to all the imperfections of a cross-sectional survey (e.g., endogeneity), this article does not apply the entire PCF. Another next step is to analyze how the cognitive characteristics of conflict shape the behavioral characteristics of conflict. This can be followed by an analysis of the feedback effects. Finally, given PCF has layered causal arguments, a structural equation model of the factor categories, characteristics of conflict, and feedback effects is needed. At the same time, comparisons of conflicts across multiple levels of action can allow for testing of how attributes of those levels (i.e., institutions) matter. Building from the PCF, this article focuses on a conceptualization and operationalization of the cognitive characteristics of conflict. Such an emphasis is distinct from other public policy studies that tend to analyze phenomena such as agenda setting, policy change, and implementation (Sabatier and Weible 2014). Closely related to PCF, the Advocacy Coalition Framework might focus on an analysis of politics by analyzing the beliefs and network patterns of coalitions, but it has yet to provide a conceptual definition

14 of policy conflicts for analyses and comparisons (Jenkins-Smith et al. 2014). At the same time, the PCF may be compatible with other theoretical approaches. We view the PCF as a framework that focuses on the relations that constitute the process within policy process research, with potential to inform our understanding of other phenomena (e.g., policy change or agenda setting) that are more conventional in this literature. Policy conflicts are ubiquitous across politics but also understudied. In part, this is because scholars have equated policy conflicts with divergent beliefs, political mobilization, or polarized network formations. Such interpretations of policy conflicts are not necessarily wrong but rather underdeveloped. Until we articulate the quiddities of policy conflicts and connect the cognitive and behavior components, any analysis of policy conflict may overlook important attributes and misinterpret policy conflicts. The purpose of this article is to provide an initial attempt to identify, measure, and parse the policy conflict concept. Acknowledgements This research was supported by the AirWaterGas Sustainability Research Network funded by the National Science Foundation under Grant No. CBET-1,240,584. Any opinion, findings, and conclusions or recommendations expressed in this article are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. Appendix See Table 3. Table 3 Descriptive statistics Dependent variables Valid cases Minimum Maximum Mean SD Divergent policy position Perceived threats Unwilling to compromise Conflict composite index Scaled explanatory variables Valid cases Minimum Maximum Mean SD Political view (-2 = extremely liberal; ?2 = extremely conservative) Diversity of experience in oil and gas Level of education Perceived risks relative to benefits Rigidity of risks and benefits Insular networks Dummy explanatory variables Valid cases N % Industry affiliation Env. or citizen group affiliation Government or other affiliation

15 References Billig, M. G., & Tajfel, H. (1973). Social categorization and similarity in intergroup behavior. European Journal of Social Psychology, 3(1), Boudet, H., Clarke, C., Bugden, D., Maibach, E., Roser-Renouf, C., & Leiserowitz, A. (2014). Fracking controversy and communication: Using national survey data to understand public perceptions of hydraulic fracturing. Energy Policy, 65, Cobb, R. W., & Elder, C. D. (1972). Participation in American politics: The dynamics of agenda-building. Baltimore, MD: The John Hopkins University Press. Cook, J. J. (2014). Who s regulating who? Analyzing fracking policy in Colorado, Wyoming, and Louisiana. Environmental Practice, 16, Cotton, M., Rattle, I., & Van Alstine, J. (2014). Shale gas policy in the United Kingdom: An argumentative discourse analysis. Energy Policy, 73, Heikkila, T., Pierce, J. J., Gallaher, S., Kagan, J., Crow, D. A., & Weible, C. M. (2014). Understanding a period of policy change: The case of hydraulic fracturing disclosure policy in Colorado. Review of Policy Research, 31(2), Heikkila, T., & Weible, C. M. (2016). Contours of coalition politics in the United States. In C. M. Weible, T. Heikkila, K. Ingold, & M. Fischer (Eds.), Debates on hydraulic fracturing: Comparing coalition politics in North America and Europe (pp ). New York, NY: Palgrave Macmillan. Hood, G. (2016). Fight over drilling could spill onto November ballots. March 9, Colorado Public Radio. sthash.cqkm1cgf.dpuf. Ingold, K., Fischer, M., Heikkila, T., & Weible, C. M. (2016). Assessments and Aspirations. In Christopher M. Weible, Tanya Heikkila, Karin Ingold, & Manuel Fischer (Eds.), Policy debates on hydraulic fracturing: Comparing coalition politics in North America and Europe (pp ). New York: Palgrave Macmillan. Iyengar, S., Sood, G., & Lelkes, Y. (2012). Affect, not ideology a social identity perspective on polarization. Public Opinion Quarterly, 76(3), Jaspal, R., & Nerlich, B. (2013). Fracking in the UK press: Threat dynamics in an unfolding debate. Public Understanding of Science, 23, Jenkins-Smith, H., Nohrstedt, D., Weible, C. M., & Sabatier, P. A. (2014). The advocacy coalition framework: Foundations, evolution, and ongoing research. In Paul A. Sabatier & Christopher M. Weible (Eds.), Theories of the policy process (3rd ed., pp ). Boulder, CO: Westview Press. Jenkins-Smith, H., St. Clair, G., & Woods, B. (1991). Explaining change in policy subsystems: Analysis of coalition stability and defection over time. American Journal of Political Science, 35, Kahneman, D., & Tversky, A. (2013). Prospect theory: An analysis of decision under risk. Econometrica, 47, McAdam, D., McCarthy, J. D., & Zald, M. N. (Eds.). (1996). Comparative perspectives on social movements. Cambridge, MA: Cambridge University Press. Rabe, B. G., & Borick, C. (2013). Conventional politics for unconventional drilling? Lessons from Pennsylvania s early move into fracking policy development. Review of Policy Research, 30, Sabatier, P. A. (1991). Political science and public policy.. PS. Political Science & Politics, 24(02), Sabatier, P. A., & Weible, C. M. (2014). Theories of the policy process (3rd ed.). Boulder, CO: Westview Press. Smith, M. F., & Ferguson, D. P. (2013). Fracking democracy : Issue management and locus of policy decision-making in the Marcellus Shale gas drilling debate. Public Relations Review, 39, Susskind, L., McKearnan, S., & Thomas-Larmer, J. (Eds.). (1999). The consensus building handbook: A comprehensive guide to reaching agreement. Thousand Oaks, CA: Sage Publications. Tajfel, H., Billig, M. G., Bundy, R. P., & Flament, C. (1971). Social categorization and intergroup behavior. European Journal of Social Psychology, 1(2), Tilly, C., & Tarrow, S. (2007). Contentious politics. New York: Oxford University Press. U.S. Energy Information Administration. (2014a). Colorado state profile and energy analysis. U.S. Energy Information Administration, Washington, DC. U.S. Energy Information Administration. (2014b). Natural gas gross withdrawals and production. U.S. Energy Information Administration, Washington, DC. dcu_nus_m.htm. Weible, C. M., & Heikkila, T. (2017). Policy Conflict Framework. Policy Sciences. (Early View). doi: /s