Abstract
I examine how social influence shapes both the adoption and implementation of 287(g) agreements—federal-local contracts that deputize local law enforcement to enforce immigration law. Drawing on organizational theory, the article highlights the role of public official associations (POAs) in diffusing interest in these agreements across counties. Using data on application letters and quasi-experimental methods, the study shows that counties with stronger ties to POAs were more likely to express interest in 287(g) agreements and to imitate language used by peer jurisdictions. It then links this imitation to variation in enforcement intensity: Counties that closely copied others’ applications were more likely to escalate their institutional commitment, including devoting more jail space to Immigration and Customs Enforcement (ICE) detainees, complying with ICE detainer requests at higher rates, and becoming more central to ICE’s detainee transfer network. This suggests that social influence may shape not only entry into immigration enforcement but also its long-term implementation.
Discussions of immigration enforcement in the United States bring to mind federal agencies such as US Customs and Border Protection (CBP) and Immigration and Customs Enforcement (ICE). Yet without state and local government cooperation, current enforcement levels could not continue. During recent administrations, for example, the majority of ICE interior arrests originated from local transfers (Capps et al. 2018). Hundreds of local jails confine noncitizens for ICE (Ryo and Peacock 2020). In certain jurisdictions, local police officers can begin noncitizens’ removal processes or recommend certain legal remedies in lieu of formal removal proceedings (American Immigration Council 2021).
Research on this phenomenon has bifurcated into two bodies of scholarship. One aims to explain the incidence and intensity of local government involvement, linking participation to the demographic, economic, and political features of the participating localities. The second aims to understand the consequences of such enforcement for noncitizens, their families, and the communities in which they live. This study considers the connections between these findings by asking whether motivations for participating in immigration enforcement influence the consequences of such participation.
I answer this question by examining 287(g) agreements, named after Section 287(g) of the US Immigration and Nationality Act (INA). These agreements allow the federal government to formally delegate immigration enforcement functions to local law enforcement. State and local governments apply to ICE to receive training and authorization to participate in immigration enforcement and, should ICE approve the application, sign agreements with ICE formalizing the arrangement. While a now burgeoning body of scholarship examines 287(g) agreements and the participation of police in immigration enforcement, very little of this research documents heterogeneity in the implementation of immigration enforcement or provides an explanation for such heterogeneity. This is a notable oversight given that the Trump administration has sought not only to revitalize the 287(g) program, but also to deputize a wide range of actors beyond local police, such as Drug Enforcement Agency officers (Hackman and Gurman 2025) and military personnel (Rodriguez 2025), to enforce immigration laws. Theory gives us little guidance on what might drive variation in how these actors interpret and implement immigration enforcement in this critical moment.
I address this gap by examining how the motivations that drive local participation—particularly those rooted in social influence—shape the intensity and character of immigration enforcement once agreements are in place. This study focuses on social influence as a motivation for counties’ participation in immigration enforcement. It is the first to identify public official associations (POAs), such as the National Sheriffs’ Association (NSA) and the Major County Sheriffs of America (MCSA), as central to creating interest in immigration enforcement among county governments. Using archival sources never before analyzed, including sheriff association publications, conference materials, and internal communications, I find that many counties’ decisions to participate were not isolated processes driven entirely by local concerns. POAs contributed to the spread of 287(g) agreements by serving as venues wherein others, such as federal authorities or issue entrepreneurs, influence counties vis-à-vis immigration enforcement. I use regression discontinuity and text reuse methods with never-before-analyzed data to show that POA participation shaped the timing and manner in which US counties first began to show interest in 287(g) agreements.
Next, the study links this social motivation for participating in 287(g) agreements to the agreements’ effects. I use difference-in-differences and regression discontinuity methods that leverage geographic and temporal variation in 287(g) agreement applications to show that counties entering 287(g) agreements with higher text reuse (a strong proxy for social influence) became more involved in immigration enforcement—even beyond the scope of 287(g) agreements—than counties with lower levels of text reuse. This includes dedicating more jail space to ICE detainees, complying with ICE detainer requests at higher rates, and becoming more central in the network of correctional facilities through which ICE shuttles noncitizens. These changes suggest that 287(g) agreements establish new organizational stakes in public offices. This potentially creates political feedback loops and further escalates commitment to immigration enforcement.
This study makes several contributions. First, it expands the literature on the diffusion of immigration enforcement policies by highlighting the role of nonpartisan professional associations—rather than partisan or ideological actors—in motivating local participation. Second, it advances understanding of the consequences of immigration enforcement by showing that variation in the intensity of implementation is shaped by the motivations underlying policy adoption. Third, it contributes to growing scholarship on sheriffs by showing how professional networks and supralocal affiliations shape local enforcement choices and institutional commitments. Fourth, it extends the study of intergovernmental agreements by showing how participation in programs like 287(g) can generate new institutional stakes and reshape local priorities—insights that apply beyond immigration to other domains of federalism by contract.
RESEARCH ON SUBFEDERAL IMMIGRATION ENFORCEMENT
This section first reviews research providing explanations for subfederal government participation in immigration enforcement. It then reviews research on the consequences of such participation.
Explanations for Subfederal Government Participation in Immigration Enforcement
As this issue’s introduction discusses, legislation in the 1980s and 1990s laid the groundwork for subfederal immigration policymaking (see Patler and Jones 2025, this issue). The Illegal Immigration Reform and Immigrant Responsibility Act (IIRIRA), notably, added Section 287(g) to the INA in 1996, allowing federal immigration authorities to delegate certain enforcement functions to states and localities through agreements with local law enforcement (DHS 2010). Through these agreements, deputized local law enforcement officers may identify and process noncitizens for removal in state and local correctional facilities, during the course of their regular duties, or both (DHS 2010, 20). Amid concerns over inconsistent practices and racial profiling, the Barack Obama administration curtailed 287(g) agreements, but the Trump administration revitalized them (Rosenberg and Levinson 2017). Currently, sixty agreements remain active (ICE 2025).
The creation of programs to encourage federal-local collaboration, nonetheless, does not explain why, when, and where localities choose to participate. To address these questions, researchers have identified many predictors of engaging in immigration enforcement under 287(g) arrangements or similar policies. Here, it is worth noting that sheriffs and their offices receive no funding to run 287(g) agreements but rather must independently support the agreements themselves. Thus, social motivations are likely a key explanation for entry into such arrangements. Consistent with this, studies often focus on local issues and characteristics as determinants. Scholars argue that desires to maintain social control in light of perceived racial, cultural, or other threats amid demographic changes are motivators (for example, Creek and Yoder 2012; Ryo and Peacock 2020; Walker and Leitner 2011). Similarly, local employers may see restrictive policies targeting noncitizens as a useful mechanism by which they can selectively threaten and control undocumented workers (Gomberg-Munoz and Nussbaum-Barberena 2011). Some homeowners may see them as a mechanism to curb the expansion of local low-income housing (Brettell and Nibbs 2011; Walker and Leitner 2011). Struggling economies may pursue jail contracts with ICE in exchange for per diem revenues (Ryo and Peacock 2020). Politicians facing public dissatisfaction may opportunistically blame immigrants for local problems and propose cooperation with ICE as a solution (Creek and Yoder 2012).
Other research suggests that supralocal factors channel local features to take on specific effects. Daniel Hopkins (2010) argues that national discourse shapes local responses to demographic change. Others have argued that ideological framing can determine the nature of a locality’s policymaking vis-à-vis noncitizens (Chavez and Provine 2009; Steil and Vasi 2014). Beyond political discourse, scholars have claimed that supralocal institutions explain why localities participate in immigration enforcement. Karthick Ramakrishnan and Pratheepan Gulasekaram (2013) maintain that political parties, particularly the Republican Party, and issue entrepreneurs have successfully leveraged post-9/11 ethnic nationalism to catalyze localities into prioritizing the policing of noncitizens. The relationship between different levels of government within parties may also matter. Lina Newton (2012) suggests that local governments enact restrictive policies not to address local issues but rather to send a signal of vertical integration with (or resistance against) the federal government and the incumbent executive. Similarly, federal or state public officials may promote local immigration policymaking in hopes that building pressure for action gets released when local governments adopt their own solutions. This model of steam valve federalism allows higher-level governments to avoid the pressure from lower-level governments to create blanket federal or state-level policies and, thereby, avoid fallout from potentially unpopular policies (Creek and Yoder 2012).
Strategic activism complements institutionalized politics to shape local immigration enforcement participation. Activist networks create channels to promote local involvement in immigration enforcement, such as actions by the American Legislative Exchange Council (Collingwood et al. 2019) or the Minuteman Project (Steil and Vasi 2014), or create opposition to such involvement, such as organizations affiliated with the UnidosUS (Steil and Vasi 2014). These forms of advocacy often involve framing immigration enforcement in ways that are locally resonant (Ramakrishnan and Gulasekaram 2013; Steil and Vasi 2014) and “targeting … jurisdictions with partisan conditions … ripe for enacting such regulation” (Ramakrishnan and Gulasekaram 2013, 1445).
While research on supralocal sources of influence offers many insights, one question that remains unanswered is what role, if any, organizations outside of partisan and movement politics play in the spread of subfederal forms of immigration enforcement. Answering this question is important because of the prevalence of nonpartisan organizations in the US and because such organizations may hold keys to explaining why localities that differ greatly in their sociopolitical makeup have converged on similar immigration enforcement policies. My study directly addresses this shortcoming by elaborating on public official associations and their contribution to the spread of immigration enforcement policies.
Consequences of Participation in Immigration Enforcement
Once localities begin to participate in immigration enforcement, research shows that such involvement has far-reaching consequences. As other articles in this issue highlight, local participation in immigration enforcement leads to significant declines in the physical and mental health of noncitizens, disrupts their life decisions and economic stability, and erodes trust in local law enforcement and civic engagement (for example, see Bennett et al. 2025, this issue; Hong et al. 2025, this issue; Kirksey and Sattin-Bajaj 2025, this issue; and Patler and Jones 2025, this issue). Involvement in immigration enforcement also has unintended consequences among those charged with enforcing the law. Amada Armenta (2017) shows how officers deputized under 287(g) arrangements increasingly act as arms of the federal government rather than as independent agents and lose focus on other services that they might provide to a community. Mai Thi Nguyen and Hannah Gill (2016) likewise argue that 287(g) distorts local law enforcement priorities. Moreover, Katharine M. Donato and Leslie Ann Rodriguez (2014) find that 287(g) programs altered officer discourse about immigrants and the reasons for which officers claimed to arrest them. Research also notes how 287(g) agreements, or similar policies such as Arizona’s “show me your papers” law, can lead to increased racial profiling (Jones-Correa and Fennelly 2009; Coon 2017).
While scholarship has documented these and other consequences of localities’ involvement in immigration enforcement, little research focuses on the heterogeneity in such consequences. Given the striking nature of the trends associated with 287(g) agreements and the fact that such agreements entail federal oversight, one might expect 287(g) agreements to have uniform consequences. But Mathew Coleman’s (2012) study of neighboring North Carolina counties that had both entered into 287(g) agreements suggests otherwise. The Durham Police Department used the program as a tool for gang suppression and only checked the immigration status of those arrested for felony offenses. The Wake County Sheriff’s Office, on the other hand, used the program as grounds to begin screening all immigrants, regardless of criminal history. The effects of the policies in the two neighboring counties diverged considerably. Wake County deported several thousand people during the same period that Durham County only deported several dozen individuals. Likewise, in a study of a federal-local program similar to 287(g) agreements, Daniel E. Chand and William D. Schreckhise (2015) find that partisan politics notably predicted the number of deportations by county. These findings suggest that identical policies can produce heterogeneous outcomes and be mediated by social dynamics.
Understanding the conditions under which the implementation of similar policies diverges, therefore, provides a clear area to extend research. This is especially the case given the decentralized nature of the US law enforcement system (Lowatcharin and Stallmann 2017) and the potential for the loose coupling between policies, formal structure, and the implementation of policies in organizations (Weick 1976). This study contributes to a greater understanding of heterogeneity in the effects of subfederal immigration enforcement by connecting the effects of 287(g) agreements back to the process by which interest in such agreements spread. This study is the first to raise the question of how motivations at the time of adoption might influence the implementation of 287(g) agreements.
AN ORGANIZATIONAL THEORY FOR THE SPREAD AND EFFECTS OF IMMIGRATION ENFORCEMENT
To develop a theory of nonpartisan, supra-institutional determinants of 287(g) adoption, I draw on the insight from organizational scholarship arguing that policies gaining social acceptance are key to understanding their spread. As Beth A. Simmons and colleagues (2007) discuss, social acceptance occurs in at least three ways relevant to this study. First, organizations draw inferences about the appropriateness of policies based on their sociocultural relationship with other organizations. This process of policy gaining acceptance relies on actors looking to others and often implicitly theorizing about what types of organizations should adopt what kinds of policies (Simmons et al. 2007). Thus, to determine whether some course of action is acceptable, decision-makers may look to those with whom they are structurally equivalent (Burt 1987, Elkins et al. 2006), psychologically proximate (Waltman 1980), or otherwise connected (Davis 1991, Sikkink 1993).
A consideration that arises from this point is the importance of organizations having awareness of precisely which other organizations are relevant in the first place. David Strang and John Meyer (1993, 491) argue that cultural categories are important for this purpose, insofar as they provide a “cognitive map [that] identifies reference groups that bound social comparison processes.” Categorization draws the boundaries of “the organizational field,” wherein potential policy adopters come to conceive of themselves as fundamentally similar (DiMaggio and Powell 1983; Strang and Meyer 1993).
Second, organizations look to exemplars in deciding whether to adopt policies. Organizations often play follow-the-leader by mimicking organizations that appear more successful (Haveman 1993). As outside observers, it becomes difficult to untangle the causal roots of another organization’s success. Followers may, accordingly, copy anything they assume contributes to desired outcomes, even if it means mimicking other organizations ritually and erroneously (Bennett 1991; Walker 1969).
Third, rather than looking to peer or leader organizations, organizations may also get information about the acceptability of policies based on information from field-specific experts and authorities. Experts theorize the effects of a policy and, in doing so, provide reasons for others to adopt it (Simmons et al. 2007). Pamela Tolbert and Lynne Zucker (1996, 183) argue that theorization involves two major tasks: identifying a general organizational failing for which a new policy or practice is a solution, and justifying the new solution. Scholars often point to the role of epistemic communities, which are networks of experts and professionals who define relevant issues and interests among actors in a policy arena (Haas 1992), in theorizing policy solutions and associated norms that lead to different organizations converging on similar solutions (DiMaggio and Powell 1983). Of course, organizations may use expert theorization as post hoc justifications rather than as motivations for policy decisions (Briscoe and Murphy 2012). In either case, expert theorization provides decision makers with a resource.
These points provide perspective for understanding how policies gain acceptance and diffuse among local governments. They highlight three mechanisms central to the spread of policies like 287(g) agreements: first, localities adopt policies by socially comparing themselves to similar or connected jurisdictions; second, localities imitate those they perceive as successful exemplars, sometimes even without clear evidence of effectiveness; and third, policy adoption is influenced by external experts who provide justifications and rationales that legitimize certain policy choices. The key implication for immigration enforcement is that local decisions about participating in programs like 287(g) may be driven significantly by these forms of social influence rather than solely by local needs or ideological motivations. Thus, the diffusion of immigration enforcement policies can occur broadly and rapidly—even across politically and demographically diverse localities—through structured organizational relationships, imitation of peers, and the theorizing activities of recognized experts or authorities.
Once policies have spread, the question remains of what effect they will have. Because the primary interest of this study is how law enforcement officers alter their behavior as a consequence of participating in immigration enforcement, I focus on consequences pertaining to law enforcement activity. The behavioral change of interest to this study is the escalation of commitment to immigration enforcement. While no study has explicitly considered this type of escalation as an outcome of localities participating in immigration enforcement, part of the variation in escalation will connect back to motives for adoption, including social motives. Consistent with this, research suggests that organizational and environmental factors moderate the strength of coupling between policy and implementation (Bromley and Powell 2012). Two relevant insights of this research are that the internalization of external influences can create tight coupling of focal policy and practices, even at the cost of disrupting other core organizational goals (Sauder and Espeland 2009), and that “motivations and skills that exist at adoption” create the contours for policy implementation insofar as such conditions “provide a form of imprinting that shapes subsequent … performance” (Weber et al. 2009, 1325).
Klaus Weber and colleagues (2009) provide a noteworthy example of these points in their study of the diffusion of national stock exchanges. They argue that those countries that created stock exchanges in response to coercive pressure from the International Monetary Fund and World Bank were more likely to ceremonially adopt stock exchanges, since such countries’ internal motivation and capacity for adoption had considerable limits. Countries that created stock exchanges as an act of social emulation, on the other hand, tended to have more material adoption, given higher levels of internal motivation and the possibilities of learning through observation of prior adopters. The extensiveness of implementation, accordingly, was a function of motives for policy adoption. Considering these insights, we might expect those localities that entered 287(g) agreements because of social influence to show the greatest escalation of commitment to immigration enforcement.
In short, this section develops the linkage between the spread of immigration enforcement policies and their subsequent implementation. Specifically, localities’ motivations at the point of policy adoption, particularly those driven by social influences such as peer imitation, exemplars, and expert theorization, have implications beyond mere diffusion. These initial motivations may imprint on organizations, shaping how intensively and durably they commit to enforcing immigration policies. Thus, localities influenced by professional peer networks and perceived policy legitimacy may not only adopt immigration enforcement practices more readily but also enforce them more vigorously and consistently. This insight provides the basis for the empirical analysis that follows, in which I examine whether social motivations behind adopting 287(g) agreements are associated with measurable differences in outcomes.
THE EMPIRICAL CASE: 287(G) AGREEMENTS AND SHERIFFS’ ASSOCIATIONS
This section develops the empirical case by identifying the institutional mechanisms through which influence operates, motivating the research design, defining key measures, outlining the empirical strategy, and presenting results that test the theory’s observable implications.
Public Official Associations and the Spread of Subfederal Immigration Enforcement
The diffusion mechanisms outlined in the previous section—social comparison, mimicry of exemplars, and expert theorization—operate through concrete organizational settings. In the case of 287(g) agreements, public official associations (POAs) like the NSA and the MCSA structured the environment in which sheriffs encountered these influences. This subsection draws on archival sources to show how these associations shaped the field of immigration enforcement: identifying relevant peers, circulating policy models, and staging encounters with purported experts.1 By examining the informational materials, events, and interpersonal connections fostered by these POAs, I show how they contributed to the spread of interest in 287(g) agreements and framed immigration enforcement as a professional norm. These patterns motivate the quantitative research design that follows, which tests whether counties exposed to POA influence were more likely to adopt 287(g) agreements and escalate their involvement in federal enforcement efforts.
For some context, POAs are national organizations whose primary membership consists of public officials and “whose purpose includes advancement of the general public interest [and] the special interests of the members” (Plant and Arnold 1993, 212). The organizational form of the POA originated in the first half of the twentieth century. POAs came in response to Progressive Era concerns about corruption and incompetence of local public offices and found a second purpose when public officials needed to effectively coordinate in response to policies of the New Deal Era that shifted ideals of intergovernmental relations from layer-cake forms of federalism, with neatly delineated authority corresponding to different levels of government, to marble-cake federalism, wherein different levels of governments have interlaced responsibilities (Plant and Arnold 1993; Wright 1990).
The activities of POAs grew to reflect these concerns. POAs allowed for the establishment of general professional standards, sharing best practices and strategies, interfacing with higher levels of government, business leaders, researchers, and other experts, and providing opportunities for professional coordination nationwide (Hertel-Fernandez 2019; Plant et al. 2010; Teaford 2002). Participation in the associations signaled a commitment to modern ideals of professionalism, expertise, efficiency, and “scientific management” (Teaford 2002, 6). The locus of assembly was, consequently, not to be ideological but instead professional. POAs, even more recently formed ones, continue to boast commitments to expertise and professionalism over ideology today.
POAs’ networks and resources also, importantly, situate them well to contribute to the acceptance of new practices and policies. POAs draw boundaries around specific sets of public officials such that POA members come to see other POA members as a relevant reference group for comparison and learning (Shipan and Volden 2012). POAs facilitate the display of exemplars by circulating publications and providing workshops to learn about policies of other public officials (Plant and Arnold 1993). POAs also link public officials to experts who can provide rationales to legitimize policies (Teaford 2002).
For a concrete example, let us consider the NSA. The association draws the boundaries around a specific set of actors whom sheriffs come to see as members of the same field, namely, other sheriffs. Thus, rather than looking to other heads of law enforcement agencies, such as police chiefs or directors of federal agencies, or to other county officials—such as supervisors or managers—NSA conferences, publications, and other proceedings help sheriffs to see each other as the relevant reference group to whom they look for cues and make comparisons. NSA activities reinforce the conception of sheriffs’ offices as fundamentally similar and relevant to each other.
Furthermore, the NSA creates ample opportunity for social learning and mimicry of exemplars. The NSA publishes the bimonthly Sheriff magazine, recurring reports, brochures, and other materials that allow sheriffs to stay abreast of policies and practices implemented by other sheriffs. A 2007 issue of Sheriff, for example, was entirely dedicated to the topic of immigration enforcement. In this issue, Jim Pendergraph (2007)—the sheriff of Mecklenburg County, North Carolina—penned an essay titled “What the Section 287(g) Program Can Do for Your Community,” wherein he described the implementation of a 287(g) agreement in his county and touted the agreement’s alleged benefits. The NSA also hosts multiple national conferences on a yearly basis. Sheriff Pendergraph, for example, mentioned learning about 287(g) agreements from another sheriff at one such NSA conference in 2005 (Caldwell et al. 2009). Besides allowing for peer-to-peer networking, such conferences also feature a wide range of seminars, workshops, exhibit rooms, committee meetings, and other events. In 2007, for example, the NSA held a conference wherein the association’s immigration subcommittee led a discussion about issues related to immigration enforcement, joined by sheriffs from across the country (McKain and McKain 2007b). As part of that conference, the NSA hosted a workshop called “The Cost of Crime: Dealing with Illegal Immigration” (NSA 2007). Later conferences offered workshops explicitly on 287(g) agreements.
The NSA also creates opportunities for sheriffs to get information about the acceptability of different immigration enforcement policies from actors who are ostensibly field-specific authorities and experts. Following the 9/11 attacks, for example, Robert Mueller (2002), then head of the Federal Bureau of Investigation, gave an NSA keynote highlighting areas of improvement for coordination between local and federal authorities for homeland security purposes. This was a similar theme—often with an explicit discussion of immigration enforcement—emphasized by subsequent NSA conference keynote speakers coming from the worlds of law enforcement, politics, and academia.2 Interaction with such actors could extend beyond giving keynotes at NSA conferences. For example, John Clark (2006), the first deputy assistant secretary for ICE, offered an NSA seminar on immigration issues affecting sheriffs and opportunities for cooperation with ICE.
While many of the keynote speakers were figures whom one might reasonably classify as experts, it is not clear what criteria the NSA applied to allow outsider participation in workshops and seminars. Investigative reporting has noted how nativist organizations such as the Federation for American Immigration Reform and the Center for Immigration Studies (CIS) made concerted efforts, starting during the George W. Bush administration, to steer the NSA and its members toward participating in immigration enforcement (Vazquez 2017). At one NSA conference, for example, a CIS representative touted the benefits of 287(g) agreements using a report based on selective statistics and misleading interpretations of academic research (Vaughn 2012a).
In short, the preceding discussion demonstrates the potential of POAs to play an active role in shaping sheriffs’ perceptions of immigration enforcement. As highlighted, the NSA certainly showed this potential through reference-group definition, policy exemplar promotion, and expert theorization. These mechanisms offer clear implications for empirical measurement: if POAs significantly influenced sheriffs, then counties with stronger POA ties should display observable patterns consistent with social influence, specifically, higher initial interest and greater imitation in their decisions to enter immigration enforcement. Likewise, consistent with the second aspect of the theory developed earlier, those counties that enter into immigration enforcement responding to social influence may ultimately exhibit a more intense policy implementation. The statistical analysis developed in the following sections explicitly tests these predictions.
Contextualizing the Statistical Analysis
I focus on the spread of 287(g) agreements among counties between 2002 and 2011.3 The study of 287(g) agreements, compared to other policies, offers advantages for empirical and theoretical reasons. To enter a 287(g) agreement, all law enforcement agencies must first express interest and apply to participate in an agreement with ICE. This leaves a paper trail subject to public records requests, while other local policies regarding noncitizens tend to be informal and left to law enforcement discretion (Decker et al. 2009). Localities’ stated rationales for entering such agreements are part of the paper trail too, something that existing research has yet to examine. The relative uniformity of 287(g) agreements in terms of aim and scope also makes them more comparable across different locations and therefore easier to identify as imitable. Moreover, 287(g) agreements enter localities into direct enforcement of immigration laws, whereas other policies put localities in more intermediary or ad hoc positions (Armenta and Alvarez 2017). Lastly, a major advantage of studying 287(g) agreements is that two quasi-natural experiments and never-before-analyzed data involving the agreements, as later discussed, allow for clearer pictures of the causes and effects of local participation in immigration enforcement.
For studying 287(g) agreements, a focus on counties also offers clear advantages. Counties comprised the vast majority (82 percent) of the law enforcement agencies participating in 287(g) agreements during this study’s time period. Sheriffs’ offices generally operate larger and longer-term jails than city police, and ICE-locality cooperation is more likely to occur in jails than on patrol (Armenta and Alvarez 2017). And, compared to police departments, sheriffs’ offices have a higher likelihood of participating in immigration enforcement and reporting that federal officials have shaped their decisions to participate in immigration enforcement (Provine et al. 2016).
To understand the spread of 287(g) agreements, I focus on two sheriff-related POAs.4 The first is the NSA. The major advantage of studying the NSA is that the association acted as an arena within which sheriffs and other actors explicitly discussed 287(g) agreements on many occasions through various mediums, as discussed earlier. Indeed, accounts from sheriffs themselves include reports of directly copying 287(g) application materials from other sheriffs following discussion of 287(g) agreements at the NSA conference (see, for example, Pendergraph 2007). And while the NSA—as an organization in an official capacity—did eventually lobby federal authorities for continuance of the 287(g) program and promote participation in the program, such actions took place after the time period of interest to this study (see, for example, NSA 2011, Vaughn 2012b). This timing is important because it provides conceptual clarity regarding the process by which the NSA might have influenced interest in immigration enforcement before 2012, insofar as the timing rules out potential top-down or coercive influences that the NSA’s official backing of a policy could have on policy adoption during this period.5
The second POA on which I focus is the MCSA. The MCSA deserves attention for two main reasons. First, like the NSA, the MCSA served as an important venue where information about 287(g) agreements spread. During the time of interest to my study, MCSA annual conferences took place during and as part of NSA conferences. MCSA attendees, thus, were likely exposed to many of the same sources of influence. Moreover, between 2005 and 2008, Jim Pendergraph was an MCSA executive officer (MCSA 2005; “Sheriff Carona” 2007). Pendergraph’s status as an MCSA officer is important because he was a notable proponent of 287(g) agreements. Beyond encouraging 287(g) agreements at conferences informally (McKain and McKain 2007a) and formally (Pendergraph 2007), Pendergraph frequently contacted other MCSA members and provided delegation tours—so many tours, in fact, that deputies complained of not having adequate time for regular work—to demonstrate the implementation of the 287(g) program (Loller 2006). In 2008, Pendergraph became the head of ICE’s Office of State and Local Coordination, in which capacity he continued meeting with NSA leadership and MCSA members and promoted participation in 287(g) agreements (Capps et al. 2011; Peters 2008).
Another reason the MCSA deserves this study’s attention is its membership criterion. MCSA membership has been historically determined by county population, with membership consisting of all sheriffs who represent counties or parishes with a population of 500,000 or more (MCSA 2022). This cutoff allows for me to employ a regression discontinuity design (RDD) to understand if participation in the MCSA influenced the timing and manner in which counties began to express interest in 287(g) agreements.
Data Used for Statistical Analysis
Online appendix table 1 details the variables used for analysis.6 This section provides an overview of the main outcome and explanatory variables.
Outcome Variables
Expressed Interest in a 287(g) Agreement indicates whether a county submitted a formal letter to ICE applying for participation in the 287(g) program in a given year. I construct this measure using public records and Freedom of Information Act (FOIA) responses from ICE and county sheriff departments. Within the framework of policy diffusion, this action marks a county’s transition from observer to adopter and signals a willingness to participate in federal-local immigration enforcement.
Interest Letter Imitation captures the extent to which a county’s first letter to ICE replicates language from earlier letters submitted by other counties. For each county, I compare its initial letter to all previously submitted letters and assign the highest observed similarity score using the Jaccard coefficient. This score measures social imitation insofar as it reflects the degree to which sheriffs borrow rhetorical frames or justifications from peer jurisdictions. Field reports confirm that sheriffs often shared or copied application letters following professional association meetings (for example, Pendergraph 2007). Prior studies use similar methods to trace diffusion and peer influence through text reuse (for example, Bail 2012; Gilardi et al. 2021). Because the distribution of maximum similarity scores is highly bimodal, I also create a binary indicator for High Imitation, which equals one if the similarity score exceeds the trough between modes.
Transitioning to outcomes of 287(g) agreements that I consider, Percent of County Jail Population Held for ICE measures the share of individuals in a county’s jail population held on behalf of ICE in a given year. This outcome reflects the degree of operational alignment between county jails and federal immigration enforcement. Compliance with Detainer Request records whether a county honored ICE’s request to hold a noncitizen beyond their scheduled release date to facilitate a potential transfer into ICE custody. The unit of analysis is the individual detainer request. This measure captures local cooperation with federal enforcement efforts at the case level. Incoming Transfers of ICE Detainees counts the number of noncitizens ICE reported transferring into a county jail on a given day. This measure captures the flow of detainees into local facilities and reflects the county’s integration into federal enforcement logistics. It is worth reemphasizing that 287(g) agreement with ICE in no way obligates counties to engage in these other activities. Thus, positive changes in these outcomes all reflect an escalation of commitment to immigration enforcement beyond 287(g) agreements.
Key Explanatory Variables
NSA Leadership equals one if a county’s sheriff served in a leadership position within the NSA during a given year. Because most sheriffs hold general NSA membership, I focus on leadership roles, which involve regular attendance at association meetings and exposure to organizational messaging.
MCSA Membership identifies whether a county belonged to the MCSA in a given year. The MCSA restricts membership to sheriffs from counties with populations above 500,000. 287(g) Active indicates whether a county held an active 287(g) agreement during a given year. This binary variable distinguishes formal adopters from applicants and provides the primary treatment variable in analyses of downstream effects.
Empirical Strategy for Statistical Analysis
This study analyzes both the spread and consequences of 287(g) agreements. The first part of the analysis investigates when counties expressed interest in joining the 287(g) program. I code a county as expressing interest in a given year if it officially submitted a letter of application to ICE. I estimate a discrete-time event history model using county-year as the unit of analysis. This model includes county fixed effects, year fixed effects, and a set of time-varying controls. The key explanatory variables indicate whether a county’s sheriff held a leadership role in the NSA or was a member of the MCSA. This approach assumes that, conditional on the included covariates and fixed effects, unobserved shocks do not systematically affect the hazard of submitting an application.
To more credibly estimate the effect of MCSA membership, I use an RDD that exploits the association’s membership rule, which restricted eligibility to counties with populations of 500,000 or more. I center the running variable (county population) at the threshold and estimate local linear regressions using triangular kernel weights. To ensure that estimates are not sensitive to bandwidth choice, I fit models using multiple symmetric bandwidths: 100,000, 125,000, 150,000, and 175,000. These bandwidths reflect a balance between theoretical relevance and empirical precision. This design assumes that counties near the threshold are similar in both observed and unobserved characteristics and that no other policy or institutional change occurs at that same population cutoff. I assess these assumptions by testing for discontinuities in the density of the running variable (using a McCrary test), checking covariate balance across the threshold, and conducting placebo tests at alternative cutoffs. The diagnostics provide no evidence of manipulation or imbalance near the threshold.
I then analyze whether counties that applied to the program imitated the language of prior applicants. To do this, I construct a dataset that pairs each county’s first letter of interest with every letter submitted by other counties prior to its own. Each observation in this dataset corresponds to a county-letter pair. I use Jaccard similarity scores for each pair and assign to each county the highest similarity score across all comparisons. This maximum score captures the degree to which a county’s application reused language from earlier submissions. Because the distribution of maximum similarity scores is highly bimodal, I define a binary indicator for High Imitation, which equals one if the county’s score falls above the valley between the two modes. This classification reflects a clear split between counties that engaged in substantial language reuse and those that did not. I use both pooled regressions and RDDs (centered at the 500,000 population threshold) to estimate whether MCSA-eligible counties were more likely to submit highly imitative letters. I also test results’ robustness to alternative similarity measures.
The second part of the analysis estimates the consequences of entering into a formal 287(g) agreement. Between 2005 and 2012, 168 counties submitted applications to ICE, but fewer than half ultimately entered the program. Some counties withdrew their applications, others were denied, and some remained in pending status. This application process allows me to compare treated counties—those that signed agreements—to counties that applied but never participated. I use a stacked difference-in-differences design that aligns counties by the year of adoption and compares outcomes in the four years before and five years after entry. Counties that never entered but applied serve as the control group. This approach assumes that, absent treatment, treated and untreated counties would have followed similar trends in outcomes. I examine pretreatment trends to evaluate the plausibility of this assumption.
To test whether counties’ motivations shaped the effects of participation, I interact the High Imitation indicator with event-time indicators in the stacked design. This interaction estimates whether counties that closely mimicked earlier letters experienced different enforcement outcomes. I estimate these models for three outcomes: Percent of County Jail Population Held for ICE, Compliance with Detainer Requests, and Incoming Transfers of ICE Detainees. For the jail population outcome, I use county-year panel data and include time-varying covariates and fixed effects. For the detainer and transfer outcomes, I use datasets with individual detainer requests and county-day observations, respectively. I estimate saturated fixed effects models that include county-by-year interactions to absorb all time-varying county-level factors. For the detainer compliance outcome, I also include person-level fixed effects to control for time-invariant characteristics of each noncitizen.
I address the assumptions specific to each analysis in greater detail in online appendix table 1, where I also report robustness checks and sensitivity analyses.
Results of Statistical Analysis
Figure 1 plots the hazard of counties expressing interest in 287(g) agreements by NSA leadership and MCSA membership status over time. It shows that counties with sheriffs in NSA leadership expressed interest in 287(g) agreements at much higher rates between the years 2002 and 2011. Similarly, it documents a higher rate of interest among MCSA members, with a notable jump in interest in 287(g) agreements among MCSA members starting in 2005.
Hazard of Expressed Interest in 287(g) Agreement by NSA Leadership and MCSA Membership, 2002–2011
Source: Author’s calculations.
Note: Kaplan-Meier survival curves show the proportion of counties that had not yet expressed interest in a 287(g) agreement between 2002 and 2011, stratified by sheriff association affiliation. For example, by 2007, approximately 40 percent of counties whose sheriffs were members of both NSA and MCSA leadership had submitted a letter of interest.
Rows (A) and (B) of table 1 display estimates from various discrete-time models predicting time to expressed interest in 287(g) agreements. For ease of interpretation, I present coefficients as changes in predicted probabilities. Thus, during the period of interest, the first model shows that NSA leadership was associated with a 2.5-percentage-point increase in the probability of expressing interest in a 287(g) agreement while holding constant control variables. MCSA membership was associated with a 7-percentage-point increase in the same probability. The next models, using an RDD framework, show that MCSA membership was associated with between a 4- and 12-percentage-point increase in the probability of expressing in 287(g) agreements, depending on the size of the bandwidths around the 500,000-inhabitant cutoff for MCSA membership.
Models Estimating Effect of NSA and MCSA Membership on Time to Expressing Interest in 287(g) Agreements and Interest Letter Imitation, by Sample and Bandwidth
Figure 2 plots measures of Interest Letter Imitation by county population. It shows that Interest Letter Imitation values tend to be higher on average among those counties with populations greater than 500,000 residents than those below 500,000 residents, which suggests that MCSA membership could have an effect on tendencies to reuse text from previously submitted letters. Rows (C) and (D) of table 1 present evidence consistent with figure 2, namely that when comparing letters of interest submitted to ICE with all those letters of interest previously submitted, those letters submitted by members of NSA leadership and the MCSA have significantly higher levels of text reuse compared to nonmembers. The RDD models in table 2 also provide similar evidence across various bandwidth ranges.
287(g) Interest Letter Imitation by County Population
Source: Author’s calculations.
Note: 287(g) Interest Letter Imitation is computed using Jaccard similarity scores between a county’s first letter of interest and previously submitted letters. The x-axis shows county population, centered at the MCSA eligibility threshold of 500,000; values above zero indicate counties above the threshold, and values below zero indicate counties below it. The diagonal lines display local linear regressions fitted on either side of the cutoff using triangular kernel weights.
Models Estimating Effect of 287(g) Agreement on Percent of County Jail Population Held for ICE, Compliance with Detainer Requests, and Incoming Transfers of ICE Detainees, by Level of Interest Letter Imitation
Row (A) of table 2 presents estimates of the effect of 287(g) agreement adoption on the percent of the county jail population held for ICE, disaggregated by counties’ levels of Interest Letter Imitation. Among counties classified as high imitators, 287(g) adoption is associated with a statistically significant increase of 0.088 percentage points in the share of the jail population held for ICE. In contrast, counties with low levels of imitation show a small and statistically insignificant decline (−0.040). The difference in effects across the two groups is 0.129 percentage points (p < 0.05), indicating that the estimated impact of 287(g) participation varies by the level of textual similarity in counties’ initial letters of interest. Figure 3 illustrates this divergence over time, with relatively flat trends among low-imitation counties and post-adoption increases concentrated among high-imitation counties.
Effect of 287(g) Agreement on Percent of County Jail Population Held for ICE (Cube Root) by 287(g) Interest Letter Imitation
Source: Author’s calculations.
Note: Lines show event-time coefficients from stacked difference-in-differences models centered on the year a county adopted a 287(g) agreement. The outcome variable is the cube root of the percent of the county jail population held for ICE, transformed to reduce skewness. Shaded areas indicate 95 percent confidence intervals. The omitted category is the year immediately prior to adoption (t = −1). Panels compare counties classified as high vs. low in Interest Letter Imitation, based on Jaccard similarity scores between each county’s first letter and previously submitted letters.
Row (B) of table 2 reports the estimated effect of 287(g) agreement adoption on the probability that a county complies with an ICE detainer request. Among counties classified as high imitators, adoption is associated with a statistically significant increase in compliance of 0.121 percentage points. In contrast, counties with low levels of imitation show a small and statistically insignificant decrease in compliance (−0.110). The difference between the two groups is 0.231 percentage points (p < 0.01), indicating that the estimated effect of 287(g) participation on detainer compliance differs by counties’ levels of Interest Letter Imitation. Figure 4 provides event-study estimates that parallel this finding, with compliance rates remaining flat or declining slightly in low-imitation counties but trending upward after adoption in high-imitation counties.
Effect of 287(g) Agreement on P(Comply with Detainer Request) by 287(g) Interest Letter Imitation
Source: Author’s calculations.
Note: Lines show event-time coefficients from stacked difference-in-differences models centered on the year a county adopted a 287(g) agreement. The outcome variable is the probability that a county complies with an ICE detainer request. Shaded areas represent 95 percent confidence intervals. The omitted category is the year immediately prior to adoption (t = −1). Panels compare counties classified as high vs. low in Interest Letter Imitation, based on Jaccard similarity scores between each county’s first letter and previously submitted letters.
Row (C) of table 2 presents estimates of the effect of 287(g) agreement adoption on the number of ICE detainees transferred into county jails. Among high-imitation counties, adoption is associated with a statistically significant increase of 0.353 detainees per county-day. Low-imitation counties show a smaller and not statistically significant increase of 0.289. The difference between the two groups is 0.064 (p < 0.05), which suggests a modest but statistically significant divergence in transfer activity following adoption. Figure 5 shows a similar pattern over time, with transfers remaining more stable in low-imitation counties and rising steadily after adoption in high-imitation counties.
Effect of 287(g) Agreement on Incoming Transfers of ICE Detainees
Source: Author’s calculations.
Note: Lines show event-time coefficients from stacked difference-in-differences models centered on the year a county adopted a 287(g) agreement. The outcome variable is the number of ICE detainees transferred into a county jail on a given day. Shaded areas represent 95 percent confidence intervals. The omitted category is the year immediately prior to adoption (t = −1). Panels compare counties classified as high vs. low in Interest Letter Imitation, based on Jaccard similarity scores between each county’s first letter and previously submitted letters.
CONCLUSION
Using a range of data sources and methodological approaches, the analysis points to two general findings. Concerning the spread of 287(g) agreements, county participation in POAs—specifically the NSA and the MCSA—predicted (1) significantly higher rates of expressing interest in 287(g) agreements and (2) significantly higher levels of copying language used by other counties in application letters for 287(g) agreements. This suggests that many counties’ decisions to participate in immigration enforcement were subject to social influence experienced because of their affiliation with a POA.
On 287(g) agreements’ consequences, I find their outcomes interact with the aforementioned social influence. Following the enactment of 287(g) agreements, counties that had imitated the application materials from other 287(g) applicants at higher levels significantly diverged from counties with lower levels of imitation in application materials in terms of their tendency to escalate the commitment of existing law-enforcement resources to immigration enforcement. Indeed, the scale of immigration enforcement significantly expanded in the former counties insofar as they dedicated a larger share of jail space for confining noncitizens for ICE, complied with ICE detainer requests at higher rates, and became more central in the network of correctional facilities through which ICE shuttles noncitizens.
Here, it may be worth noting this study’s limitations. For one, this study only considers one type of social influence—peer imitation among public officials—as measured through text reuse in application letters for 287(g) agreements. This approach captures a narrow but observable channel of influence: rhetorical mimicry following participation in professional associations. But it is likely that other forms of social influence, like shared consultants or legal counsel, also play a role in shaping both the decision to participate and the intensity of policy implementation. Moreover, it is possible that something other than social influence led to higher levels of Interest Letter Imitation and higher levels of the aforementioned outcomes independently. But it seems unlikely that anything other than social influence—in some form—led to pairs of counties submitting letters that contain large sections of identical text—including typographical errors. Moreover, while it is possible that unobserved variables could explain selection into social influence that would affect Interest Letter Imitation and the aforementioned outcomes, my attempts to detect indicia of selection have revealed no clear trends. That said, while I have taken efforts to reduce potential sources of bias through careful design and robustness checks, the estimates should not be interpreted as identifying causal effects with certainty.
Notwithstanding these limits, these findings offer insights for several bodies of scholarship. First, while research on the spread of subfederal immigration enforcement has pointed to supralocal influences contributing to this spread (see, for example, Steil and Vasi 2014), such research has largely highlighted influences that are explicitly partisan or political in nature. My archival analysis of POA materials—including NSA and MCSA conference proceedings, publications, and correspondence—and my statistical analysis illustrate the importance of previously overlooked, nonpartisan professional associations as nonpartisan sources of such influence. Given the prevalence of nonpartisan organizations and the fact that organizational research has argued that “the character of American federalism encourage[s] … [an] association-driven diffusion” of policy often by associations working outside of existing partisan allegiances and power (Clemens 1997, 98), this is an important area for research on subfederal immigration enforcement to develop.
Second, this study builds on existing research looking at the consequences of immigration enforcement. Prior studies have shown that new federal immigration policies lead to increased commitment of resources by federal authorities (Andreas 2000), subfederal immigration policies can fuel public hostility toward immigrants (Flores 2017), and subfederal law enforcement officers adjust their behavior when involved in immigration enforcement (Armenta 2017; Donato and Rodriguez 2014). My research expands on these by showing how motivations for entering into immigration enforcement help explain variation in the subsequent escalation of commitment to immigration enforcement.
Third, this study contributes to a growing body of scholarship examining the political and institutional power of sheriffs in contemporary law enforcement. Recent work highlights the unique autonomy sheriffs possess as elected officials and their ability to shape local criminal justice and immigration policy with minimal oversight (Farris and Holman 2024; Pishko 2024). My findings show that sheriffs do not make enforcement decisions in isolation but often in response to signals and incentives generated by supralocal professional networks. In particular, I argue that POAs serve as influential sites of policy diffusion, where sheriffs encounter federal officials, model policies, and peer endorsements that encourage participation in immigration enforcement. These organizational settings help explain not only which sheriffs pursue programs like 287(g), but also may have further downstream effects on how much institutional capacity sheriffs ultimately dedicate to federal enforcement efforts. By linking the social and institutional motivations for entering immigration enforcement to concrete implementation outcomes, this study shows how sheriffs’ professional affiliations can shape the scope and intensity of their cooperation with federal immigration authorities.
These findings also hold implications for ongoing discussions on 287(g) agreements and immigration federalism. Now that President Trump has returned to office and begun revitalizing the 287(g) program, the findings of this study carry renewed policy relevance. Dozens of agreements remain active, and recent signals from the Trump administration suggest expanded efforts to enlist local law enforcement in federal immigration efforts. To the extent that jurisdictions adopt these agreements in response to social influence—rather than through independent policy evaluation—they may commit resources to enforcement in ways that persist beyond the agreements themselves. This has implications not only for enforcement levels but also for broader immigration policymaking. As Pratheepan Gulasekaram and S. Karthick Ramakrishnan (2013) argue, when subfederal governments take on a larger role in immigration enforcement, they may generate political feedback loops that constrain future federal legislative action by normalizing local control. The dynamics identified in this study may also extend beyond 287(g) agreements. States like Texas have recently initiated unilateral enforcement efforts (Barragán and Svitek 2022), and model legislation promoted by interest groups may further diffuse similar policies across jurisdictions (Hertel-Fernandez 2019). Future research can examine whether these newer enforcement regimes—often rooted in partisan dynamics—generate similar institutional commitments and policy effects or whether their consequences differ depending on the motivations and contexts of adoption.
Finally, this study also holds implications for understanding an increasingly important but understudied legal device: the intergovernmental agreement. While only a minority of all counties at any given time participate in 287(g) agreements, it has become far more common for state and local governments to enter into agreements with federal immigration authorities to participate in immigration enforcement in other ways—such as using their correctional facilities to confine noncitizens in ICE custody (Ryo and Peacock 2020). Bridget Fahey (2020) identifies the type of arrangement that exists in a 287(g) agreement as an example of federalism by contract, wherein multiple governments coauthor contracts that function like regulations or statutes. Federal authorities’ access to state driver’s license databases to conduct facial recognition searches, military transfers of equipment to local police, or Medicaid’s operationalization are all achieved through similar agreements. Many such agreements remain secret (Fahey 2020). While much of the scholarly work that can improve government policy relating to intergovernmental agreements centers around doctrinal questions—of which rules apply or which courts govern, for example—my study contributes to an empirical understanding of such agreements. I show that sociological and organizational processes shape the performance of such agreements in ways that may fundamentally alter participants’ priorities and resource allocation. Thus, beyond 287(g) agreements in particular, this study offers a glimpse into the power of intergovernmental contracts to establish new stakes and interests in local governments. These empirical considerations should inform discussions of the costs and benefits of intergovernmental agreements in the context of immigration enforcement and other areas.
FOOTNOTES
↵1. This section draws on an original review of primary sources, including back issues of Sheriff magazine, NSA and MCSA conference programs, press releases, IRS Form 990 filings, and other organizational documents from 2000 to 2012. I also examined contemporaneous news coverage and ICE communications to reconstruct how 287(g) agreements were presented, discussed, and disseminated within POA networks. These sources provide insight into the informational environment that shaped sheriffs’ perceptions of immigration enforcement and the role of professional peer networks in policy diffusion.
↵2. For example, subsequent speakers included: former US Attorney General Alberto Gonzalez (“Attorney General Gonzales” 2005); US Senator Joseph Biden (“Biden” 2005); former Department of Homeland Security Director Janet Napolitano (NSA 2009); former ICE Director John Morton (“National Sheriffs’ Association 70th” 2010); and criminology professor Simon Perry, who specializes in homeland security and policing (Peluso 2013).
↵3. I limit years studied to 2002–2011 for theoretical and empirical reasons. First, this range of years covers the early institutionalization period of 287(g) agreements, before the Obama administration began a phaseout of many agreements in 2012 (Kolker 2021). Federal funding for the 287(g) program also saw a notable decline shortly after this period (Kolker 2021). Moreover, in 2011, the POAs that I study eventually endorsed 287(g) agreements, which makes untangling the processes by which participation in POAs influenced county decision-making more difficult.
↵4. These POAs are in no way representative of all or most POAs. This article’s aim is not to generalize to all POAs.
↵5. During other periods, this may be a well-founded concern because the NSA is one of the four parent associations of the Commission on Accreditation for Law Enforcement Agencies, which accredits law enforcement agencies. The NSA often coordinates with federal authorities in deciding on the allocation of grants and the provision of training, technical assistance, and other services (Bowman 2001). These aspects of the NSA may make certain sheriffs more likely to implement NSA-backed policies.
↵6. The online appendix can be found at https://www.rsfjournal.org/content/11/4/26/tab-supplemental.
- © 2025 Russell Sage Foundation. Peacock, Ian G. 2025. “Imitate, Then Escalate: Social Influence and Its Consequences for the Subfederal Deportation System in the United States.” RSF: The Russell Sage Foundation Journal of the Social Sciences 11(4): 26–48. https://doi.org/10.7758/RSF.2025.11.4.02. Thanks to Asad Asad, Ingrid Eagly, Jacob Foster, Brad Jones, Cree Jones, Caitlin Patler, Huyen Pham, Emily Ryo, Roger Waldinger, Edward Walker, and three anonymous reviewers for their feedback. Funding from the following sources supported this research: American Sociological Association; John Randolph Haynes and Dora Haynes Foundation; and University of California, Los Angeles. Direct correspondence to: Ian Peacock, at ipeacock{at}uchicago.edu, 1111 E. 60th Street, Chicago, IL 60637, United States.
Open Access Policy: RSF: The Russell Sage Foundation Journal of the Social Sciences is an open access journal. This article is published under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.
REFERENCES
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.