Chapter 7: Sanctuary Cities, Focusing Events, and the Solidarity Shift: A Standard Measurement of the Prevalence of Victims for the Narrative Policy Framework

Aaron Smith-Walter*, University of Massachusetts Lowell
Emily Fritz, University of Massachusetts Lowell
Shannon O’Doherty, University of Massachusetts Lowell



Numerous state and local jurisdictions across the United States have adopted policies limiting their cooperation with federal deportation efforts carried out by Immigration and Customs Enforcement (ICE). Sometimes referred to as “sanctuary cities,” these jurisdictions interpret federalism in a way that resists active participation in federal immigration enforcement. Employing the Narrative Policy Framework (NPF), we analyze 164 public consumption documents to examine policy narratives disseminated by interest groups engaged in the policy debate surrounding sanctuary cities between 2010 and 2017. Using data derived from a content analysis of these documents, we develop a new measure, the solidarity shift, to capture the prevalence of victims in policy narratives; we find there are significant differences in the narrative strategies employed by advocates and opponents of sanctuary jurisdictions, with opponents’ narratives demonstrating more active responses to external events and a higher proportion of victims, relative to other characters. We also find that the killing of Kathryn Steinle in San Francisco can be seen as a focusing event because of the narrative actions of anti-sanctuary city advocates and their reliance on the solidarity shift, which resulted in significant changes to anti-sanctuary city narrative strategies.


*Corresponding author:

To cite the chapter: Smith-Walter, Aaron, Emily Fritz, and Shannon O’Doherty. 2022. “Sanctuary Cities, Focusing Events, and the Solidarity Shift: A Standard Measurement of the Prevalence of Victims in the Narrative Policy Framework”, in Narratives and the Policy Process: Applications of the Narrative Policy Framework, Michael D. Jones, Mark K. McBeth, and Elizabeth A. Shanahan(eds.), Montana State University Library, 162-193.



      During the 2016 presidential election campaign, Donald J. Trump captured the national spotlight—and the Republican Party nomination—as he staked out a forceful position against undocumented immigrants by promising to “build a wall” to keep them from entering the United States via the US.–Mexico border and to defund sanctuary cities that limit cooperation with federal authorities. A series of tweets related to the killing of Kathryn Steinle by an undocumented immigrant in San Francisco, a sanctuary city, illustrated Trump’s opposition to the current state of immigration enforcement. In these tweets, Trump expressed his condolences to the Steinle family, tied her death to an illegal immigrant, and asserted that crimes committed by undocumented immigrants could be curtailed by the building of a border wall (Brown, n.d.).

      In this chapter, we explore the relationship between the casting of certain characters in policy narratives as an example of what narrative policy framework (NPF) scholars have identified as narrative strategies. Narrative strategies are how policy stories are constructed to achieve particular ends (Jones et al., 2014). For example, political actors who are interested in changing policy might deploy a narrative that portrays their opponents as more powerful and malicious than they are: a strategy named the devil shift (Sabatier et al., 1987; Shanahan et al., 2013). This strategy appears when interest groups want to mobilize the base to take political action (Merry, 2016) or when a policy arena is highly intractable because of antagonistic policy beliefs (Sabatier et al., 1987). Although the NPF has used the devil shift to measure the prevalence of villains as a narrative strategy and has developed a corresponding measure to examine the emphasis that policy actors place on heroes (the angel shift), there has been much less attention paid (with notable exceptions, e.g., Brewer, 2019; Smith-Walter, 2018) to formulating a standard measure that focuses on victims in policy narratives.

      We seek to contribute to existing NPF literature on three important concepts relevant to the study of policy narratives surrounding sanctuary cities: (a) the importance of focusing events, (b) the need for a victim-centric measure analogous to the devil–angel shift, and (c) the importance of federalism for investigating sanctuary city policy debates. This objective gives rise to three distinct research questions:

Research Question 1: To what extent do competing coalitions in the sanctuary city debate differ in their use of the angel shift and devil shift as narrative strategies?

Research Question 2: To what extent do competing coalitions differ in their use of the number of victims as a narrative strategy?

Research Question 3: To what extent do competing coalitions differ in their conception of federalism as a policy belief?

      We will answer these questions in this chapter by first reviewing the framework’s key components and the literature. We then present the case, data, and methods, followed by our findings. We close the chapter with a discussion of the implications of our findings and the chapter’s limitations.

Narrative Policy Framework

      The NPF posits that stories, or policy narratives that actors deploy in a policy arena, are important objects of study; they can work as a resource to mobilize, demobilize, convince, cajole, and attract or repel mass publics, political elites, and bureaucratic officials. The NPF posits that narratives are composed of context-specific narrative content and generalizable narrative form. The content of a narrative is germane to the policy area in which it is deployed and is resistant to generalizations regarding its impact on policy outcomes. This means the policy narratives on climate change policy are likely to feature characters (e.g., oil companies) and relationships (e.g., modern industry’s reliance on cheap fossil fuels) that are not necessarily illuminating when applied to immigration policy. However, the NPF adopts a structural understanding of narrative form and proposes that at least four aspects of policy narratives, (a) characters, (b) setting, (c) plot, and (d) moral of the story (Shanahan et al., 2017) are foundational to narratives and are amenable to generalization across policy areas using standard social science methods.

      The characters include the hero, who works to bring about a policy solution, and the villain, whose actions cause or threaten to cause harm to a victim. The setting is the social, political, economic, and legal background in which the story is told. The plot relates how the characters interact with one another and the setting through time, and the moral of the story is the solution to the problem that the hero champions (McBeth et al., 2014b). These characters are a key component of the framework, and the presence of at least one character and the presence of a policy referent are required for a policy narrative to exist (Shanahan et al., 2013, p. 457).

      Narrative strategies are posited to exist as narrative content and are thus in need of some “anchor” to established systems of meaning to combat the problem of narrative relativity.[1] Anchoring narrative content to structural components, such as the devil–angel shift and other character-based approaches, has been the most common approach to addressing the issue of narrative relativity, although it does not exhaust all possible approaches to making content more generally applicable to multiple policy domains. For example, the causal mechanisms laid out in Deborah Stone’s (2012) work have been used as narrative strategies. These studies have found that narratives around bison restoration on public lands that featured villains who intentionally cause harm are more persuasive than those where the harm is mechanical (Shanahan et al., 2014). Different causal mechanisms were also found to be more convincing in a survey experiment on campaign finance reform, although this study found the mechanical cause induced greater support for (and less resistance to) public financing than intentional causation (Jorgensen et al., 2018).

      NPF scholars have explored causal mechanisms, such as the growth or reduction in the scope of policy conflict (McBeth et al., 2007), distributions of costs and benefits (Bragg & Soler, 2017; Gupta et al., 2014), and policy actors presenting themselves as “winning” or “losing” in a policy conflict (Gupta et al., 2014; McBeth et al., 2007; Shanahan et al., 2013). However, the most frequently employed and best-tested narrative strategy in the NPF is the devil–angel shift; this is partly because of its association with the advocacy coalition framework (ACF) and partly because it allows characters to be quantified and statistically analyzed. This chapter expands the character-based approach of the devil–angel shift by applying similar measurement logic to victim characters featured in policy narratives.

      We now turn our attention to the three theoretical concerns that undergird this chapter. First, we examine the importance of the relationship between an incident occurring and the role that victims play in crafting a narrative that effectively creates a distinct focusing event. Second, the lack of a standard measure for the prevalence of victims (vis-à-vis other characters in NPF studies) and the NPF’s existing measure of victim-centric narratives, the impotent shift, although extremely valuable in certain cases, does not have the broad applicability needed. Finally, we must acknowledge and account for the vital role that federalism plays in policy development and implementation, especially in those policy areas (e.g., sanctuary cities) characterized by conflict between constitutionally distinct spheres of governmental authority.

Focusing Events and NPF Literature

      A focusing event, defined as an event that can advance the priority of an issue on an agenda by commanding attention in a sudden and noteworthy fashion (Birkland, 1998, p. 53), is a key concept in many policy process studies. As such, the NPF has integrated the concept of focusing events into a number of studies (Lawlor & Crow, 2018; McBeth & Lybecker, 2018; McBeth et al., 2014a; Stauffer 2020). Birkland illustrated the link between a focusing event and narrative strategy by noting that focusing events can serve as a touchstone for groups trying to mobilize support because the event may serve to reorganize the policy priorities on a decision maker’s agenda. This reshuffling of priorities can prompt an effort by interest groups to expand the issue to include new policy actors or to resist those efforts and maintain the status quo in the policy subsystem (Birkland, 1998, p. 53). Another important connection is that focusing events are characterized by their harmful or potentially harmful nature. This creates a situation where victims become an integral aspect of policy narratives seeking to use a focusing event to mobilize support.

      Birkland (1998) proposed that pro-change groups responding to focusing events will pursue mobilization in a number of ways, including fundraising appeals and invitations to become members of advocacy organizations. Other mobilization strategies are those that ask members and fellow travelers to contact elected representatives to express their concerns, engage in boycotts, or attend public demonstrations. Merry (2018) found that fundraising communications were more likely to rely on the angel shift and that calls for political action were more likely to invoke the devil shift, which supports the dynamic Birkland outlined.

      Several NPF scholars have employed the notion of focusing events, including an examination of policy narratives being used to construct a social crisis around increasing rates of obesity in the United States (McBeth et al., 2016). McBeth et al. (2016), drawing on the history of tobacco regulation, noted the process for successfully defining a public crisis, “depends on who is blamed for a problem, who is harmed, and who the advocates for change are” (p. 142, emphasis added). Indeed, the need to tell a story that includes sympathetic victims has been identified as a recommendation for public health officials to keep in mind when constructing their communications, but the study reported no investigation of victims.

      In a second study, Stauffer (2020) dealt with the implementation of reform to child and adult protective services in Switzerland. The study focused on the distinct effects that cultural differences in the French- and German-speaking cantons had on the structure of implementation and how the murder of two children in German-speaking Zurich served as a focusing event to catalyze discourse and mobilize public opinion.

      However, the most expansive examination of the link between focusing events and the NPF may be McBeth and Lybecker’s (2018) examination of the rise of sanctuary cities as a policy issue in the United States. This study informs our work most directly in the current chapter. McBeth and Lybecker asserted that policy narratives play a role in creating both a focusing event and a flow from the event itself. In this way, they captured the more expansive version of the focusing event (Birkland & Warnement, 2016), which emphasizes their contingent nature and the need for organized advocacy groups to construct the event in a manner that maximizes the chances that policy action will occur.

      McBeth and Lybecker (2018) focused on the emergent nature of the issue of sanctuary cities and the ambiguity surrounding an established definition to investigate how policy actors used their power to advance particular images of these entities in competing policy narratives. The conflict over sanctuary cities saw substantial increases in media attention following the tragic shooting of Kathryn (“Kate”) Steinle, which the authors argued became a focusing event for the latest round of elevated discord (McBeth & Lybecker, 2018, p. 873). The study used the devil shift to measure the policy narratives published by Breitbart News and advanced by the GOP during the debate on the floor of the US House of Representatives on H.R. 3009, the Enforce the Law for Sanctuary Cities Act. They discovered a strong devil shift in both the Breitbart (-0.608) and GOP (-0.593) narratives. It noted that Steinle’s accused killer, Jose Ines Garcia Zarante (aka Juan Francisco Lopez Sanchez), was the primary villain and Steinle the primary victim of the Breitbart narratives. However, the GOP narratives featured President Obama as the primary villain and the public and Steinle as the primary victim.

      The authors argued that this adaptation allowed for a quick expansion of the overall scope of the problem, from a single incident in San Francisco to a nationwide epidemic of crimes committed by undocumented immigrants (McBeth & Lybecker, 2018, p. 885). They noted the policy narrative that featured Steinle as a victim was successful; similarly, horrific events did not manage to become focusing events likely because the Steinle case combined “the right villain, the right innocent victim, the right hero arriving on the scene, [and] the right setting” (McBeth & Lybecker, 2018, p. 885). With this in mind, we hypothesize the following:

H1: Anti-sanctuary narratives will demonstrate an increase in the devil shift following the shooting of Kathryn Steinle.

As a corollary to this hypothesis, we also hypothesize the following:

H2: Anti-sanctuary narratives will demonstrate less of a devil shift, and pro-sanctuary narratives more of a devil shift, following President Trump’s issuance of Executive Order (EO) 13768.

If this is the case, it suggests we can help identify focusing events by finding those stories that work to identify “the right villain” and “the right hero”; however, we also must be able to identify “the right innocent victim,” which allows the policy narrative to effectively make the move from a particular event to a more general problem (McBeth & Lybecker, 2018, p. 885). The NPF can rely on the devil–angel shift to assist as an indicator of the first two components of McBeth and Lybecker’s formulation. However, the existing measure cannot begin to account for the latter movement because victims traditionally have been measured in NPF studies without considering their proportionality to heroes and villains.[2] In this chapter, we argue that NPF scholarship should become more attentive to the prevalence of victims in policy narratives, explore existing attempts to measure these characters, and propose a general-purpose measurement that can serve to add a quantitative measure to the NPF that augments the devil and angel shifts, encouraging the expression of all three character types in relation to one another. Before we move on to that, however, we must briefly explore the development of the devil–angel shift in the NPF and what has been discovered about narratives by its deployment.

The Development of the Devil–Angel Shift

      The development of the devil shift as a measure of character-based narrative strategy in the NPF is relatively straightforward, but it does contain one crucial mutation that generates a key blind spot in the NPF and that this chapter aims to address. To this end, we review the existing literature to gain a better understanding of the theory behind the measure, the various approaches to using it, and the findings from key studies.

      The origin of the devil shift concept can be traced to Sabatier et al. (1987), who explored the tendency for competing elites to perceive their opponents as more malicious and more powerful than they truly are (p. 451). The ACF and the NPF share an affinity for empirical hypothesis testing (Shanahan et al., 2011) and adherence to a social science standard that is “clear enough to be proven wrong” (Sabatier, 2000, p. 137; Jones & McBeth, 2010). The first mention of the devil shift in an explicitly NPF article is found in Shanahan et al. (2011), and it appears as a theoretically derived hypothesis imported from the ACF and was first used for empirical application in Shanahan et al. (2013). The initial use of the measure sought to capture a particular aspect of the devil shift as a strategy by examining the percent of total references to self as hero minus the percent of total villain references to other as villain divided by the total number of heroes plus the number of villains (Shanahan et al., 2013, p. 466). This means the devil–angel shift as reported in 2013 would produce a measure that did not include heroes who were not synonymous with the narrator.

      However, Heikkila et al. (2014) noted the devil–angel shift measure described in the Shanahan et al. (2013) piece had a typographical error in the formula for computing the measure.[3] They clarified the measure with Dr. Shanahan, finding the characters to be counted as heroes were not only those identical to the narrator but also all heroes. As such, they used the shift as the “number of references to one’s own group/coalition as a hero, minus the number of references to the opposing group/coalition as a villain, and then divides that difference by the number of references to one’s own group as a hero plus the number of references to the opposing group as a villain” (Heikkila et al., 2014, p. 193). This expansion to one’s coalition as a hero means any hero character was now clearly included in the calculation, not just those references that the group made to itself as a hero, and it has become the preferred measure for computing the devil shift in NPF studies.[4]

      In any case, several other NPF studies adopted the earlier formulation of the devil–angel shift. For example, Crow and Berggren (2014) reported the devil–angel shift by counting self as hero minus opponent as villain divided by the total number of characters (p. 148), as did Lebel and Lebel (2018). This is an important observation because it appears others adopted the revised calculation (Gottlieb et al., 2018; Merry, 2016; Smith-Walter, 2018; Smith-Walter et al., 2016; Stephan, 2020) in their computation of the devil–angel shift, thus incorporating a complete accounting for the balance of positive and negative characters in the narrative. These differences are especially salient for the development of our victim-focused narrative measure and will be addressed in further detail below. Before that discussion, we delve into the findings from the devil–angel shift in NPF literature.

 Devil–Angel Shift in NPF Studies

      Although the devil shift is a widely used measure of narrative strategy across existing NPF studies, for the sake of brevity, here we will examine only two groups of studies within extant NPF literature: gun policy studies (Merry, 2016, 2019; Smith-Walter et al., 2016) and energy policy studies (Gottlieb et al., 2018; Heikkila et al., 2014; Shanahan et al., 2013; Stephan, 2020). This review is intended to illustrate the importance of the standardized devil–angel shift measures in comparing and contrasting studies within policy arenas and across cases and the potential value that a similar victim-focused measurement might have on future efforts to synthesize and systematize case-based NPF findings.

Gun Policy

      Gun policy, a highly charged and relatively intractable policy arena, has been the focus of numerous NPF studies and has produced interesting findings. For example, Smith-Walter et al. (2016) found that National Rifle Association (NRA) narratives had a larger devil shift and that it used this strategy more frequently than the Brady Campaign. They also found there were no statistically significant differences between the number of villains appearing in NRA and Brady Campaign narratives in publications intended for their own memberships, although differences between the use of heroes and victims were statistically significant.

      Contrary to findings in Smith-Walter et al. (2016) and the NPF hypothesis that the devil shift is more likely to emerge in policy debates seen as intractable (p. 376), Merry’s (2016) analysis of 10,000 tweets from the Brady Campaign and the NRA found the Brady Campaign used more villains than the NRA, and the differences in the devil–angel shift between the two groups were small (0.15 for Brady and -0.04 for the NRA). In her 2019 investigation, Merry found there were differences between the use of the devil and angel shifts between the gun rights and gun control coalitions, with the gun control advocates using both the angel and devil shifts more than gun rights organizations (Merry, 2019, p. 899), which is inconsistent with earlier findings in NPF studies in this policy area (Merry, 2016; Smith-Walter et al., 2016).

      However, Merry (2016, 2019) and Smith-Walter et al. (2016) both found that gun control advocates used victims in their narratives far more frequently than did the NRA (p. 383).[5] This suggests that although the devil–angel shift results did not align with each other, the heavier reliance of gun control advocates on victims did seem to hold over individual three narrative studies. These findings suggest the role of victims and their relationship to other characters should command more attention from future NPF scholarship.

Energy Policy

       We now turn our attention to another area that has received substantial attention from NPF researchers: energy policy. Gottlieb et al. (2018) noted that prior NPF research found that industry and environmental groups were more likely to portray themselves as heroes than they were to cast other groups as heroes, and narratives from environmental groups were more likely to feature villains than industry-disseminated policy narratives (Heikkila et al., 2014). They also cited Blair and McCormack (2016), who found that conservative sources were more likely to feature industry players as heroes, whereas environmentalists were the protagonist of choice for liberals. They used these findings to hypothesize that coalitions will deploy characters in distinct ways that depend on which side of the debate the narrator finds themself. This is contrasted with findings from a study based on traditional NPF theorizing (Shanahan et al., 2013) that found the winning coalition more frequently used the angel shift, whereas the devil shift was the strategy most often used by the losing coalition.

      In their review of four New York county fracking policies, Gottlieb et al. (2018) found that the hypotheses derived from the winning/losing perceptions of the coalition did not hold, and extant NPF research by Crow and Berggren (2014), Heikkila et al. (2014), Shanahan et al. (2013), and Blair and McCormack (2016) demonstrated that anti-fracking groups used victims more than pro-fracking narratives; this suggests (at least in this case) that knowing which groups relied more heavily on victims was a better indicator of policy stance (pro or con) than either the devil or angel shift.

      This review is by no means exhaustive, but it does indicate that across these studies, the proportion of victims used seems to vary by the coalition, and this variation corresponds to Gottlieb et al.’s (2018) hypothesis that policy actors are likely to use different narrative strategies based on which side of the conflict they find themselves instead of whether they portray themselves as winning or losing. Therefore, we hypothesize the following:

H3: There will be differences between pro- and anti-sanctuary narratives concerning the use of the devil shift.

The Impotent Shift

      The development of the devil–angel shift is important for NPF scholars to note because the split between these distinct calculations was replicated in recent work by Brewer (2019) in his investigation of the policy narratives surrounding the Columbia River Crossing project in the US Northwest. This piece paid special attention to the role of victims in policy narratives and stands as a notable exception to general underdevelopment of a measurement strategy to address the prevalence of victims in policy narratives. Brewer (2019) used the NPF as a descriptive framework to evaluate the policy debates surrounding a bridge-building project in the US Pacific Northwest and explored whether a narrative strategy exists where the narrator presents themself as a victim. In this work, Brewer (2019) coined the term impotent shift to refer to a policy narrative “that emphasizes the victimhood of the narrating individual/group while understating their position as the hero in solving a policy problem” (p. 503).

      The impotent shift is an important innovation that augments and expands measures of policy strategies at the disposal of NPF scholarship. However, the operationalization of the impotent shift does not aim to capture a general ratio of victims to heroes and villains. By building the measurement for the impotent shift on the foundations of the angel–devil shift measure as formulated in Shanahan et al. (2013), the impotent shift focuses attention on how the narrator characterizes their own role; the definition links the narrator to a narrative strategy in which they assume the victim role while downplaying their efficacy as a heroic bearer of the policy solution (Brewer, 2019, p. 503).

      Although this approach is useful in exploring how the narrator positions themself, it does not capture the overall ratio of victim characters present in the narrative relative to other characters. Brewer’s approach is more nuanced and is likely to be more precise when seeking to identify and measure the use of the impotent shift as a narrative strategy, but the research reported in this chapter rests on an adaptation of the devil–angel shift as reformulated by Heikkila et al. (2014, p. 192) that sought to express the difference between the total number of heroes and villains in a narrative compared to the total number of characters presented as heroes and villains.

The Solidarity Shift: A Theoretical Basis

      This research focuses on the victims presented in policy narratives and their role in building compelling stories. We rely on Nadler’s (2020) recognition of two distinct yet fundamentally related human psychological needs, belonging and independence, to lay the foundation of a victim-centric measure that we term the solidarity shift. Unlike the impotent shift, the solidarity shift does not focus on the presentation of the narrator as a victim; instead, it focuses on the need for the narrator to convince their audience to take action on behalf of a third party—in essence, to act in solidarity with the victims portrayed in the policy narrative. When the narrator presents a narrative in which they highlight the harm done to others and the need to intercede on other groups’ and individuals’ behalf, they are acting on the need for a group sense of belonging, but as Nadler (2020) noted, those providing the assistance may also be “an assertion of superiority, and dependency on others is shameful” (p. 13).

      Indeed, Nadler (2020) recounted that Roman philosopher Seneca distinguished between two motivations for assisting others. The first of these, beneficius, was the desire to alleviate others’ suffering out of a sense of compassion and a lack of any expected reciprocity (p. 14). The second motivation for helping others was named munus, which was rendering assistance to others with “the desire to gain fame, honor and prestige in return for kindness” (p. 14). Importantly, both motivations can be expressed by helping others; however, actions intended to strengthen feelings of belonging are rooted in social solidarity (beneficius), whereas actions that desire to demonstrate the dominance of one group over another (munus) are geared toward the production of relationships of dependence and the maintenance of social hierarchies (Nadler, 2020, p. 15).

      Based on the construction of the impotent shift, we believe this construct may best reflect a use of the victim character as one that serves to create and justify hierarchical relationships, munus. Using the impotent shift, Brewer found neither coalition victimized itself as a narrative strategy (2019, p. 513). However, Brewer noted that “both coalitions saw the use of the victim character as an effective strategy and heavily referred to ‘citizens’ as the victims of the perceived policy problem” (2019, p. 514). As constructed in this chapter, the solidarity shift captures the overall use and deployment of the victim character and its prevalence compared to other characters in the narrative. The solidarity shift does not distinguish between the use of victims motivated by beneficius versus munus, but it does capture the general use of victims. This broad measure could be used in conjunction with the impotent shift to devise a distinction between (a) narratives using victims to call for audience participation and involvement in a policy conflict and (b) narratives seeking to establish the narrator as a conduit for change, reinforcing and reifying hierarchical relationships.

      One way to understand the underlying social psychological basis for the solidarity shift is by employing the social identity perspective, which holds that an individual’s identification as a member of a particular group or groups is incorporated into their notion of self, and when group membership appears to be an important factor in social situations, “an interpersonal interaction is actually an intergroup one” (Nadler, 2020, p. 139). Another component within the social identity perspective is the intergroup helping as status relations (IHSR) model, which is key to understanding intergroup dynamics within structurally unequal systems such as human political formations. The relevant portions of this model suggest assistance given from a privileged group to an underprivileged group is likely to be assistance that functions to cause the less-advantaged group to incur debts to the advantaged group while simultaneously raising the powerful group’s prestige and status.

      The dynamics of this relationship express the view that “solving the low-status group’s problem is the responsibility of the higher-status group” (Nadler, 2020, pp. 153–154). This logic is similar to the logic undergirding the impotent shift in that the hero character—who has the agency as well as a policy proposal, resources, or other attributes that can address the problem—takes on the role of the victim and uses the empathy generated for the victim’s plight to increase their power and efficacy. This type of relationship is one in which the high-status group seeks to assist others while maintaining its own place of social dominance in the hierarchy. The NRA is an excellent example of this type of victim utilization; it often casts itself as the target of unconstitutional infringement on its Second Amendment rights or attacks by antagonistic mainstream media outlets. This is then parlayed into requests for money, which it uses to fight for its preferred policy aims.

      Although interested in the role of the victim, the solidarity shift does not focus on when the hero adopts the victim persona. Instead, the solidarity shift is interested more directly in the prevalence of victims employed in the policy narrative offered by the narrator—a measure that is more in line with the formulation of the devil–angel shift as operationalized by many NPF studies. Astute readers will note this distinction does not preclude the logic of noblesse oblige as a motivation behind their claims to act on behalf of the harmed. However, it does open the door to a motive for helping others by altering those unjust structures producing harm. This type of assistance is egalitarian in its desire because it seeks to eliminate the cause of harm by enhancing the autonomy and agency of the victims, often by appealing to moral outrage (Van de Vyver & Abrams, 2015). This “leveling” desire seeks to build a common identification and solidarity between victims and supporters to mobilize and empower them. As such, we hypothesize the following:

H4: There will be differences between pro- and anti-sanctuary narratives concerning the use of the solidarity shift.

The Importance of Federalism

      Sanctuary cities in the United States can be traced to a response by churches and local communities to the increasing number of people fleeing civil wars in Central America in the 1980s. Eventually, four US states implemented sanctuary policies that provided services and support to immigrants regardless of immigration status, in contravention to the federal stance of denying them refuge (Amdur, 2016, p. 95). Congress responded to this movement by passing the Illegal Immigration Reform and Immigrant Responsibility Act of 1996 (IIRIRA) (Amdur, 2016; Juárez et al., 2018; Macías-Rojas, 2018).

      Under the IIRIRA, federal funding for immigration enforcement significantly increased, and a local–federal cooperative program for immigration enforcement, the 287(g) program, was established (Amdur, 2016, p. 99; Juárez et al., 2018). The 287(g) program permitted local law enforcement agencies to enforce immigration law on behalf of the federal government. Some subnational governments warmly embraced 287(g), whereas others met it with virulent opposition. These latter jurisdictions balked at “the attendant financial burdens and harm to police work” (Amdur, 2016, p. 89) that involvement in immigration enforcement might entail. The 287(g) program enjoyed some initial successes but ultimately generated serious controversy because of racial profiling by local law enforcement agencies.

      Another federal immigration enforcement program, “Secure Communities,” was created during the George W. Bush presidency and implemented during the Obama administration (Rodriguez, 2017). This program leveraged the long-standing partnership between local law enforcement agencies (LEAs) and the Federal Bureau of Investigation (FBI), whereby arrested individuals’ fingerprints were submitted to the FBI. Secure Communities relied on the fact that the FBI forwarded these fingerprints to US ICE. If ICE decided it wanted to take the arrested individual into custody, it would issue a detainer request to the LEA to hold that individual until ICE could collect the person in question (Kandel, 2016).

      Secure Communities proved to be extremely effective at removing undocumented immigrants from the United States. Because the Secure Communities mechanisms were activated by the arrest and not the conviction of an individual, vast numbers of people were arrested, detained, taken into custody by ICE, and deported. Johnson (2016) noted that the program resulted in the deportation of “approximately 400,000 non-citizens a year in the first six years of the Obama presidency” (p. 1016). Under Secure Communities, undocumented immigrants reported rising anxiety and fear because LEAs were seen as agents of immigration enforcement (Bhatt, 2017) and not as public servants engaged in protecting residents’ lives and properties.

      Increasing subnational resistance to Secure Communities led the Obama administration to replace it with the Priority Enforcement Program (PEP). PEP utilized the same mechanisms as Secure Communities, but it refrained from requesting detainers from local LEAs for individuals who were merely arrested and focused instead on those convicted of severe crimes (Johnson, 2016). PEP was operative between 2015 and 2017 until President Trump’s EO 13768 reactivated Secure Communities. This order was a partial fulfillment of a campaign promise to take aggressive action against individuals present in the United States without authorization. Given the Trump administration’s orientation during this period, we hypothesize the following:

H5: Anti-sanctuary advocates will be more likely to cast the federal government as a hero.


      As the above paragraphs illustrate, the policy conflict around sanctuary cities is embedded in the complex web of US federalism. An additional complication is the fact that no legal definition of a sanctuary city exists. However, it is generally understood to relate to the level of cooperation given by subnational governments to immigration enforcement efforts undertaken by ICE. For example, ICE relies on an administrative procedure known as a “civil immigration detainer request” to ask LEAs to maintain custody of undocumented immigrants following their arrest (Manuel, 2015). However, states and localities cannot be compelled to honor these requests because this amounts to an unconstitutional “commandeering” of state resources by the federal government. Because detainer requests are a lynchpin in contemporary federal immigration enforcement strategy, sanctuary policies generally address the level of compliance with these requests or the state and local procedures governing the collection of information regarding residents’ immigration status. As such, our definition of a “sanctuary jurisdiction” for this research is any subnational government that through legislation, municipal ordinance, administrative policy, or officially sanctioned means limits or restricts the information collected concerning an individual’s immigration status and/or routinely declines to honor ICE civil immigration detainer requests.

      Having explored the importance of federalism in sanctuary city policy, we return to McBeth and Lybecker’s (2018) study of policy narratives on sanctuary cities and the link they identify between Steinle as a victim of crime-ridden sanctuary cities and Trump’s condemnation of these jurisdictions. This allowed the forging of a policy narrative that helped elevate (or perhaps highlight) the public’s interest in crimes committed by undocumented immigrants more generally (McBeth & Lybecker, 2018, p. 884). Our research is complementary to that of McBeth and Lybecker, but it is distinct because we explore policy narratives before and after Steinle’s shooting and explore whether new policy narratives related to sanctuary cities emerged (2018, p. 887).

Data and Methods

      This study focuses on the policy narratives discovered in media sources and interest groups cited in two daily newspapers, the Boston Globe and the New York Times. We chose these sources because of their geographic location, their place in the top 10 papers in the United States by circulation (Agility PR, 2020), and the fact that both Boston and New York City have active sanctuary policies. These newspapers served as a starting point we used to identify anti-sanctuary and pro-sanctuary advocates active in this policy arena.

      We conducted a search in the Boston Globe and the New York Times archives between 2010 and 2017 for mentions of sanctuary cities to identify published news articles pertaining to the sanctuary city debate in the United States. After examining the articles from this search, we identified key pro- and anti-sanctuary organizations mentioned in the articles’ text. We further narrowed our pool of organizations by only including organizations with websites. The logic behind this choice was that if someone was exposed to the sanctuary city controversy via mainstream newspapers, then they would be likely to turn to the website of the organization mentioned in the paper to learn more. We then searched each organization’s website for the term “sanctuary city,” which yielded 559 documents from the organizations listed below.

  • American Immigration Lawyers Association (AILA)
  • American Civil Liberties Union (ACLU)
  • Boston Globe
  • Center for Immigration Studies (CIS)
  • Centro Presente
  • Families for Freedom
  • Irish International Immigration Center
  • The Center for American Progress
  • The National Council of La Raza (now UnidosUS)
  • The National League of Cities
  • The New York Times
  • Service Employees International Union
  • US Conference of Mayors
  • Make the Road New York
  • Migration Policy Institute
  • Major County Sheriff’s Association

We then evaluated these documents for their conformity to the following criteria:

  1. Must contain a policy narrative
  2. Must reference “sanctuary” within the text
  3. Must have been published between January 1, 2010, and December 31, 2017

After eliminating documents that fell outside these guidelines, 282 documents remained. We then randomly selected 164 documents to code, which we then coded to create the data analyzed in this study. We arrived at this number of documents by calculating the required sample size to achieve a 95% confidence level with a ±4.88% margin of error. We coded for the presence or absence of characters (heroes, villains, and victims). Researchers coded for the stance of the document narrator on sanctuary policies. Narrators in 64 documents approved of sanctuary policies; 66 narratives displayed opposition; and 31 documents were unclear in their stance in that they mentioned the controversy but did not take a stand on whether they supported or opposed sanctuary city policies[6] (researchers disagreed about the nature of three narratives, and we ultimately excluded those narratives from the analysis). We generated the codebook using inductive and deductive techniques (see Appendix A for intercoder reliability metrics).

Tests and Measures

      To test our hypotheses, we used two existing NPF measures: the devil–angel shift (Heikkila et al., 2014) and McBeth et al.’s (2010) measure of federalism. The third measure was the solidarity shift, which we developed theoretically in the early parts of this chapter. Our statistical analysis employed the Kruskal–Wallis H test (the nonparametric equivalent of one-way ANOVA) and the Mann–Whitney U test (the nonparametric equivalent of an independent samples t-test) to test our hypotheses.

      We used the Kruskal–Wallis H and Mann–Whitney U nonparametric tests for several reasons. First, by constructing a series of histograms, we concluded that the distribution of data for our variables of interest was skewed. Second, the number of narratives present in the media environment during the period under study was limited in reference to our two focusing events, with pro-sanctuary narratives being scarce before the Steinle shooting (N = 5, 7.8% of all pro-sanctuary narratives) and anti-sanctuary narratives being reduced significantly following the issuance of EO 13768 (N = 10, 15.2% of all anti-sanctuary narratives).

Devil–Angel Shift Measure

      The devil shift has been used in previous NPF research and deals with the prevalence of villains in policy narratives. The devil shift is described as an argument that presents a group’s “opponents as ‘devils,’ i.e., as being more powerful and more ‘evil’ than they actually are” (Sabatier et al., 1987, p. 451). As such, in this chapter, we adopt the following formulation put forth by Heikkila et al. (2014, p. 192).

      To compute the devil shift, we relied on the following formula for a document-level statistic:

HV / H + V =


V = Total number of villains

H = Total number of heroes

This resulted in a value ranging from -1 to 1. Negative values are indicative of the presence of the [ . . .], and positive values are labeled as an angel shift.

Federalism–Localism Measure

      A key factor in the development and enforcement of immigration policy in the United States is federalism. Immigration enforcement traditionally has been recognized as a federal duty, which suggests anti-sanctuary advocates would be more likely to point to federal actors as heroes because they desire more stringent enforcement. By the same token, those in favor of the actions of subnational governments to establish sanctuaries would be more likely to lionize state and local governments, whose cooperation (or lack thereof) with federal programs could significantly degrade the efficacy of immigration enforcement. As such, we modified the measure proposed by McBeth et al. (2010) by substituting heroes for “allies” and restricting these categories to those characters with a clear federal, state, or local affiliation. This modification is required because the measure in McBeth et al. (2010) initially was devised to explore the evolution of policy views of a single interest group, and it thus held the narrator constant as the narrative’s hero. Therefore, it conceived all other supporters of the policy as allies. In our study, the identity of the narrator was not constant, so we did not use the hero–ally distinction; this meant those with a role in advancing a possible solution were included in the hero category. This resulted in the calculation of the federalism measure, as follows:

F – (S + L) / F + (S + L) = Federalism


F = Federal-level heroes

S = State-level heroes

L = Local-level heroes

The Solidarity Shift Measure

         NPF scholarship has concerned itself with the power of villains (Zanocco et al., 2018) and heroes (Jones, 2014); however, victims have received less attention. This is especially important for this issue because undocumented individuals, who are most likely to be harmed by increased enforcement, cannot generally cast votes[7] for lawmakers. As such, we hypothesize that painting a compelling picture of those hurt by action against sanctuary jurisdictions would be fundamental to convincing those with formal standing in the electoral process to take action. Therefore, pro-sanctuary narratives should feature more attention to those hurt by the current state of affairs. To measure the solidarity shift of a policy narrative, we computed a measure that operates on a -1.00 to +1.00 scale.

VI – (H + V) / VI + (H + V) = Solidarity shift


VI = Total victims

H = Total heroes

V = Total villains

Higher values indicated a higher proportion of victims relative to the other total count of the two characters.


      The first factor to note about our findings is that the total number of characters in pro- and anti-sanctuary narratives did not demonstrate a statistically significant difference (p = .979); what differed were the types of characters that the two groups employed. The average number of heroes deployed by pro-sanctuary narratives was 2.38 per narrative, whereas anti-sanctuary stories featured only 1.83 heroes per story. This difference was statistically significant (p = 0.033). Pro-sanctuary narratives demonstrated lower numbers of villains in their narratives (2.56 vs. 5.14, p < 0.001) and a higher number of victims (4.63 vs. 2.44, p < 0.001). With these contextual differences established, we now turn to the findings directly relevant to our hypotheses.


Table 1. Results of Hypothesis Testing

Hypothesis Supported
Yes No

H1: Anti-sanctuary narratives will demonstrate an increase in the devil shift following the shooting of Kathryn Steinle.


H2: Anti-sanctuary narratives will demonstrate a decrease in the devil shift following President Trump’s issuance of EO 13768.


H3: There will be differences between pro- and anti-sanctuary narratives concerning the use of the devil–angel shift.

H4: There will be differences between pro- and anti-sanctuary narratives concerning the use of the solidarity shift.

H5: Anti-sanctuary advocates will be more likely to cast the federal government as a hero.

H1: Anti-sanctuary narratives will demonstrate an increase in the devil shift following the shooting of Kathryn Steinle.

      Although the policy narratives demonstrated some interesting changes following the shooting death of Steinle, we did not find confirmation for H1; the only statistically significant differences in the structure of anti-sanctuary narratives following the shooting were a change in the federalism measure (with -1.00 meaning only state/local heroes and +1.00 meaning only federal heroes), moving from one measure that was slightly more likely to mention state and local governments as heroes (mean value of -0.0522) to a narrative structure with a heavier emphasis on federal actors as heroes (mean value of .3810). No such shift was seen in pro-sanctuary narratives. This lends support to McBeth and Lybecker’s (2018) findings that focusing events may open narrative windows to which groups can respond or remain committed to existing narrative formulations.

      The narratives’ devil–angel shift did not change following the shooting, nor did a statistically significant change emerge in the mean number of heroes, victims, or villains deployed in any of the narratives. This suggests a closer look at the heroes used by anti-sanctuary narratives is needed, and we now examine that aspect. The most common heroes in anti-sanctuary narratives before the shooting featured particular pieces of legislation (35% of narratives), state government officials (26%), local law enforcement (26%), ICE (22%), and anti-sanctuary activists (17%). After the shooting, anti-sanctuary activists were characterized as heroes in 40% of the narratives; likewise, ICE increased its heroic portrayal to 30% of the post-shooting narratives, as well as legislation (28%), Congress (23%), and Donald Trump (23%). This pattern of local to general character casting comports with McBeth and Lybecker’s (2018) findings when they traced the narratives from Breitbart to the speeches of GOP members on the House floor.


Table 2. Heroes in Sanctuary City Narratives (Pre- and Post-Shooting of Kathryn Steinle)

Hero % in Pre-Shooting Narratives % in Post-Shooting Narratives Change Mann–

Whitney U


Sig. (2-tailed)

Cohen’s d
Legislation 0.35 0.28 -0.07 460.5 0.565 0.149 (Small)
State Gov. Officials 0.26 0.2 -0.24 377.0 0.003* 0.716 (Medium)
Local Law Enforcement 0.26 0.14 -0.12 434.5 0.227 0.300 (Small)
ICE 0.22 0.30 +0.08 452.5 0.464 0.180 (Small)
Anti-Sanctuary Activists 0.17 0.40 +0.23 385.0 0.068 0.517 (Medium)
Congress 0.9 0.23 +0.14 422.5 0.147 0.384 (Small)
Donald Trump 0.0 0.23 +0.23 379.5 0.013* 0.762 (Medium)

*Significant at the p < 0.05 level.

  Of these changes, two are statistically significant. The first shift is the casting of state government officials as heroes in one-quarter of the anti-sanctuary narratives before the shooting of Steinle to a virtual absence in those stories following the shooting. The second is the emergence of Donald Trump as a key hero in post-shooting narratives from a complete absence in pre-shooting narratives. This transition also bears a striking resemblance to the focus on Obama as a villain in the GOP floor speeches (McBeth & Lybecker, 2018) and points to a narrative pattern of moving to establish the “right hero” and the “right villain” to construct an effective narrative to construct a focusing event. This finding shows that the policy entrepreneurs at the CIS echoed Donald Trump’s series of tweets in the aftermath of the shooting, which sought to acknowledge the tragedy and to use the event to call for a policy response and vilify opponents. On July 3, 2015, Trump tweeted: “My heartfelt condolences to the family of Kathryn Steinle. Very, very sad!” This was followed by a second tweet later that day attacking Marco Rubio (a competitor for the 2016 Republican presidential nomination): “@marcorubio what do you say to the family of Kathryn Steinle in CA who was viciously killed b/c we can’t secure our border? Stand up for US.” Then, pivoting to a policy statement in support of Kate’s Law (H.R. 3004, 2017), 10 days later Trump tweeted: “I absolutely support Kate’s Law—in honor of the beautiful Kate Steinle who was gunned down in SF by an illegal immigrant” (Trump Twitter Archive, 2020).

H2: Anti-sanctuary narratives will demonstrate less of a devil shift, and pro-sanctuary narratives more of a devil shift, following President Trump’s issuance of EO 13768.

      After the Trump administration issued its EO titled “Enhancing Public Safety in the Interior of the United States” (EO 13768), we noted several changes in the narrative environment and the narrative form employed by both sides of the debate. The number of pro-sanctuary narratives issued in the 85 months before EO 13768 (n = 31) averaged 0.36 per month, but this number spiked to 15.97 per month (n = 33) following it. Whereas the number of anti-sanctuary narratives in the pre-EO period (n = 56) averaged 0.65 narratives per month, this increased to 4.84 per month following the administration’s action (n = 10). Although the data are limited given the small sample size for anti-sanctuary narratives in the post-order period, the trend suggests that overall narrative production increased on both sides, but a much greater increase was seen on the pro-sanctuary side.

      Although we noted a quantitative change in the number of narratives, contrary to our hypothesis, we did not see any significant change in the use of the devil–angel shift following the issuance of EO 13768. Using the Kruskal–Wallis H test for pro-sanctuary narratives to test pre- and post-EO narratives (pre = 32.02, post = 32.95, H = 0.41, p = 0.840), we saw no evidence for our hypothesized increase in the angel shift. Likewise, the anti-sanctuary stories did not evince a statistically significant change (pre = 32.08, post = 41.45, H = 2.050, p = 0.152). This suggests that although a decisive shift in the policy arena resulted in each side increasing its overall narrative output, it did not seem to affect the use of either the angel or devil shift as a narrative strategy. Similarly, it did not affect the use of the solidarity shift by pro-sanctuary (pre = 32.55, post = 32.45, H = 0.000, p = .984) or anti-sanctuary sides (pre = 34.04, post = 30.55, H = 0.283, p = 0.595). Groups also did not alter their portrayal of federalism; no differences were seen between pro-sanctuary (pre = 31.21, post = 33.71, H = 0.348, p = 0.555) or anti-sanctuary stories (pre = 32.10, post = 33.71, H = 0.942, p = 0.332). This finding supports Gottlieb et al.’s (2018) and Chang and Koebele’s (2020) findings that the orientation toward a policy is more indicative of a narrative strategy than is a group’s positioning as either “winning” or “losing” in a policy contest (Shanahan et al., 2013).

      We did not observe differences in the devil–angel shift between pre- and post-EO 13768 anti-sanctuary narratives, but there were differences in the absolute number of villains and victims and the total number of characters in the narratives. As with the narrative variation associated with Steinle’s shooting, we saw significant differences emerge only in anti-sanctuary policy narratives. We saw villains in anti-sanctuary narratives shift from a mean of 5.54 villains per narrative to only 2.90 villains following EO 13768 (Mann–Whitney U 140.0, p = 0.007). Pre-EO 13768 anti-sanctuary narratives had a total mean use of 2.61 victims, whereas post-order narratives deployed a mean of 1.50 victims (Mann–Whitney U 167.5, p = 0.040). The mean number of all characters fell from 10.07 in pre-order narratives to 5.70 in anti-sanctuary post-order narratives (Mann–Whitney U 127.5, p = 0.006, Cohen’s d = 1.085). This might suggest that the need to push a narrative with more characters becomes (at least temporarily) less important following a substantive “victory” in a policy contest.

      Those narratives appearing before EO 13768 featured sanctuary jurisdictions (93% of narratives), undocumented immigrants (84%), criminals (77%), President Obama (57%), and local government officials (30%) as the most common villains. After the order, we found that sanctuary jurisdictions were characterized as villains in 70% of the narratives; criminals were portrayed as villains in 60%; undocumented immigrants in 50%; and Obama, local law enforcement, and local government officials each in 20%.


Table 3. Villains in Sanctuary City Narratives (Pre- and Post-EO 13768)

Villain % in Pre-EO


% in Post-EO


Change Mann–

Whitney U


Sig. (2-tailed)

Cohen’s d
Sanctuary Jurisdictions 0.93 0.70 -0.23 216.0 0.032* 0.593 (Medium)
Undocumented Immigrants 0.84 0.50 -0.34 185.0 0.016* 0.746 (Medium)
Criminals 0.77 0.60 -0.17 233.0 0.267 0.359 (Small)
Local Government Officials 0.30 0.20 -0.10 251.0 0.508 0.225 (Small)
Obama 0.57 0.20 -0.27 176.0 0.032* 0.801 (Large)
Local Law Enforcement 0.23 0.20 -0.03 271.0 0.825 0.071 (Small)

*Significant at the p < .05 level.

     After Trump signed the EO, there were significant changes in the frequency with which sanctuary jurisdictions, undocumented immigrants, and President Obama were featured as villains. The overall number of villains in the anti-sanctuary narratives fell, and three of the six most prevalent villains also demonstrated a statistically significant decrease.

      Narratives appearing before the EO showed that victims of crimes committed by undocumented immigrants (50% of narratives), the general public (41%), public safety (41%), ICE (16%), and taxpayers (14%) were the most common victim types. After EO 13768, victims of crimes committed by undocumented immigrants dropped as a class of victim to only 40% of the narratives, the general public increased to appearing in 50%, and families increased to appearing in 30%. We noted that public safety, ICE, the US Constitution, taxpayers, and the economy were all featured as victims in 20% of narratives.


Table 4 . Victims in Sanctuary City Narratives (Pre- and Post-EO 13768)

Victim % in Pre-EO Narratives % in Post-EO Narratives Change Mann-

Whitney U


Sig. (2-tailed)

Cohen’s d

(effect size)

Victims of Crimes 0.50 0.40 -0.10 252.0 0.563 0.194 (Small)
General Public 0.41 0.50 +0.10 255.0 0.602 0.176 (Small)
Families 0.13 0.30 +0.17 231.0 0.158 0.409 (Small)
Public Safety 0.41 0.20 -0.21 221.0 0.209 0.456 (Small)
ICE 0.16 0.20 +0.04 269.0 0.761 0.101 (Small)
Taxpayers 0.14 0.20 +0.06 264.0 0.645 0.154 (Small)
U.S. Constitution 0.02 0.20 0.18% 229.0 0.011* 0.575 (Medium)
The Economy 0.13 0.20 0.07% 259.0 0.528 0.184 (Small)

*Significant at the p < 0.05 level

H3: There will be differences between pro- and anti-sanctuary narratives concerning the use of the devil–angel shift.

      The statistics related to the devil–angel shift showed that pro-sanctuary narratives (n = 64) had only a slight devil shift (mean -0.0753). Anti-sanctuary narratives (n = 66) were found to have an average devil shift of -0.495, indicating these stories contained a higher ratio of villains to heroes than did the pro-sanctuary narratives. Using the Mann–Whitney U test, pro-sanctuary narratives were found to be significantly different from anti-sanctuary narratives in their use of the angel shift, with a considerably strong effect size (U = 974.0, z = -5.316, p = 0.00, Cohen’s d = 0.999). The direction of the effect and the statistical significance lend support to our hypothesis that anti-sanctuary groups would be more reliant on the use of the villain character in their policy narratives. This finding provides evidence that anti-sanctuary narratives present stories with a higher ratio of villains to other characters, which may help catalyze their strategy of employing a view of immigrants as criminals (e.g., García Hernández, 2013; Cházaro, 2016).

H4: There will be differences between pro- and anti-sanctuary narratives concerning the use of the solidarity shift.

       Our fourth hypothesis sought to determine whether differences were observable between policy narratives in the realm of sanctuary policy debates regarding the proportion of characters who were victims as opposed to heroes or villains. The mean value of the solidarity shift for pro-sanctuary narratives was 0.506 and for anti-sanctuary narratives was 0.609. The results of the Kruskal–Wallis H test indicated there were certainly differences between the narratives along this line (H = 7.93, p = 0.019), and further examination with Mann–Whitney U found there was a significant difference between pro- and anti-sanctuary narratives; however, the effect size was weak (U = 1643.5, z = -2.188, p = 0.029, Cohen’s d = -0.307). Again, we see that anti-sanctuary narratives employed a higher ratio of victims compared to pro-sanctuary narratives. This fact further plays into the strategy of depicting immigrants as predators who harm law-abiding US citizens. Indeed, the field of rhetorical battle could be interpreted as one dealing with competing conceptions of public safety as well as how the integration of federal, state, and local authorities can best be structured to aid law enforcement in achieving its goal of protecting the public from crime.

H5: Anti-sanctuary advocates will be more likely to cast the federal government as a hero.

      Given the importance of federalism in this issue and the generally recognized federal predominance in immigration policy, we hypothesized that anti-sanctuary narratives would cast the federal government (and its agents) as heroes. The mean value for pro-sanctuary narratives was 0.266 and for anti-sanctuary narratives was 0.697. The Kruskal–Wallis H test found that differences existed between the narratives (H = 11.934, p = 0.003), and anti-sanctuary narratives favored federal heroes by a wide margin. Using Mann–Whitney U, we found pro- and anti-sanctuary narratives differed from each other, and this difference was statistically significant, with a medium effect size (Becker, n.d.) on Cohen’s d (U = 1475.50, z = –3.070, p = 0.002, Cohen’s d = -0.0584).

      This finding is counter to recent scholarship, which found that partisanship generally was not predictive of any particular preferences associated with federalism (Rendleman & Rogowski, 2020, p. 17).[8] However, Rendleman and Rogowski (2020) noted large majorities of both Democrats (76.2%) and Republicans (79.1%) prefer subnational governments take the lead in law enforcement (p. 14). Interestingly, the opposite pattern was demonstrated in terms of foreign affairs, with only 22.6% of Democrats and 20.9% of Republicans preferring subnational dominance over this policy area. Sanctuary cities may represent a case where overall preferences for local control of law enforcement and federal dominance in foreign affairs collide. This dynamic is especially interesting when considered in tandem with another recent fight revolving around federalism concerns, the Affordable Care Act (ACA), where states with Republican governors sued the US Department of Health and Human Services over the individual mandate and the Medicaid expansion to claim states were victims of federal overreach (Smith-Walter, 2018). Instead, the argument we saw presented over sanctuary cities inverted the traditional positions of liberals and conservatives, with anti-sanctuary conservatives demanding that states toe the federal line on immigration policy and pro-sanctuary liberals arguing for subnational governments’ right to create their own communities free from federal meddling (Young, 2006).


      In this chapter, we set out to answer three major research questions and extend the work done by McBeth and Lybecker (2018) on the utility of the concept of focusing events (Birkland & Warnement, 2016) in the study of policy narratives on the emerging public controversy surrounding sanctuary cities. Specifically, these questions were as follows:

Research Question 1: To what extent do competing coalitions in the sanctuary city debate differ in their use of the angel shift and devil shift as narrative strategies?

Research Question 2: To what extent do competing coalitions differ in their use of the number of victims as a narrative strategy?

Research Question 3: To what extent do competing coalitions differ in their conception of federalism as a policy belief?

      Looking at the characteristics of anti-sanctuary narratives using the devil–angel shift, McBeth et al.’s (2010) measure of federalism, and the newly devised solidarity shift, we found that the murder of Kathryn Steinle was followed by a narrative movement toward federal-level actors as heroes by anti-sanctuary actors. We also found that anti-sanctuary narratives demonstrated a higher mean score (0.609) than pro-sanctuary groups (0.506) on the solidarity shift, and the difference was statistically significant (p = 0.29). This means that anti-sanctuary stories featured a higher number of victims relative to the other characters, although the pro-sanctuary narratives contained a higher average number of victims per narrative. When we couple this finding with the results of the devil–angel shift measure, we can begin to fully appreciate the value that the solidarity shift adds to the NPF’s ability to characterize policy narratives. For example, we see the anti-sanctuary narratives had a higher devil shift score, a tendency to feature federal-level actors as heroes, and a higher reliance on victims relative to the number of heroes and villains.

      Taken together, the “right innocent victim” in a narrative sense might mean appreciating the types of victims used in the narratives, the ratio of victims used relative to other characters, and an effective pairing of the victim with a suitable hero who is capable of effectively confronting the villain. As such, we should take the quality and quantity of victims employed in narratives quite seriously because doing so provides a more complete picture of the relationship among the three key character types used in NPF studies. The solidarity shift can help round out a system of summary statistics better captures the internal dynamics of policy narratives across policy subsystems.

       When exploring pro-sanctuary narratives, we found these narratives were more angel-shifted, less solidaristic, and more friendly toward local and state authorities. This contrasts with anti-sanctuary narratives, which were more devil-shifted, more solidarity-shifted, and more oriented toward federal authorities. Initially, this result might seem illogical. After all, are not sanctuary cities all about the harm done to undocumented immigrants? Should not victims feature more prominently? We saw that although pro-sanctuary narratives featured more victims than anti-sanctuary policy narratives (4.63 vs. 2.44), anti-sanctuary narratives had a higher proportion of victims compared to all characters. However, it is possible that the absolute number of victim characters may not be as important as the specific proportion of characters who are victims relative to heroes and villains.

       Additionally, we failed to see any major changes in the narrative strategy on the part of pro-sanctuary advocates following either Steinle’s shooting or the issuance of EO 13678. This suggests they were unable to effectively create a counternarrative to the one that anti-sanctuary advocates developed using Steinle’s death as a focusing event. This led us to suspect that dynamism surrounding policy narratives is something that should not be overlooked when conducting NPF studies. The movement from local to federal heroes in anti-sanctuary narratives is one indicator for which NPF scholars and analysts can look when attempting to identify casting in policy narratives that may result in the construction of a focusing event. Indeed, policy actors and interest groups might consider replicating the approach used in this chapter to create a conceptual map of the media space (both printed and digital) and the narratives circulating within it. This map could be updated with a certain regularity, made easier by developing auto-coding methods (see Wolton et al., in this volume) to gain insight into the occurring narrative adaptations. When particular narrative formulations make the leap from the media environment to the political agenda, with casting adaptations from the particular and the local to the general and the national, narrative scholars would do well to sit up and take notice.

       Future NPF research efforts might consider the creation of a system of narrative collection and analysis in vital policy areas; this system could be conceived as analogous to the monitoring of the spread of communicable diseases performed by epidemiologists. In other words, when the mutations within the set of characters slow appreciably and stabilize around a single basic assemblage of characters (the “right hero,” “right villain,” and “right innocent victim”), narrative scholars should be alert for the potential for the narrative to “jump species” from media and interest group communication to authoritative policy-making venues. This is especially applicable if the narrative features a recent transition to a hero (or heroes) who controls institutions with the authority to act on a particular policy controversy.

       Such a monitoring program might also explore whether the narratives produced by the opposing coalition in response to the emergent focusing event manifested an identifiable strategic response. One such strategic response might be the production of narratives intent on expanding the scope of conflict. NPF scholars could identify this strategy by paying attention to the proportion of heroes and victims (an angel shift, coupled with a moderate to large solidarity shift with a stable federalism measure) featured in their policy narratives. An alternative strategy might seek to create narratives featuring dangerous villains at other levels of government (a devil shift with a large solidarity shift and a federalism measure swinging away from the government level where the opposition chalked up its recent win), with the idea that shifting to alternative policy venues might provide a strategic advantage to “check” future political gains made by their opponents. In any event, it is important to stress that future research will benefit from mapping the ebb and flow of a narrative’s emphasis on all three primary characters contained in policy narratives.

       Although this study has sought to illuminate the differences in policy narrative deployed by various groups, it is subject to certain limitations. First is the limited diversity in the groups that authored anti-sanctuary narratives. The CIS featured all the narratives in our study, although multiple different contributors authorized it to do so. This may present a limitation on our findings’ generalizability to parts of the nation where the New York Times and the Boston Globe are not major media outlets. Another weakness concerns the fact that the areas we explored (i.e., Boston and New York City) have political cultures that are strongly aligned with the Democratic party. In addition, more than 28% of Boston’s population and 37% of New York City’s population are foreign-born (US Census, 2017a, 2017b). Many of the interest groups included in this study are national in scope, but the inclusion of media from metropolitan areas that do not have sanctuary ordinances and those with lower numbers of foreign-born residents should feature in future research.


Agility PR Solutions. (2020, January). Top 10 US newspapers by circulation.

Amdur, S. E. (2016). The right of refusal: Immigration enforcement and the new cooperative federalism. Yale Law & Policy Review, 35, 87–160.

Becker, L. A. (n.d.). Effect size calculators.

Bhatt, R. (2017). Pushing an end to sanctuary cities: Will it happen? Michigan Journal of Race & Law, 22, 139–162.

Birkland, T. A. (1998). Focusing events, mobilization, and agenda setting. Journal of Public Policy, 18(1), 53–74.

Birkland, T. A., & Warnement, M. K. (2016). Refining the idea of focusing events in the multiple-streams framework. In R. Zohlnhöfer & F. W. Rüb (Eds.), Decision-making under ambiguity and time constraints: Assessing the multiple-streams framework (pp. 0–0). ECPR Press.

Blair, B. D., & McCormack, L. (2016). Applying the narrative policy framework to the issues surrounding hydraulic fracturing within the news media: A research note. Research and Politics, 1–3.

Bragg, D. D., & Soler, M. C. (2017). Policy narrative on applied baccalaureate degrees: Implications for student access to and progression through college in the United States. Journal of Vocational Education & Training, 69(1), 123–146.

Brewer, A. M. (2019). A bridge in flux: Narratives and the policy process in the Pacific Northwest. Review of Policy Research, 36(4), 497–522.

Brown, B. (n.d.). Trump Twitter archive.

Cházaro, A. (2016). Challenging the “criminal alien” paradigm. UCLA Law Review, 594, 595–664.

Executive Order 13768. (2017, January 25). Executive order: Enhancing public safety in the interior of the United States.

García Hernández, C. C. (2013). Creating crimmigration. Brigham Young University Law Review, 1457–1515.

Gupta, K., Ripberger, J. T., & Collins, S. (2014). The strategic use of policy narratives: Jaitapur and the politics of siting a nuclear power plant in India. In M. D. Jones, E. A. Shanahan, & M. K. McBeth (Eds.), The science of stories: Applications of the narrative policy framework in public policy analysis (pp. 0–0). Palgrave MacMillan.

Heikkila, T., Weible, C. M., & Pierce, J. J. (2014). Exploring the policy narratives and politics of hydraulic fracturing in New York. In The science of stories: Applications of the narrative policy framework in public policy studies (pp. 0–0). Palgrave MacMillan.

IIRIRA. (1996). Illegal Immigration Reform and Immigrant Responsibility Act of 1996.

Johnson, K. R. (2016). Doubling down on racial discrimination: The racially disparate impacts of crime-based removals. Case Western Law Review, 66(4), 993–1037.

Jones, M. D. (2014). Cultural characters and climate change: How heroes shape our perception of climate science. Social Science Quarterly, 95(1), 1–39.

Jones, M. D., & McBeth, M. K. (2010). A narrative policy framework: Clear enough to be wrong? Policy Studies Journal, 38(2), 329–353.

Jones, M. D., & Radaelli, C. M. (2015). The narrative policy framework: Child or monster? Critical Policy Studies, 9(3), 339–355.

Jorgensen, P. D., Song, G., & Jones, M. D. (2018). Public support for campaign finance reform: The role of policy narratives, cultural predispositions, and political knowledge in collective policy preference formation. Social Science Quarterly, 99(1), 216–230.

Juárez, M., Gómez-Aguiñaga, B., & Bettez, S. P. (2018). Twenty years after IIRIRA: The rise of immigrant detention and its effects on Latinx communities across the nation. Journal on Migration & Human Security, 6(2), 74.

Kandel, W. A. (2016, September 8). Interior immigration enforcement: Criminal alien programs. Congressional Research Service, R44627.

Lawlor, A., & Crow, D. (2018). Risk-based policy narratives. Policy Studies Journal, 46(4), 843–867.

Lebel, L., & Lebel, B. (2018). Nexus narratives and resource insecurities in the Mekong region. Environmental Science & Policy, 90, 164–172.

Littlefield, C. (2015, July 24). Sanctuary cities: How Kathryn Steinle’s death intensified the immigration debate. LA Times.

Manuel, K. M. (2015). Immigration detainers: Legal issues. Congressional Research Service.

McBeth, M. K., Clemons, R. S., Husmann, M. A., Kusko, E., & Gaarden, A. (2014a). The social construction of a crisis: Policy narratives and contemporary U.S. obesity policy. Risk, Hazards & Crisis in Public Policy, 4(3), 135–163.

McBeth, M. K., Jones, M. D., & Shanahan, E. A. (2014b). The narrative policy framework. In C. M. Weible & P. Sabatier, Eds., Theories of the policy process (3rd ed., pp. 0–0). Westview Press.

McBeth, M. K., & Lybecker, D. L. (2018). The narrative policy framework, agendas, and sanctuary cities: The construction of a public problem. Policy Studies Journal, 46(4), 868–893.

McBeth, M. K., Shanahan, E. A., Arnell, R. J., & Hathaway, P. L. (2007). The intersection of narrative policy analysis and policy change theory. Policy Studies Journal, 35(1), 87–108.

McBeth, M. K., Shanahan, E. A., Hathaway, P. L., Tigert, L. E., & Sampson, L. J. (2010). Buffalo tales: Interest group policy stories in Greater Yellowstone. Policy Sciences, 43(4), 391–409.

Merry, M. K. (2016). Constructing policy narratives in 140 characters or less: The case of gun policy organizations. Policy Studies Journal, 44(4), 373–395.

Merry, M. K. (2019). Angels versus devils: The portrayal of characters in the gun policy debate. Policy Studies Journal, 47(4), 882–904.

Offutt, L. (2018, August 2). Federal appeals court: Executive order denying funding to ‘sanctuary cities’ unconstitutional. Jurist.

Qualtrics. (2021, March 3). Sample size calculator.

Rendleman, H., & Rogowski, J. C. (2020). Americans’ attitudes toward federalism.

Rodriguez, C. (2017). Enforcement, integration, and the future of immigration federalism. Journal on Migration and Human Security, 2, 509.

Sabatier, P. (2000). Clear enough to be wrong. Journal of European Public Policy, 7(1), 135–140.

Sabatier, P., Hunter, S., & McLaughlin, S. (1987). The devil shift: Perceptions and misperceptions of opponents. The Western Political Quarterly, 40(3), 449–476.

Schneider, A., & Ingram, H. (1993). Social construction of target populations: Implications for politics and policy. The American Political Science Review, 87(2), 334–347.

Shanahan, E. A., Adams, S. M., Jones, M. D., & McBeth, M. K. (2014). The blame game: Narrative persuasiveness of the intentional causal mechanism. In M. D. Jones, E. A. Shanahan, & M. K. McBeth (Eds.), The science of stories: Applications of the narrative policy framework in public policy analysis (pp. 0–0). Palgrave MacMillan.

Shanahan, E. A., Jones, M. D., & McBeth, M. K. (2011). Policy narratives and policy processes. Policy Studies Journal, 39(3), 535–561.

Shanahan, E. A., Jones, M. D., McBeth, M. K., & Lane, R. R. (2013). An angel on the wind: How heroic policy narratives shape policy realities. Policy Studies Journal, 41(3), 453–483.

Shanahan, E. A., Jones, M. D., McBeth, M. K., & Radaelli, C. M. (2017). The narrative policy framework in theories of the policy process (4th ed., C. Weible, Ed.). Westview Press.

Smith-Walter, A. (2018). Victims of health-care reform: Hirschman’s rhetoric of reaction in the shadow of federalism. Policy Studies Journal, 46(4), 894–921.

Smith-Walter, A., Peterson, H. L., Jones, M. D., & Marshall, A. N. R. (2016). Gun stories: How evidence shapes firearm policy in the United States. Politics & Policy, 44(6), 1053–1088.

Stauffer, B. (2020). Culture, institutions, focusing events and their effect on narratives – Differences in the public controversy on the Swiss Child and Adult Protection Authorities [Manuscript in preparation].

U.S. ICE. (2017, May 19). Secure Communities.

Van de Vyver, J., & Abrams, D. (2015). Testing the prosocial effectiveness of the prototypical moral emotions: Elevation increases benevolent behaviors and outrage increases justice behaviors. Journal of Experimental Social Psychology, 58, 23–33.

Varsanyi, M. W., Lewis, P. G., Provine, D. M., & Decker, S. (2012). A multilayered jurisdictional patchwork: Immigration federalism in the United States. Law & Policy, 34(2), 138–158.

Voting by Aliens, 18 U.S.C. § 611 et seq. (1996).

U.S. Census Bureau. (2017a). Quick facts – Boston City, Massachusetts.

U.S. Census Bureau. (2017b). Quick facts – New York City, New York.

Young, E. A. (2006). The conservative case for federalism. George Washington Law Review, 74(5), 874–887.


Chapter Appendixes

Appendix A. Intercoder Reliability Statistics

Code Percent Agree Krippendorf’s Alpha (nominal) N Disagree N
Policy Solution 93.30% 0.923 33 492
Stance of Narrator on Sanctuary Cities 98.20%








American Values 99.20% 0.746 4 492
Anti-Sanctuary Advocates 98.40% 0.847 8 492
Asylum Seekers 100% 1.0 0 492
Congress 95.10% 0.738 24 492
Criminal Justice System 99.40% 0.931 3 492
Criminals 98% 0.92 10 492
DACA Recipients 100% 1 0 492
Democrats 100% 1 0 492
DHS 98.60% 0.884 7 492
Domestic Violence Service Providers 99.80% 0.857 1 492
Economy 99.20% 0.941 4 492
Elderly 99.80% 0.94 1 492
Executive Order 98.00% 0.88 10 492
Faith Leaders 99.80% 0.94 1 492
Families 98.80% 0.90 6 492
Federal Agencies or Officials 99.40% 0.959 3 492
Health Care System 100% 1 0 492
Human Rights 100% 1 0 492
ICE 97.80% 0.896 11 492
Immigrants 98.40% 0.928 8 492
Immigration Reform 99.60% 0.749 2 492
Law Enforcement 99.60% 0.922 2 492
Law Makers 98.00% 0.743 10 492
Legal Immigrants 99.80% 0.922 1 492
Legislation 96.10% 0.862 19 492
Local Government or Officials 95.90% 0.875 20 492
Local Law Enforcement 98.40% 0.947 8 492
Muslims 99.80% 0.94 1 492
Non-Immigrant Visa Holders 100% 1 0 492
Non-Sanctuary Jurisdictions 100% 1 0 492
Obama 98.60% 0.902 7 492
Other 98.50% 0.819 15 984
Poor People 100% 1 0 492
Powerful People 100% 1 0 492
President Donald Trump 98.40% 0.931 8 492
Prisons-Jails-Corrections 99.80% 0.888 1 492
Pro-Sanctuary Advocates 98.60% 0.911 7 492
Protestors 100% 1 0 492
Public Safety 98% 0.912 10 492
Racial Ethnic Groups 99.80% 0.856 1 492
Refugees 99.80% 0.951 1 492
Republicans 99.80% 0.959 1 492
Researchers 99.60% 0.666 2 492
School District 99.80% 0.922 1 492
Stance of Narrator on Sanctuary Cities 98.20% 0.972 3 164
State Government or Officials 97.60% 0.848 12 492
Supreme Court 100% 1 0 492
Taxpayers 99.40% 0.92 3 492
Technology 100% 1 0 492
The Constitution 98.20% 0.844 9 492
Undocumented-Illegal-Aliens 98.20% 0.946 9 492
U.S. Immigration Services 99.40% 0.931 3 492
Veterans 99.80% 0.666 1 492
Victims of Crime Committed by Undoc. Immigrants 98% 0.881 10 492


[1] A detailed discussion of narrative relativity is outside the scope of this chapter, but see Shanahan et al. (2018, pp. 175–178).

[2] We will not explore the movement from the particular to the general that McBeth and Lybecker (2018) laid out; however, we will propose a measure that can allow NPF scholars to more fully incorporate the complete traditional character triad in future NPF analyses.

[3] It was also in Shanahan et al. (2013) that the devil shift’s positive pole was labeled the angel shift, creating a measure that associated hero-heavy narratives with a positive strategy. The formulation reported in Heikkila et al. (2014) was the actual measure computed by Shanahan et al. (2013, as cited in Heikkila et al., 2014, p. 202, endnote 9).

[4] However, Heikkila et al. (2014) also noted the possibility that even by taking all heroes and villains into account with the corrected angel and devil shift, the measure may be too blunt, and it lacked the dimension of power—a key conceptual component of the devil shift (p. 200) and an important strand of the discussion that Merry (2009) would later address. Merry’s identification of an actor’s power, in addition to evilness, as an underplayed aspect of the devil shift is important. Although this chapter does not adopt her recommendation for using Schneider and Ingram’s (1993) social construction of target populations as a means for categorizing characters as weak or powerful, the solidarity shift measure could be adapted easily to the approach detailed in her 2019 piece.

[5] While the number of victims compared across gun rights and gun control coalitions was not reported in the text of the findings section, Figure 1 of Merry (2019, p. 894) indicated there were numerical differences in the two coalitions, with gun control groups using more victims than gun rights groups.

[6] These organizations were primarily the professional organizations of local and state law enforcement officers who took pains not to comment on policy but opted to concentrate on the danger that local involvement in immigration enforcement would have on community policing efforts. As such, we removed those narratives from this analysis.

[7] [i] Connecticut, Delaware, and New Mexico allow noncitizens to vote in municipal elections.

[8] When examining ideology instead of partisanship, Rendleman and Rogowski (2020) found that those individuals with higher scores on a measure of liberal ideology preferred a stronger federal government, and those with a more conservative ideology preferred greater subnational governmental power; however, the nature of this study and that of Smith-Walter (2018) is such that effectively disentangling ideology and partisanship based on analyzing policy narratives is impossible.


Icon for the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License

Narratives and the Policy Process: Applications of the Narrative Policy Framework by Michael D. Jones; Mark K. McBeth; and Elizabeth Shanahan is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, except where otherwise noted.

Digital Object Identifier (DOI)

Share This Book