Shaping the Scope of Conflict in Scotland’s Fracking Debate: Conflict Management and the Narrative Policy Framework Pre-publication version (4 Nov 2019) of article published in Review of Policy Research

: This study applies a narrative lens to policy actors ’ discursive strategies in the Scottish debate over fracking. Based on a sample of 226 newspaper articles (2011 – 2017) and drawing on key elements of the narrative policy framework (NPF), the research examines how policy coalitions have characterized their supporters, their opponents. and the main regulator (Scottish government). It also explores how actors have sought to expand or contain the scope of conflict to favor their policy objectives. Empirically, only the government strives for conflict containment, whereas both pro- and anti-fracking groups prioritize conflict expansion through characterization contests and the diffusion and concentration of the costs/risks and benefits of fracking. In theoretical terms, the study proposes that insights from Pral le’s (2006) conflict management model, which emphasizes symmetrical strategies of conflict expansion by both coalitions, is a potential tool to revise extant NPF expectations about the different narrative strategies of winning and losing coalitions. Moreover, the fact that policy actors mostly employ negatively rather than positively framed characters in their narratives may be a valid expectation for similar policy conflicts, particularly under conditions of regulatory uncertainty.


Introduction
The 'shale revolution' in natural gas production, reliant on hydraulic fracturing technology (or 'fracking'), began in the US in the late 2000s. In the absence of extensive federal regulation, states and provinces across the US and Canada have chosen a variety of regulatory paths, ranging from moratoria to moderate regulations to permissive rules (Carter & Eaton, 2016). Such diversity of regulatory approaches has also become a global pattern, and Europe has emerged as a region where precautionary trends and moratoria have been especially pronounced. For instance, France, Germany, Luxembourg, Bulgaria, and all devolved administrations in the United Kingdom have instituted moratoria or bans on the technology (Gamper-Rabindran, 2018; Van de Graaf, Haesebrouck, & Debaere, 2018).
Comparative analyses have evaluated numerous factors linked to permissive and precautionary regulation of fracking. Structural variables -such as mining law and mineral rights, the relative abundance of shale gas reserves, the significance of fossil fuels for the national economy, import dependence, and the institutional environment (planning law, multi-level vs. centralized regulation) -are deemed important (Keeler, 2016;McGowan, 2014). But factors related to political agency -such as governments' political orientation and the relative power of the fossil fuel industry and environmental groups -may be equally decisive in shaping the regulatory outcome (Bomberg, 2017a;Keeler, 2016;Van de Graaf et al., 2018). Scholars have convincingly argued that the most critical elements in the development of restrictive regulations may be the effectiveness of anti-fracking campaigns and the level of public concern in a given jurisdiction (Neville et al., 2017;Van de Graaf et al., 2018).
It follows that the chief purpose of anti-fracking movements is influencing public opinion. Above all, this means turning fracking into a salient issue on the political agenda, encouraging policy-makers to respond with reassuring messages and/or more restrictive regulations. Alongside the mobilization of resources, enabling access to policy-makers and the media, and the exploitation of political opportunities, such as regulatory failures or elections, an important skill of movements is the use of "compelling framing" (Piggot, 2018, p. 942). The resonance of frames and the perceived effectiveness of narrative strategies have become an important focus for the literature on contests between pro-and anti-fracking groups (Bomberg, 2017b;Evensen, 2018;Neville et al., 2017). While these studies reveal numerous tailored framing strategies, they share a concern with thematic analysis of frames and counter-frames, examining which frames are over-or underrepresented, and which frames will be most resonant with broader audiences. Many studies demonstrate, in effect, the impact of a successful expansion of the scope of conflict, even if most do not use this language 1 . Originally theorized by Schattschneider (1975), a narrow scope of conflict favors established lobby groups and often entrenches the status quo, while an expansive scope of conflict attracts new audiences and may decisively alter the outcome of a policy conflict. Based on a longitudinal content analysis of four Scottish newspapers published between May 2011 and December 2017, this article employs insights from the scope of conflict and agendasetting to pursue two objectives: (1) to examine debates over shale gas policy in Scotland, a devolved region of the United Kingdom, and (2) to contribute to the development of the narrative policy framework (NPF).
Scotland's permanent moratorium on the technology (imposed in October 2017) rests on an ambiguous structural context and requires consideration of political agency. In energy terms, Scotland exhibits a 'dual identity', being a major producer of offshore oil and gas, while also hosting a substantial, and rapidly growing, renewable energy industry. The power of incumbent fossil fuel industries and their contribution to public finances and employment militated against a moratorium. Conversely, extraction of shale gas could potentially undermine Scotland's ambitious climate policy targets, aiming to achieve net-zero greenhouse gas emissions by 2045.
The empirical contribution of this article consists of exploring what the narrative strategies of political actors reveal about the Scottish debate over fracking. Informed by framing and narrative studies of conflict over fracking in Scotland (Stephan, 2017) and in other contexts (Bomberg, 2017a;Evensen, 2018;Heikkila, Pierce, Gallaher, Kagan, Crow, & Weible, 2014;Zanocco, Song, & Jones, 2018), I examine some of the main narrative elements used by policy actors, show how both coalitions have sought to shape the scope of conflict, and shed light on the government's role in conflict containment.
Furthermore, a key contribution of this article is theoretical. In contrast to the frame-based literature, the narrative policy framework focuses on the structural elements of actors' narrative interventions. Narratives have common elements, such as characters, which can be tracked over time, revealing the narrative strategies of competing coalitions of actors. NPF scholars have explored a variety of policy conflicts over the last decade (Gupta, Ripberger, & Collins, 2014;McBeth, Shanahan, Arnell, & Hathaway, 2007;McBeth, Shanahan, Hathaway, Tigert, & Sampson, 2010;Shanahan, Jones, McBeth, & Lane, 2013). Yet, their attempts to account for the scope of conflict and for the narrative strategies of winning and losing coalitions have often remained tentative. The recent victory of the anti-fracking coalition in Scotland offers an opportunity to revisit these aspects. Drawing on insights from Pralle's (2006) conflict management model and previous NPF studies, I propose two different sets of expectations. While NPF scholars have often focused on differentiating the narrative strategies of winning and losing policy coalitions, insights from the conflict management model suggest symmetrical strategic behavior by both coalitions to expand the scope of conflict in their preferred direction. These revised expectations are only partly confirmed in this study, but the results suggest their potential validity and should prompt further research. There are also noteworthy trends about the campaign tone of the coalitions, as defined by the use of supportive and antagonistic narrative characters (the devil-angel shift). Investigating these patterns advances our understanding of how different coalitions deploy narrative strategies to shape the scope of conflict through characterization contests and through the narrative diffusion and concentration of costs/risks and benefits of fracking. It also offers future research avenues for potentially linking regulatory trends (e.g. regulatory certainty vs. uncertainty) to narrative strategies.
I proceed by first introducing the empirical case study of Scotland. Then, after reviewing insights provided by the scope of conflict and the conflict management model, I outline key elements of the NPF, explain the methodology, and summarize and discuss the main findings. The core arguments as well as implications for the NPF and the wider literature on fracking are presented in the concluding section.

The Case: Fracking in Scotland
While Scotland has limited powers in the area of energy policy, the Scottish government has the ability to permit or deny fracking operations through its planning process (Cairney, McHarg, McEwen, & Turner, 2019;Shapovalova, 2018). Moreover, exploration licenses for onshore oil and gas exploration were devolved by the UK government in February 2018. Drawing on an existing account of the controversy (Stephan, 2017) and on regional (Scottish) newspapers, a survey of key political events and policy decisions (Table 2) allows me to separate the controversy into three distinct phases (Table 1): (1) Emergence and mobilization; (2) Stalemate; (3) Endgame and permanent ban. Identification of the three phases 2 in the Scottish case study rest on a careful interpretation of the political dynamics, specifically the movement from relative regulatory certainty (phase 1) to uncertainty (phase 2) and back to more certainty (phase 3). Each shift is marked by the changing stance of the main regulatorthe Scottish government and the governing Scottish National Party (SNP). The three phases approximately coincide with key policy decisions, such as the moratorium. This qualitative interpretation reflects the changing discourse of the government and the SNP. The observed shifts are either preceded by momentous political events (e.g. the 2014 independence referendum) or motivated by looming elections and competition with Scottish Labour (e.g. in spring 2016). In turn, this pattern helps to designate stronger (winning) and weaker (losing) coalitions during phases 1 and 3. The rationale behind this is further explained in the section on expectations. sceptical' of fracking Apr 2016 ----------Phase 3 (endgame and permanent ban) begins ----------Apr 2016 SNP election manifesto vows that fracking will not be permitted unless it can be proven beyond any doubt that it is not harmful to public health and the environment Apr 2016 IPSOS-MORI poll (Taylor, 2016) finds fracking is no. 6 (out of 11) on the list of voters' priorities for the Scottish Parliament 2016 election In phase 1, Scottish anti-fracking groups were galvanized by a combination of factors, including the growing opposition in the US (Mazur, 2016), early discussions across Europe about shale gas exploration (McGowan, 2014), and applications for local planning permission for test drilling. However, the shale gas industry was in a promising (winning) position during this phase. The Scottish government was not opposed to exploring new extractive possibilities, initially declaring that "[u]nconventional gas offers huge potential as long as development and use is consistent with environmental objectives" (Sunday Herald, 29 May 2011). Resistance was primarily voiced in relation to specific local projects. It took some time for anti-fracking demands to escalate from more stringent planning rules to a moratorium or ban. This demonstrates the gradual growth in strength and confidence of the anti-fracking movement.
Phase 2 was initiated in October 2014, in the aftermath of the Scottish Independence Referendum (18 Sep 2014). It quickly transpired that the governing Scottish National Party (SNP) would pursue a more cautious fracking policy -a stance partly motivated by the opportunity to compare itself favorably to the (at the time) openly pro-fracking UK government in Westminster. Furthermore, with fracking continuing to grow in salience and with important elections approaching in 2015 and 2016, a moratorium was announced in January 2015. Under growing pressure by activists, the Scottish Greens, and the Scottish Labour Party, the moratorium was intended to provide some respite for the Scottish government and the governing SNP. Similar to  analysis of the moratorium in New York State, the associated regulatory uncertainty meant that no clear winning coalition can be determined for this phase. The Scottish government invoked an ostensibly neutral 'evidence-based approach', while both coalitions strove to persuade policymakers and the public of their respective policies -exploratory drilling with safeguards vs. restrictive designs ranging from local referendums to a national ban on fracking.
Phase 3 began with the publication of the SNP's election manifesto (April 2016) which reiterated skeptical comments made by Scotland's First Minister Sturgeon. She challenged proponents of fracking to prove there would be "no risk to health, communities or the environment" (Daily Record, 21 April 2016). In the same month, the salience of fracking for the public was highlighted by an IPSOS-MORI poll listing the most important political issues for Scottish voters (Taylor, 2016). Although a number of research reports and a public consultation were still to come, the SNP's manifesto firmly signaled the ascendancy of antifracking positions and marked out the associated movement as likely winners. The permanent moratorium announced in October 2017 was followed by a legal challenge brought by industry which, however, was rejected in June 2018.

The Scope of Conflict and Conflict Management
The Scottish case represents a politically salient and dynamic policy dispute that has carried on for a considerable period of time. It serves to illuminate the notion of the 'scope of conflict' which originates from Schattschneider's (1975) work on the strategic expansion and containment of political conflict. What Schattschneider considered to be 'normal' politics consists of a tight control over the definition of the policy problem by a limited number of actors. This has also been termed a policy monopoly (Baumgartner & Jones, 1993) and is more likely to occur in policy areas that have relatively low salience and are not subject to regular public debate. Schattschneider (1975) described such insulated political arenas as vulnerable to the 'mobilization of bias'. Policy-makers would be less accountable to public opinion and would come under intense pressure from organized interest groups. Opening up such closed arenas -dubbed the 'socialization of conflict' -will spark the 'contagion' of political conflict and thus widespread public debate. Such conflict expansion will strongly influence the nature and number of participants or audiences, the arguments presented, the ideas and values invoked and, ultimately, the outcome of the policy debate.
There is evidence that highly salient and controversial political topics are associated with policy outcomes that correspond more closely with prevailing public opinion (Burstein, 2003;Page & Shapiro, 1983;Shapiro, 2011). Some issues, such as 'morality politics' (Haider-Markel & Meier, 1996), may be reliably salient by their nature, but the majority of policy issues will have to be actively politicized to enlarge the scope of conflict. Linking an issue with overarching cultural values -such as liberty, justice, or freedom of speech -can be a powerful recipe for making conflict "contagious" and for delivering policy changes (Schattschneider, 1975, p. 7). Groups who believe they will benefit from conflict expansion will work towards this end, particularly by using the mass media to involve new actors and to influence public opinion. Conversely, those interested in preserving a closed or more stable policy subsystem will seek to contain the scope of conflict. The media, in particular, has proven influential in setting the agenda for both policy-makers and the public (McCombs, 2004). It represents the "ultimate means of conflict expansion" (Brown, 2002, p. 136).
Schattschneider's pioneering work has also informed the literature on agenda-setting and issue definition. Cobb and Elder (1983) theorized the movement of policy issues from the general 'systemic' agenda to the 'institutional' agenda which preoccupies policy-makers. Conflict expansion, often through issue (re)definition, is the primary means through which the size of the audience is increased and the elevation of an issue can be accomplished. However, other political actors may use the countervailing strategy of conflict containment, frame the issue very narrowly, and attempt to shrink the audience. In conjunction, these seminal contributions from earlier decades laid out a productive research agenda which, to this day, continues to generate valuable insights (Jones & Baumgartner, 2005;Jones, 2016;Pralle, 2006).
Conflict expansion and containment have been integrated into core NPF hypotheses about narrative strategies (Shanahan, Jones, McBeth, & Radaelli 2017) and some NPF studies have closely examined the associated political dynamics (Gupta et al., 2014;McBeth et al., 2007). While they have found considerable evidence of conflict expansion, containment strategies have received less attention. The present article confirms this pattern of activity for pro-and anti-fracking groups and additional insights from the agenda-setting literature can furnish an explanation. Davis and Hoffer (2012), for instance, have applied such a perspective on 'agenda denial' to U.S. fracking regulation. Of particular importance in the Scottish debate over fracking are attempts at conflict containment made by the ostensibly neutral Scottish government rather than by pro-fracking groups. 'Symbolic placation' involves relatively modest actions to avoid discussion of more radical policy solutions or it seeks to undermine opponents' momentum by postponing the debate (Cobb & Ross, 1997, p. 34ff.).
An important reason why conflict containment has received less attention in NPF studies is that, beyond the initial stages of a policy conflict, efforts at conflict containment are likely to be overshadowed by efforts at conflict expansion. A sufficiently large audience has now become involved and will likely drive efforts at conflict expansion by both the incumbent and the challenger coalition. This type of 'mature' policy conflict, in which expansive narrative contestation becomes a regular occurrence, is not fully represented by Schattschneider's (1975) original conceptualization which reflects "an enduring and static structure of competition" (Pralle 2006, p. 221). Yet, it is captured by Pralle's (2006) model of conflict management which brings together many of the insights generated by the agenda-setting literature. Her framework recognizes that issue (re)definition continues well beyond the initial stages of agenda-setting and conflict initiation. Actors in policy conflicts are more likely to "muddle through" under conditions of "ongoing conflict and competition [… which] complicate the strategic choices for advocacy groups" (Pralle 2006, p. 221). Such prolonged and dynamic discursive contests between coalitions of policy actors arise when containment strategies have failed to close down the debate.
As the political salience of an issue increases, and as all coalitions pursue strategies of issue expansion, Pralle (2006) highlights several dimensions to be considered: (1) linkages with other policy issues, including the use of culturally resonant storylines; (2) constructing boundaries, often in a geographical sense (e.g. from local to national or global level) (3) problem ownership, that is, authoritative and compelling issue definition and redefinition; (4) characterization contests (naming and shaming through narrative elements); and (5) the formation of organizational and discursive coalitions.
In operationalizing the narrative scope of conflict through their analysis of narrative strategies, NPF scholars have consistently built on Schattschneider's (1975) conceptualization. Pralle's work (2006) is cited by many NPF studies, but its implications for analyzing policy disputes deserve to be explored more extensively, beginning with an overview of the NPF.

The Narrative Policy Framework (NPF)
Created by a group of scholars in the 2000s (McBeth et al., 2007;McBeth, Shanahan, & Jones, 2005), the NPF investigates the role of narratives in policy disputes. In part, it is inspired by contributions from scholars in the post-positivist tradition, and narrative policy analysis in particular (Roe, 1994;Schön & Rein, 1994;Stone, 2012). But the NPF offers a more systematic approach to assessing how narratives are deployed in policy conflicts and how they may shape public opinion and policy outcomes. Systematically examining policy narratives also allows NPF scholars to measure the policy beliefs held by different actors. However, for the purpose of this article, the NPF's first function -the identification of policy narrative strategies employed by policy actors -is of primary importance. The NPF adopts a structuralist conception of narratives: they are not unique to particular situations and contexts, but can be distilled into common elements. Fully-fledged policy narratives consist of several narrative elements: First, the setting describes the policy context and can involve assumptions about geographical conditions, political institutions, the economic situation, cultural patterns, etc. (Shanahan, Jones, & McBeth, 2018;Shanahan, Jones, McBeth, & Radaelli, 2017). Second, a variety of characters appear in policy narratives, such as heroes, villains, and victims. They exhibit some intentionality or are subject to the actions of other characters. Third, a plot acts as a mechanism to connect all narrative elements. It is often interpreted through generic plotlines defined by Stone (2012). Fourth, the moral of the story (or policy solution) involves a purposive action to improve the situation or delivers a verdict on a policy idea.
While these elements are important building blocks of policy narratives, this article concentrates on the use of characters by opposing groups. An exploration of narrative characters is frequently at the core of NPF studies and this can largely be explained by research findings gathered at the micro level of analysis. At this level, NPF research examines the impact of narratives on individual and public attitudes, often through survey techniques. For instance, Jones (2014) concludes that affect for hero characters positively influences US respondents' policy preferences on climate policy. Specifically on fracking, Zanocco, Song, & Jones (2018) find that the use of villain characters shapes the US public's attitudes to fracking policies.
Given the significance of characters for the NPF, building on McBeth et al's (2005) work, Merry (2016) deploys a broader cast of characters to create a more nuanced typology. 'Opponents' are less strongly criticized than villains or are forcefully encouraged to endorse a particular policy solution. 'Allies' are actual or potential supporters. They are not praised in the same way as heroes, but have some potential to turn into heroes. For instance, the Scottish government is at times constructed as an ally in policy narratives. But when the government's decisions are challenged, for instance in favor of a complete ban on fracking, the characterization is coded as an opponent. To enable comparisons with previous NPF studies, I later create two new aggregate categories: supporters (heroes and allies) and antagonists (villains and opponents).

Narrative Characters and the Devil-Angel Shift
A key narrative strategy that can be linked to Pralle's (2006, p. 25) notion of 'characterization contests' is the 'devil-angel shift'. It denotes the discrepancy in the use of heroes and villains as characters in policy narratives (Shanahan et al., 2013). Originally drawn from the advocacy coalition framework (Sabatier, Hunter, & McLaughlin, 1987), the devil shift is a rhetorical strategy to describe members of the opposing coalition as particularly malevolent and powerful (i.e. villains). It is a basis for naming and shaming tactics, for rallying a group's supporters, and for demoralizing the opponents. By contrast, the angel shift describes the "glorification" of members of one's own coalition by designating them as hero characters (Merry, 2017, p. 2). It is designed to reassure and motivate coalition members.
Winning coalitions are assumed to use the angel shift more frequently, whereas losing coalitions will tend to use the devil shift. However, so far the results of existing studies have been inconsistent. While Schlaufer's (2018) and Merry's (2016and Merry's ( , 2017 research on gun policy debates in the US have found little statistical difference between the coalitions. Such variation makes the devil-angel shift an area ripe for repeated testing. Based on Pralle's (2006) notion of 'characterization contests', I follow her argument that "groups who want to expand conflict vilify enemies to align supporters with the goals of the movement." While hero characterizations appear especially suitable for maintaining and motivating one's own coalition, the above findings from micro-level analysis suggest that, in particular contexts, a variety of character types may function to enhance the persuasiveness of policy narratives and to gain wider support in society. Hence, actors' differential use of characterizations may ultimately derive from a strategic decision to adopt a negative or positive campaign tone in a policy conflict (Schlaufer, 2018, p. 107). There is a large body of extant research on messaging in electoral campaigns (Pattie, Denver, et al., 2011;Fridkin & Kenney, 2012), but actors in ongoing policy disputes are equally likely to deliberately shape their narratives to maximize the impact. I will therefore use the notion of negative and positive campaign tone as shorthand for the relative strength of the devil or angel shift.

The Narrative Scope of Conflict
While there are grounds to associate an increased use of characters with the expansion of conflict, NPF scholars have also operationalized the latter concept in different ways. First, following Stone's (2012) observation that coalitions strive to be seen to defend a recognized public interest rather than narrow special interest, scholars postulate that winning groups will highlight diffuse, widely shared benefits and only admit concentrated costs or risks that affect few, if any, actors. Conversely, losing groups will emphasize widespread costs or risks, while only acknowledging concentrated gains for few, if any, actors (Gupta et al., 2014;McBeth et al., 2007). The nature of benefits and costs/risks is not always clarified in policy narratives on fracking, but the vast majority of these instances relates to economic or environmental and health issues. Importantly, several NPF studies have found empirical evidence for a consistent pursuit of conflict expansion and containment. Gupta, Ripberger, & Wehde (2018, p. 132) find that coalitions' narrative strategies regarding nuclear power in the US differ in "theoretically predictable ways", and similar results are reported for nuclear policy in India (Gupta et al., 2014). These findings confirm earlier studies that identified significant differences between winning and losing coalitions in the controversy over buffalo conservation in the Greater Yellowstone Area in the US (McBeth et al., 2007;McBeth et al., 2010). Winning groups will work to stabilize their own coalition, demobilize the opposition, preserve the status quo, and thus contain the scope of conflict (Shanahan, Jones, & McBeth, 2011). Their opponents will construct a divisive 'loser's tale' that aims to expand their coalition and galvanize both new and existing supporters (McBeth et al., 2010).

Expectations
Besides the empirical value of the Scottish case, I have identified several theoretical areas that would benefit from further investigation, including the use of narrative characters, the devil-angel shift, and the narrative scope of conflict. The overarching purpose is to explore whether insights from the conflict management model can be successfully integrated with existing NPF suppositions and inform future research. In line with the conflict management model, given the propensity of competing groups to engage in issue expansion and to battle "on the same rhetorical turf" (Pralle, 2006, p. 6), I expect both coalitions to use relatively symmetrical narrative strategies and to predominantly focus on conflict expansion. Previous research on gun policy debates in the US has detected a similar narrative pattern (Merry, 2016, p. 390;Merry, 2017, p. 18). This may be because the 'socialization' of the conflict (Schattschneider, 1975) is already in full swing once the news media has stepped up its reporting on the issue in 2011. Efforts to contain or minimize conflict over an already salient issue may therefore be overshadowed by efforts at conflict expansion.
First, because strategies of conflict expansion can be linked to characterization contest (Pralle 2006), one should expect a relatively frequent use of narrative characters.
Expectation 1a: Both coalitions will use narrative characters with similar frequency. Heroes (and allies), villains (and opponents), and victims will be used by both coalitions for the purpose of conflict expansion -to persuade and galvanize a wider audience.
Second, I follow McBeth et al. (2007) and Gupta et al. (2014) on NPF-based analysis of conflict expansion and containment. The two opposing coalitions will seek to expand the scope of conflict in a direction that favors their objectives, while trying to contain it in areas that are detrimental to their cause. Thus, pro-fracking groups will seek to expand the scope of conflict by highlighting widely diffused benefits, while anti-fracking groups will seek to contain these aspects by stressing the concentration or absence of such benefits. Furthermore, the antifracking coalition will seek to expand the scope of conflict by emphasizing the costs/risks, whereas the pro-fracking coalition will seek to contain these aspects by underscoring the concentration or absence of such risks. However, drawing from Pralle's (2006) conflict management model, from Merry's (2016; study of US gun policy, and from empirical findings by  about fracking policy in New York State, I expect conflict expansion strategies to be used more frequently than conflict containment strategies.
Expectation 2a: Rather than pursuing containment strategies, both pro-and antifracking coalitions will primarily invoke benefits and costs/risks to expand the scope of conflict. The pro-fracking coalition will mainly focus on the diffusion of benefits. Conversely, the anti-fracking coalition will mainly focus on the diffusion of costs/risks.
The above expectations contrast with some of the hypotheses developed and tested by NPF scholars in which the status of winning and losing coalitions plays a key role. To ascertain this status, earlier NPF studies relied on subjective signaling through policy narratives created by policy actors themselves (i.e. self-portrayal) (McBeth et al., 2007;McBeth et al., 2010). However, this could potentially distort the results if policy actors deliberately present themselves in a way that does not reflect their actual level of political influence Merry, 2018). Other NPF studies have therefore generalized about winners and losers by considering policy outcomes (Crow & Berggren, 2014;Gupta et al., 2018;. But this approach may overlook the dynamic nature of a policy conflict in which "winners and losers may trade places over time" (Merry, 2018, p. 750).
In a recent comparison of NPF hypotheses, Gottlieb, Oehninger, & Arnold (2018, p. 799) examine conflicts over fracking policy in different New York municipalities and cast doubt on the centrality of the "winning-losing dichotomy". In their study, a simplified version of the original NPF approach (self-portrayal as winning or losing) produces the expected results: the devil shift is mainly used by losers and the angel shift by winners. But when coding winner and losers by overall policy outcome, the results reveal a devil shift by both coalitions. The authors conclude that, by going beyond the winning-losing dichotomy, a distinction between anti-fracking narratives and pro-fracking narratives yields the most consistent results: profracking actors use the angel shift, while anti-fracking actors display a significant devil shift.
As outlined earlier, the present article is based on three clearly differentiated phases of the Scottish policy conflict over fracking. This allows for a more dynamic approach focused on the position of the main regulator for fracking (Scottish government) and it takes account of degrees of regulatory un/certainty. Whereas the pro-fracking coalition is in a stronger (winning) position in phase 1, during which the government remains cautiously supportive, the anti-fracking coalition is stronger (winning) in phase 3. A stalemate reigns in phase 2 under the temporary moratorium which allowed the Scottish government to consider the scientific evidence and the political risks of permitting shale gas development.
Therefore, a second set of expectations is designed to connect with the results of extant NPF studies about winning and losing coalitions. It assumes that they pursue markedly different narrative strategies. Winning coalitions generally use fewer narrative components in their statements, feel less compelled to expand the conflict, have more interest in containment and stabilization, and adopt a more positive campaign tone (angel shift). Comparing the two sets of expectations enables me to draw conclusions about previous NPF findings and to evaluate whether the conflict management model may be used to revise NPFrelated hypotheses.
Expectation 1b: As conflict expansion is more crucial for losing coalitions than for winning coalitions, they are likely to employ more characters than winning coalitions. Expectation 1c: As a negative campaign tone is likely associated with losing coalitions, they will use more antagonistic characters and perform a devil shift. Winning coalitions will adopt a positive campaign tone, use comparatively more supporting characters, and will perform an angel shift.
Expectation 2b: Winning coalitions will more frequently promote conflict containment than conflict expansion, while losing coalitions will more frequently promote conflict expansion than conflict containment.

Methodology
Previous meso-level NPF analysis often relied on content analysis of documents produced by policy actors, such as public consumption documents (Gupta et al., 2014;McBeth et al., 2007;Shanahan et al., 2013), internal newsletters (Smith-Walter, Peterson, Jones, & Reynolds Marshall, 2016), and social media messages (Gupta et al., 2018;Merry, 2016). But several studies have also examined newspaper articles (Blair & McCormack, 2016;Crow et al., 2017;Crow & Lawlor, 2016;Gottlieb, Oehninger, & Arnold, 2018;Huda, 2018;Kear & Wells, 2014;Shanahan, McBeth, Hathaway, & Arnell, 2008). For some of the key policy actors in Scotland's fracking debate, neither social media nor public consumption documents reach back to 2011. In this article, I therefore focus on narrative components in newspaper articles. To an extent, this choice complicates the analysis. Not only is the content of newspapers shaped by writers and editors, but the stories told in newspaper articles also often do not include all the narrative elements defined by the NPF (Crow et al., 2017;Huda, 2018). Rather, articles typically feature relatively short statements by one or several actors. Nevertheless, this drawback is partly counterbalanced by the considerable investigative scope offered by newspaper archives, covering many years of policy debate, numerous political actors, and mirroring the dynamic, shifting nature of narrative contestation. Moreover, newspapers play a crucial role for coalitions seeking to expand the scope of conflict (Brown, 2002), and they help to set the agenda for communication strategies devised for other media (Crow et al., 2017, p. 634).
The Scottish newspapers selected for content analysis do not display substantial political bias and are therefore expected to summarize or quote a comparable amount of statements from both pro-and anti-fracking policy actors. Among the chosen newspapers -The Herald, The Scotsman (incl. its Sunday edition), The Sunday Herald, and the Daily Record & Sunday Mail -only the last paper can be reliably categorized as left-of-center on the political spectrum. It also carries fewer articles on fracking than other daily newspapers, which limits possible bias in the overall dataset. Furthermore, the four Scottish newspapers address a national (Scottish) rather than primarily local audience, and traditionally concentrate on Scottish issues. Using the keywords 'shale gas', 'fracking', and 'hydraulic fracturing' for searching on the Nexis database, newspaper articles were gathered for all years under consideration. Once a policy timeline had been ascertained, specific time periods were chosen (purposive sampling) to both yield a sufficient amount of articles and to ensure that key policy events were included, such as important policy decisions, elections, or the publication of reports (see Table 3).
Search results (n = 1004) were manually filtered to remove duplicates and to meet several criteria, yielding a total of 226 articles for content analysis. Articles had to relate directly to the Scottish debate over fracking and had to contain statements (or summaries of statements) by policy actors that indicated a clear policy position which would allow for their categorization into one of the two coalitions. In accordance with NPF conventions, in addition to a clear policy stance they had to include a recognizable narrative character (Shanahan et al., 2013). These inclusion criteria screened out the majority of articles, including most newspaper editorials, but retained most opinion pieces and letters to the editor. The coalitions identified in this manner match the 'discourse coalitions' summarized by Stephan (2017): environmental NGOs, the Green Party, and local groups constitute the bulk of the anti-fracking coalition and are progressively joined by other political parties. The pro-fracking coalition mainly consists of oil and gas companies, the Conservative Party, and later the GMB trade union. Opting for broadly defined discourse coalitions of actors has the advantage of capturing a plethora of interventions by actors with rhetorically similar policy objectives. It also leads to a more complete picture of narrative dynamics in the fracking debate.
While scholars have theorized the mass media as a potential contributor to policy debates in its own right (Shanahan et al., 2008), this study treats newspapers as a conduit, that is, an approximate record of actors' narrative interventions. As articles were analyzed chronologically, the principal coder ensured that, if several newspapers reported on the same statement by a policy actor, its narrative components were only coded once (i.e. in the first instance). Statements were thus required to be original, and their merely repetitive presence in different newspaper articles did not skew the results. This was also checked when testing for intercoder reliability. Coding was performed at document level. Thus, each specific narrative component (e.g. the anti-fracking coalition labelling environmental/local groups as a hero) was only recorded once per document. But this coding practice still allows for the counting of several characters of the same category in the same newspaper article. For example, in addition to portraying environmental/local groups as a hero, the anti-fracking coalition might also describe the Scottish Greens as a hero.  (Crow & Berggren, 2014;Gupta et al., 2014;McBeth et al., 2007;Schlaufer, 2018;Shanahan et al., 2018;Shanahan et al., 2013;Smith-Walter et al., 2016). The complete codebook can be found in the appendix. Descriptions of characters are located in the earlier section on the NPF. Heroes and allies as well as villains and opponents are differentiated by qualitative nuances that are subject to interpretation. An example of an ally is petrochemicals company INEOS declaring that it still has "confidence" in the Scottish government's evidence-based approach despite its "hardening […] stance against unconventional oil and gas" (Herald, 28 May 2016). An example of an opponent is when antifracking groups challenge the Scottish government to adopt a ban on shale gas: "The Scottish Government needs to stand up to this as it has the powers to put the brakes on fracking" (Scotsman, 19 August 2014). As practiced by most NPF scholars, (e.g. Shanahan et al., 2013), the measurement of the devil-angel shift is based on counting the instances of hero and villain characters through a simple formula: (number of heroes) -(number of villains) / (total number of heroes and villains). This produces a result on a continuum between -1 (devil shift) to 1 (angel shift). To enable comparison with previous findings by other scholars, I retain this simplified representation. The original concept of the devil shift also sought to incorporate the perceived power of opponents, but has proved difficult to operationalize Merry, 2017). It is common for studies on public policy to concentrate on either the 'evilness' or the 'power' dimension of the devil-angel shift (e.g. see Katz, 2018;Vogeler & Bandelow, 2018, p. 719).
Coding for conflict expansion focuses on narratives that diffuse either the costs/risks of fracking or its benefits. In the latter case, actors would, for instance, claim that "fracking was key to tackling fuel poverty" (Daily Record, 26 January 2015). Coding for conflict containment refers to narratives that concentrate either the costs/risks of fracking or its benefits. The former case also encompasses assertions of safety (i.e. no genuine risk). Costs/risks and benefits are categorized as either undefined, economic, or health/environmental. Drawing on previous studies, deductive coding was used for characters and conflict expansion/containment.
Regarding the coding process, initially two coders independently processed the first 25 newspaper articles. They then reconciled their differences and updated the codebook accordingly. Overall intercoder reliability was tested by the second coder who examined a randomized subset (20 percent) of the whole dataset. This process confirmed adequate to good levels of reliability for each of the main narrative components, ranging from 81.6 percent for villains to 95.9 percent for benefits cited by anti-fracking actors.

Findings and Discussion
This section sets out the main findings of the study, both in quantitative measures and through qualitative discussion. Unless otherwise stated, the results below are rendered as mean numbers of narrative components per newspaper article, as this is the clearest representation of their overall presence. T-tests were carried out to establish statistical significance (see the p values below the figures).

Narrative Characters
To shed light on the use of characters in the Scottish fracking debate, I initially compare figures of the three main character types -heroes, villains, and victims (Figure 1). While the number of villains is relatively similar across coalitions, anti-fracking groups employ approximately three times more hero and victim characterizations (p < .01 or .05).  (2016), I also examine the frequency of opponents and allies (Figure 2). This reveals that the anti-fracking coalition designates four times more opponents than the profracking coalition (p < .01). To enable a comparison with the extant NPF literature, I also calculate the combined tally of supporting characters (heroes and allies = supporters) and the combined tally of antagonistic characters (villains and opponents = antagonists). The results show that, overall, the anti-fracking coalition uses approximately twice as many supporting and antagonistic characters as the pro-fracking coalition (Figure 3: p < .01). Considering aggregate character categories as the main comparator, expectation 1a for pro-and anti-fracking coalitions is disconfirmed, as the anti-fracking coalition is clearly more narratively engaged than the pro-fracking coalition (Figure 3: p < .01; significant differences between pro-and anti-fracking coalitions), at least through the medium of the Scottish press. To explore the second set of expectations, I count the number of narrative characters for winning and losing coalitions. Based on a timeline and the shifting discourse of the Scottish government and the SNP, I earlier identified winning and losing coalitions as trading places over time. Thus, the winning pro-fracking coalition in phase 1 becomes a losing coalition in phase 3. Adding up the narrative interventions for each coalition from both phases allows for a comparison with some earlier findings of NPF studies and also helps to paint a more complete picture of the policy controversy (Figure 4). A notable result is the greater use of heroes (in phase 3 and overall: p < .01) by the winning coalition. Figures 5 and 6 display additional characters and aggregate categories (supporters and antagonists). Compared to Figures 2 and 3, which display the use of these characters by proand anti-fracking coalitions, there is a strong equalizing effect, bearing in mind that winning and losing coalitions can only be designated for two out three phases.
Thus, expectation 1b about the use of characters by winning and losing coalitions is partly disconfirmed. The average frequency of narrative characters overall is relatively similar, even if winning groups deploy more supportive characters (Figure 6: p < .05). Generally, the relatively low use of supporters (i.e. heroes and allies) in this dataset contrasts with more frequent mentions of heroes in other NPF studies (e.g. Shanahan et al., 2013;Smith-Walter et al., 2016). This could be an artefact of newspapers' penchant for attributing blame and raising political tensions (Nie, 2003). Alternatively, in light of possible cultural differences (Gupta et al., 2014), it raises the question of whether heroes (and allies) might be used less regularly in settings outside the US where most NPF case studies have been located. The exploration of narrative characters also offers an opportunity to consider which specific policy actors are designated as characters and by whom. This part of the analysis is not directly linked to the two sets of expectations outlined earlier. However, firstly, it offers a valuable empirical comparison with extant literature on fracking; and, secondly, it can be read in conjunction with the later section on the narrative scope of conflict (expectation 2a) because it equally indicates a relatively symmetrical narrative pattern for both coalitions.
The most prominent Scottish policy actors are listed in Table 4 (in percentages). It is not surprising that the pro-fracking coalition often casts industry in a positive light and regularly presents environmental and local groups as antagonists. Conversely, anti-fracking groups commonly cast their counterparts as antagonistic characters. Regarding the most prominent antagonistic characters, the results show the same percentage (40%) for the only major 'neutral' policy actors -the Scottish government and the SNP. Verification of the detailed coding results allows me to conclude that much of the remainder (50-60%) designates members of the opposing coalition. This balanced pattern contrasts with research on the fracking debate in New York State. Here,  found that the antifracking coalition designates pro-fracking actors twice as often as villains (in percentages) than vice versa. Another important finding is that the Scottish government and, by extension, the governing SNP, are frequently targeted through narrative characterizations. This indicates strategic intent by the two coalitions who recognize the centrality of the government's decisionmaking power in this area. Beyond a ritualistic criticism of the opposite coalition, there is political -and party-political -acumen involved, as the ranking of the increasingly antifracking Scottish Labour Party demonstrates.

The Devil-Angel Shift
The interpretation of the devil-angel shift used by most NPF scholars compares the extent to which coalitions portray their adversaries as villains with the extent to which they identify themselves as heroes. To facilitate comparisons with previous studies, I consider the categories of supporters (heroes, allies) and antagonists (villains, opponents). As the Scottish government and the SNP were not part of either coalition (Stephan 2017), characterizations referring to them were discounted before calculating the scores below (Figures 7 and 8).
Although it is tempting to count the Scottish Government as part of the anti-fracking coalition in phase 3, this is contradicted by the government's frequent designation as an opponent by anti-fracking policy actors. This conforms to the original definition's focus on distinct policy coalitions and is key to understanding the trend over time. 3 Based on this definition, there is a strong but weakening devil shift by the anti-fracking coalition across the three phases (see Figures 7 and 8). By phase 3 it has turned a strong devil shift into a mild angel shift, matching its rising political fortunes.
Overall, the results offer a mixed response to expectation 1c about negative and positive campaign tone. While the winning coalition retains a mild devil shift overall, it performs a mild angel shift in phase 3 -compared to a devil shift by the losing coalition (Figure 8: p < .05). Regarding the wider NPF literature, this finding is not wholly consistent with studies revealing that winning coalitions perform an angel shift (Schlaufer, 2018;Shanahan et al., 2013). Nor does it unreservedly support NPF studies which find that the devil shift indicates losing coalitions (McBeth et al., 2010;Shanahan et al., 2011).
To move beyond the winning-losing dichotomy, Gottlieb et al's (2018) study of New York State municipalities attempts to link pro-fracking groups and the angel shift, on the one hand, and anti-fracking groups and the devil shift on the other. This pattern is not confirmed by the Scottish case. However, the results accord with studies that detect a widespread use of villains in policy narratives and directs our attention to the cumulative amount of characters being deployed (Merry, 2016(Merry, , 2017. It also aligns with   Pro-fracking Anti-fracking supposition that regulatory uncertainty promotes the devil shift, namely in phases 1 and 2 of the Scottish fracking debate. Given that the likely regulatory outcome is even more transparent by phase 3 than in phase 1, greater regulatory certainty may thus be linked to an increase in supporting characterizations -or less vilification -and therefore to the angel shift.

The Narrative Scope of Conflict
Efforts by the two coalitions to diffuse and concentrate benefits or costs/risks yield a relatively even picture. Table 5 highlights (in bold) the mechanisms through which coalitions reinforce their narratives by expanding or containing the scope of conflict. Confirming expectation 2a, both pro-and anti-fracking coalitions are more intent on diffusion tactics than on concentrating benefits or costs/risks (p < .01 or .05). Conversely, expectation 2b is disconfirmed because the switch from winning coalition (indicated by round brackets) to losing coalition [indicated by square brackets] and vice versa does not trigger a substantial realignment. Other relevant observations include the fact that coalitions' narrative strategies do not change significantly across the different phases. Moreover, considering the percentages of the main concerns stated, a strong specialization in types of benefits or costs/risks prevails, with a focus on either economic or health and environmental factors.  Finally, the minor importance of conflict containment in the above analysis allows for consideration of insights from the agenda-setting literature, as outlined earlier, especially regarding the crucial role of the main regulator. The Scottish government clearly pursued what Cobb and Ross (1997) would term a strategy of symbolic placation which can be categorized as conflict containment. Across the three phases, the government's statements grew ever more cautious. Based on the content analysis, Table 6 shows that the government's early interest in fracking (position 1) was soon displaced by feigning ignorance (position 2) and offering a moderately strengthened planning regulation (position 3). The government's mantra then became the evidence-based approach to fracking (position 4), but it too was sometimes watered down to an evidence-based informed approach that takes account of public opinion (position 5). By phase 3, the government's position was largely precautionary, vowing to proceed only if all risks could be excluded (position 6). Key decisions in phases 1 and 2, -strengthening planning law and later imposing a temporary moratorium -were designed to dampen the political conflict and reduce electoral risks. A second part of this strategy, also consistent with Cobb and Ross' (1997) notion of symbolic placation, was to postpone decisions by invoking an evidence-based approach and by commissioning an expertscientific panel (in 2013) and a series of detailed reports (in 2015). This ultimately ill-fated attempt to reassert control over the policy agenda by containing the scope of conflict was similar to the actions of governments in New York State (Katz, 2018) and Quebec (Montpetit, Lachapelle, & Harvey, 2016). The strategy's chances of success were further harmed by invoking more stringent Scottish fracking regulation for political gain -on the eve of the Scottish independence referendum (Sep 2014). This briefly disrupted conflict containment by emphasizing the contrast with a more 'gung-ho UK government at Westminster. But it can be considered a temporary, if damaging, exception to the quest for conflict containment.

Conclusion and Implications
In conclusion, the above analysis offers empirical insights into the Scottish policy dispute over fracking that equally have comparative value for the wider literature on fracking and other conflicts over energy policy. A mildly hospitable political context for shale gas extraction soon turned more challenging, and ultimately hostile, when anti-fracking actors used distinct narrative elements to target pro-fracking actors and, frequently, the Scottish government. Mainstream Scottish newspapers carried significantly more narrative characters deployed by anti-fracking policy actors. Both coalitions engaged in characterization contests by using narrative characters to expand the scope of conflict, but efforts by pro-fracking actors proved less effective and were further undermined by party competition over stricter regulation, both within Scotland and regarding the UK government. Thus, the narrative components and strategies I have reviewed represent important ways of shaping the scope of conflict. And they are implicated in the eventual policy outcome -a permanent ban on fracking. By enabling systematic identification and assessment, the NPF enhances our understanding of the dynamics of adversarial strategies in policy conflicts over fracking and beyond.
A key empirical observation is that conflict containment was comparatively less important than conflict expansion -even for pro-fracking actors (see Table 5). Once the debate had begun to gather pace, there were fewer incentives for delivering reassuring messages of safety or, for the anti-fracking side, of minor economic benefits. Containment narratives by pro-fracking actors were also weakened by the ongoing scientific debate over the health and environmental risks of fracking, which made it difficult to unreservedly guarantee its safety. However, the role of conflict containment cannot be fully grasped without considering the role of the main regulator, in this case the Scottish government, and the political danger posed by fracking as a salient and electorally relevant issue on the public agenda. The government prioritized efforts in favor of conflict containment, mirroring the dynamics in several other political contexts.
Furthermore, a key purpose of this research is to explore theoretical aspects that could underpin future studies using the NPF or the scope of conflict. Integrating the NPF with assumptions and concepts drawn from the conflict management model further enriches the analysis and may lead to revised theoretical expectations.
First, a major contribution of this study to NPF scholarship is a more consistent integration of Pralle's (2006) model of conflict management. Considering the Scottish fracking debate as a relatively mature policy conflict, expectations 1a and 2a emphasize symmetrical strategies of conflict expansion by both pro-and anti-fracking coalitions. This is only confirmed for conflict expansion through benefits and costs/risks (expectation 2a). However, to this can be added the findings about both coalitions' analogous targeting of the Scottish government and the opposing coalition (approx. a 40%/60% split). The conflict management model warrants further study because it usefully reflects the long-standing emphasis on conflict expansion in NPF studies and offers a connection between conflict expansion and the frequent use of characters (characterization contests). It can also be applied to another aspect of conflict expansion that was not explored in this study: an analysis of policy surrogates, the deliberate linkage of a policy issue to overarching political controversies (McBeth et al., 2007, Schattschneider, 1975, for instance Scottish climate policy. Additionally, the conflict management model is particularly suited to exploring long-running and 'mature' or intractable conflicts and could be further enriched with reflections on conflict containment and agenda denial from the agenda-setting literature (e.g. Cobb & Ross, 1997).
Second, the integration of the conflict management model yields more promising results than extant NPF propositions about the narrative strategies of winning and losing coalitions (expectations 1b, 1c, 2b). Using a policy timeline informed by the changing discourse of the main regulator (Scottish government) to determine winning and losing coalitions is a feasible approach, although it depends on careful, case-specific interpretation and may not work equally well for ongoing conflicts. In this study, the results for expectations regarding winning and losing coalitions are either mixed or negative. Overall, my findings reinforce the mounting skepticism about the predictability of winning and losing narrative strategies (Gottlieb et al., 2018;Merry, 2018).
Third, although winning and losing coalition have limited explanatory power in this case study, there are various ways of defining and coding the status of winning and losing in a policy conflict (Gottlieb et al. 2018). Some of these approaches may be easier to replicate than my case-specific, interpretive approach based on the changing discourse of the main regulator. Additionally, in the Scottish case only two out of three phases could be reliably categorized and more revealing results might arise from examining longer-running policy conflicts. Thus, it could be argued that the dynamic nature of the debate may not be fully represented by the three phases. Annualized (year by year) analysis may provide an alternative perspective (e.g. see Weible, Olofsson, Costie, Katz, & Heikkila, 2016). Yet, amid doubts over the utility of the winning-losing dichotomy, examination of the devil-angel shift (or campaign tone) does offer a clear trend in this case study: over time, the anti-fracking coalition both reduced its negative characterizations of the opposing coalition and targeted the Scottish Government with a mixture of criticism and forceful encouragement. My findings support  proposition that greater regulatory uncertainty may promote the devil shift. This is particularly evident in phases 1 and 2, whereas phase 3, by which time the policy outcome is more transparent, records less vilification, bearing in mind that the devil-angel shift does not take account of references to the 'non-aligned' Scottish government. Thus, there may be important underlying reasons for the trend regarding positive and negative campaign tone. These trends could be tested in future studies and may be incorporated into future NPF expectations. They might equally inform other framing and discursive approaches.
This article also draws attention to some methodological considerations that could be taken up by future NPF studies. First, following Merry (2016), coding additional characters generates a more nuanced picture of narrative characterizations. The present study demonstrates the differential use of antagonists (and of characters overall) between the two coalitions due to the anti-fracking group's frequent reliance on opponents. Focusing only on heroes, villains, and victims may ultimately achieve similar results, but risks leaving some of the more 'muted' narrative characters as uncategorized (Huda, 2018). Increasing precision in and greater convergence of codebook instructions in NPF studies would help researchers avoid potential inconsistencies, regardless of whether or not they code additional characters. This also applies to inclusion criteria for the sample, the practice (in this study) of treating newspapers as a conduit of policy debates by only coding unique (rather than repeated) narrative components, and the choice of calculating mean occurrences per newspaper article rather than the mere presence/absence of narrative components.
Second, the analysis of newspaper archives proves to be a practical approach to identifying a large number of policy narrative components on a given issue. Yet, in comparison to other documents (e.g. campaign literature, newsletters), the ambiguous status of newspapers as both conduits and potential contributors to policy debates (Shanahan et al., 2008) might place the results of newspaper analysis in a separate category of its own. Some of the findings in this study, such as the preponderance of antagonistic characters or the consistent targeting of the Scottish government, may be related to journalists' preference for conflictual debate, negatively framed arguments, and for holding the government to account (Montpetit et al., 2016;Nie, 2003). Future studies comparing newspaper analysis with other relevant data sources would be useful to determine whether a general bias exists across different policy areas and countries.
Notes 2 Other NPF studies have equally used a tripartite differentiation of phases to investigate the changing strategies of policy actors over time.  examined shale gas policy in Colorado and distinguished between the formal policy process and time periods before and after this process. In their study of wildfire policy (also in Colorado), Crow et al. (2017) regard external conditions (wildfire seasons) as demarcating the different phases. Thus, while these distinctions are all motivated by a longitudinal analysis of policy narrative practices, they are not based on a single formula. Rather, they are adapted to the particular context of each case study. 3 Without discounting, the general devil shift orientation would be even stronger and more enduring.