How universities facilitate city network socialization through knowledge exchange on immigrant integration

Funding information Open Society Foundations, Grant/Award Number: OSIFE-grid.453407; Paul Hamlyn Foundation, Grant/Award Number: PHF-grid.481059.1; British Academy, Grant/AwardNumber: PF21∖210066 Abstract Universities increasingly facilitate aswell as study social change alongside decision-makers and users. This is partly in response to demands for civic engagement and demonstrable public impact. While existing scholarship has critically examined these collaborations’ impacts, less work addresses either how they develop or what benefits they confer to participants— especially those operating at local scales. By examining two university-initiated networks comprising 28 cities in 12 European countries working on immigrant integration issues, we show how two-way knowledge exchange among researchers and municipal policymakers can foster peer learning and coproductive dynamics. We argue that these exchanges socialize cities intomore cohesive groupswith shared goals and agendas, particularly in low-salience policy areas, as integration can be in some national contexts. Moreover, universities and universityaffiliated researchers play unique roles in facilitating this process. Our results have theoretical implications for multi-level and networked governance and offer practical guidance for designing knowledge exchange initiatives.


INTRODUCTION
Universities are potentially well-placed to facilitate peer learning and knowledge exchange with groups beyond academia, thanks to their abilities to produce knowledge and convene a range of expertise. These exchanges comprise a core function of city to city networking on socially and politically important issues, including immigrant and refugee integration. 1 Yet despite increasing concerns about the political drivers and consequences of seeking research impact (e.g., Smith et al., 2020), as well as renewed interest in higher education as a means of promoting civic engagement and democratic values (e.g., Hartley, 2009;Hazelkorn, 2016), the roles of universities have been under-examined and under-theorized in the domains of migration at the city level. Instead, researchers in city networks are usually conceived as filling narrowly defined roles of external experts who, at most, actively cultivate statuses as 'researcherentrepreneurs' (Russeil & Healy, 2015).
By contrast, as work in adjacent scholarly and policy areas shows, universities and researchers associated with them potentially play more active roles within networks by anchoring, brokering, and fostering knowledge exchange through structured activities and programming (Lepore et al., 2021;Meyer, 2010). Yet there is a striking lack of evidence showing how and in what circumstances this happens, which presents problems for better theorizing about how successful knowledge exchange might arise as well as generating guidance for practice and policy-making.
In response, we present a novel and empirically grounded example of how academic researchers and city-level practitioners relate to each other (the 'City Working Group Model'). This involved a sustained programme of knowledge exchange which itself emerged from experimentation and revision through prior rounds of knowledge exchange inspired by some aspects of co-production and participatory action research methods. These two programmes, 'Action towards Inclusion in Europe' (2014)(2015)(2016) which preceded the 'Inclusive Cities' programme (ongoing since 2017) form the basis for this analysis. Inclusive Cities is a knowledge exchange initiative working with 12 UK cities aiming to support them in developing and implementing changes in their approach to migrant integration and inclusion at the local level. We particularly focus on the distinct roles played by universities and academics in these city networks, which speaks to growing interest in the 'local turn' (Caponio & Borkert, 2010) as a way of understanding policy-making and public debates surrounding politically consequential issues.
By tracing the development and execution of these two programmes as exemplars of a 'City Working Group model' , we make three contributions. First, we extend existing theories of city networks to include consideration of universities as agents of socialization in their own rights, but ones which are also motivated by civic demands at local and global levels. Second, we conceptually and empirically establish what a new model of university-facilitated knowledge exchange among cities looks like. Specifically, we identify the processes involved in such exchanges, as well as the outputs and outcomes arising from the model. Relatedly, we also examine the extent to which these outcomes can be measured. Third, we advance methodological understanding about how researchers can be productively embedded within these programmes of knowledge exchange.

CITY NETWORKS, SOCIALIZATION, AND THE CIVIC ROLES OF UNIVERSITIES
In many respects, universities have always had varying levels and types of connections to the places in which they are located. Such civic engagement has evolved from detached and elite-led modes of knowledge production centred around scientific discovery for its own sake to solving societal problems with the consent and involvement of groups beyond academia (Hazelkorn, 2016;Hughes & Kitson, 2012). This trend, manifested in different ways around the world (e.g., McIlrath et al., 2012), has supported shifts towards civically minded objectives that prioritize impact and strengthening links with local users-particularly when the research in question has been funded through taxpayers' contributions. Moreover, as universities pivot towards addressing complex global problems that require collaboration across traditional disciplinary knowledge bases, as well as public and private enterprise, they increasingly rely on networks and international consortia to deliver such impact (Goddard, 2018). Consequently, we echo how prior scholarship conceives of universities as institutions that link multiple scales and places.
While recent work has productively identified practical and theoretical challenges associated with communitybased research on policy-relevant topics such as international development (e.g., Lepore et al., 2021), it also highlights a relative lack of understanding of how such networks operate. All the more reason, then, for detailed and contextualized analyses of specific examples which can provide roadmaps for future interventions. As we will explain in our empirical material, one of the key dimensions of the City Working Group model that we present involves learning from and exchanging ideas with peers to develop appropriate responses to immigrant integration and inclusion issues. We argue that, in line with their broader civic purposes outlined above, universities play unique and important roles in facilitating these multidirectional processes. City networks themselves seem to recognize this importance: in a recent survey of existing networks, Lacroix (forthcoming) found that 17% (11 out of 64) explicitly mentioned supporting research as a key objective. Specific to migration and immigrant integration, a review of some of the most visible city networks in Europe working in this field observes how universities (and individual high-profile scholars) can seed and sustain networking activities (Oomen, 2020). Yet these roles have been under-theorized in the context of scholarship on and about city networks. To address this gap, we connect concepts from sociology, evidence and policy-making studies, and urban studies to show how universities can play a vital role in city networks' operations and objectivesparticularly in politicized issue areas like integration.

City networks as agents of socialization
Our first step involves conceiving of city networks as 'agents of socialization' that enable constituent cities to develop their 'sense of groupness and an identity of [being a] national or international player' (Lacroix and Spencer, this issue).
Here, we split this concept into two parts: networks' 'agent' status, and 'socialization' as both process and outcome.
While there are considerable debates about what constitutes agency, particularly in migration contexts (e.g., Bakewell, 2010), we use this term to draw attention towards the ways that collective networks can sometimes exert greater influence than any individual city member. Indeed, for many city networks, this is an explicit goal that motivates municipal interactions in the first place (Flamant, this issue).
But in what contexts, and over which entities, do these networks exert influence? As Lacroix and Spencer (this issue) observe, city networks span national and international levels: they are multi-scalar in nature, crossing geopolitical boundaries to connect diverse actors including local and national governments, civil society organizations, businesses and corporate interests, and citizens and migrants themselves. This is the social and political reality into which city networks fit. As our empirical material shows, constituent cities share experiences and knowledge among themselves as they engage in these connecting activities. This gives rise to processes of socialization that ultimately enhance cities' capacities and collective identity under the banner of a given network.
Prior work has tried to express this mode of socialization through various mechanisms, including policy diffusion (Shipan & Volden, 2008) and 'city to city learning (C2CL)' (Haupt et al., 2020). Two features distinguish the latter kind of learning from other forms of urban cooperation (see Shefer, 2019), which make it relevant for our case: (1) its scope is tightly focused on cities, and more precisely the policymakers operating within them; and (2) it prioritizes the forms of knowledge involved, as well as how that knowledge is transmitted. Yet there remain key gaps in how C2CL is currently understood. Notably, universities are missing from the picture-a striking omission given their unique abilities to generate and gather critical masses of knowledge and knowledge producer), as well as convene users around these areas.
To be clear, scholarship in some parts of urban studies and higher education studies has considered universities' wider roles in cities. For example, universities have variously been characterized as 'anchor institutions' (Goddard et al., 2014;Harris & Holley, 2016) that confer varying levels of economic benefits to regions, through attracting and retaining workers in key sectors (see Hawthorne, 2018). In this sense, universities exist within wider migration patterns that involve urban areas (Walton-Roberts, 2021). Beyond their economic impacts, universities also gather knowledge and resources around locally relevant issues, and link key players together to address shared goals (Goddard, 2018). These activities may have overtly civic objectives, such as addressing social issues or raising staff members' and students' awareness of community-based organizations (Hartley, 2009).
Meanwhile, broader shifts in universities' governance, particularly those receiving public funding, increasingly motivate or even require academic researchers to engage with users in other sectors (Smith et al., 2020). This provides incentives for some universities to become more entrepreneurial and proactive in seeking out collaborations, particularly if their geographic location allows for easier connections with key industries (e.g., Mok & Jiang, 2020) or if they already possess a degree of international visibility (e.g., Caponio, 2018). As a result, prior work in this area tends to evaluate the outcomes of networking activities involving universities, such as by measuring the extent to which research outputs are successfully transformed into commercial products (Huggins et al., 2008).

Universities within city networks: Knowledge exchange as a vector for socialization
By contrast, we argue that greater theoretical and empirical attention to how universities engage with city-level actors (rather than documenting whether they do so) on issues that stretch beyond the immediate locality provides a mechanism by which city network socialization can happen. Here, we can learn from the field of evidence and policy-making studies, which has devoted much attention to understanding how knowledge producers (e.g., universities) relate to decision-makers in policy-making. 2 Theoretically, these groups tend to exist in one of four relationships: (1) knowledge production impacting policy-making in an instrumental manner; (2) politics and policy-making leading knowledge production; (3) the two areas mutually constituting and impacting each another in a co-productive manner; or (4) occupying separate, distinct spheres of activity that have-and operate by-different sets of logic, meanings, and values (Boswell & Smith, 2017; see also Weiss, 1979).
The first arrangement is an idealized version of public impact, deriving from the tradition of science communication in which academic expertise is clearly preferable on certain topics (Allen et al., 2019). The second relationship, by contrast, is probably more reflective of settings that lack clear and objectively 'true' outcomes. The risk of this approach is that 'research is constantly being influenced by policy and politics and that efforts to bring researchers and policymakers closer together are likely to exacerbate this in ways that may not be desirable . . . At worst, the impact agenda will lead to the increasing politicization of research (and an associated reduction in academic freedom)' (Boswell & Smith, 2017). There is the risk that an ever-closer relationship with policymakers will drive research agendas to the exclusion of research questions that are not either politically palatable, or technically 'useful' to the policymaker audience.
While all four modes can be present in city-level engagement, we particularly want to focus on the co-productive mode as one which is most likely to contribute towards cities' (or more precisely, city-level decision-makers') socialization as part of transnational networks. This is for two reasons. First, the presence of two-way relationships among universities and city partners means that knowledge, experiences, and norms are more likely to flow among all members of the network, rather than simply being stockpiled by one or a few partners. Second, learning from other citiesinstead of from university partners that merely dispense academic knowledge-potentially fosters intangible yet vital qualities of group identification, self-awareness, confidence, institutional buy-in, and a shared sense of self-importance that contribute to socialization.
But how does this happen? The concept of 'knowledge exchange' as it has developed in the evidence and policymaking literature provides some clues. Successful exchanges tend to involve a degree of fluidity and iterative revision among partners: while steps such as defining key problems or identifying actors that could help with a specific task tend to appear somewhere in the exchanges, they do not necessarily have to happen in a set order. Indeed, the outcomes from these steps may be revisited and modified as part of the mutual learning process (Ward et al., 2012). They also can involve a specific individual or organization serving as a 'broker' as they move among the participants and facilitate exchanges (Knight & Lyall, 2013). Therefore, our main contribution is to empirically demonstrate how knowledge exchange, brokered by a university partner, can become a vector for city networks' socialization. While prior work has argued that new knowledge can arise from exchange and brokerage activities involving many kinds of intermediaries (Meyer, 2010), we have specifically argued that universities are uniquely positioned to fulfil a role of the 'honest broker' in issue areas characterized by polarization and politicization. Then, by using a case of knowledge exchange occurring within the domain of migrant integration and inclusion, we extend this argument to focus attention on how specific processes and practices generate socialization dynamics among cities. Not only does this develop theorization about the role of universities within city networks, it also presents insights for policymakers and practitioners who may either turn to or be approached by universities as project partners.

THE ROLE OF UK LOCAL AUTHORITIES IN INTEGRATION: GAPS AND OPPORTUNITIES
The 'local turn' in migration studies (Caponio & Borkert, 2010) highlights the need to document and understand the dynamics surrounding integration efforts and policy-making as they occur at subnational levels. For example, Spencer and Charsley (2016) note that most integration processes and their effects are felt at the local level. However, gaps in local governments' capacity to act can exist at this level. This is especially the case in the United Kingdom, which lacks a national integration strategy and does not place requirements on local governments to develop their own strategies. 3 Instead, 'the role of UK local government in the reception and inclusion of newcomers has often been one of low salience and limited capacity, whereby a highly centralized system combines with a lack of clarity as to the role of local government in the inclusion of newcomers, to produce relative inertia' (Broadhead, 2020). Similarly, although several city networks address this topic, especially across Europe, this limited capacity has consequences for the characteristics of networking activities in the United Kingdom. Compared to their European counterparts, they tend to be more limited in scope, focused on more targeted policy engagement, led by civil society rather than local government, and prioritize advocacy goals rather than strategic or planning objectives such as improving service delivery. 4 What is more, the United Kingdom's devolved system that incorporates four constituent nations (England, Wales, Scotland, and Northern Ireland) adds further complexity: migration governance remains a reserved matter at the UK government level, but integration is devolved. As such, the devolved administrations are either developing or have already implemented their own approaches. For example, the Scottish government has outlined its own 'New Scots' strategy (Scottish Government, 2018).
Notwithstanding this, studies investigating the multi-level governance of integration have 'revealed much more "entrepreneurship" by local governments not only in developing their own integration policies but also in setting their ideas on the agenda of national governments' (Scholten, 2016). Although some local authorities do not want to deviate from existing national policy or to lead developments within policy areas such as integration, others who feel the effects of a lack of (national) policy guidance and action have chosen to take stronger leadership roles.
This reality shaped the development of the Inclusive Cities programme in two ways: first, by valuing academic leadership and facilitation as a means of initiating such a network, and second, by responding to the lack of capacity (perceived or real) within local government. 5 With respect to capacity, local government does not possess the material resources or organizational ability to set and maintain an agenda on the issue: after all, these policy questions are neither statutory responsibilities nor priorities for extra funding from central government. Yet local government has also been generally ignored as a domain for knowledge exchange-especially on migration and immigrant integration.
Therefore, these issues have not attracted additional resources in the UK context.
As a result, the Action for Inclusion project was motivated by the observation that 'exclusion is perpetuated in part by the lack of opportunity for policymakers, service providers and civil society actors at the local level to access, discuss and act upon evidence from research and from other cities, on practices and initiatives that have helped to foster inclusion' (Spencer, 2016). This drove the development of the model and its iterative character that involved peer learning TA B L E 1 How the Inclusive Cities Programme incorporates academic research into its design

Features of immigrant integration and inclusion Adaptation into the Inclusive Cities Model
Involves both newcomers and longer-term residents

Conceives integration as a two-way relationship between newcomers and residents
Requires a range of actors to take shared responsibility Requires each city to establish a taskforce comprising local stakeholders, which provides feedback on the planning process and takes forward initiatives (mirroring horizontal governance: see Scholten, 2016) Multidimensional in character, to include structures (e.g., labour market participation); participation in society, culture, and civic life; and identity Uses an action planning model comprising specific actions and timescales across thematic areas, including (1) development of an inclusive narrative to motivate the actions, (2) policies promoting inclusive economic growth, (3) connections across communities, (4) the mainstreaming of inclusion within existing public services such as health, housing, and social services, and (5) civic representation and participation (collectively known as the 'Inclusive Cities Framework': see Broadhead & Kierans, 2019) Implemented mostly at the local level Orients policy engagement exclusively at local levels as well as academic research-specifically, as shown in Table 1, a model of integration (Spencer & Charsley, 2016) which informs its core principles and design.

THE CITY WORKING GROUP MODEL: UNIVERSITY-FACILITATED PEER LEARNING ON IMMIGRANT INTEGRATION
The city working group model comprises five core features which we argue lead to more meaningful knowledge exchange: (1) intentional recruitment and selection of complementary and highly-engaged cities, (2) links between research evidence and practice, (3) a researcher-facilitated discussion space, (4) a structured programme to enable and recognize progress, and (5) peer learning among cities, as well as between cities and university researchers.
These features arose out of a preceding process of knowledge exchange called 'Action towards Inclusion in Europe' that ran between 2014 and 2016 and involved 16 European cities working to secure reforms in city practices that addressed the exclusion of marginal communities from services and civic life. 6 It prototyped the City Working Group model in its organization, which comprised one representative from each city who took responsibility for helping their respective city team foster a positive sense of local belonging in a key area of policy or practice using actionoriented learning exchange. Collectively, these separate working groups focused on three thematic areas: cohesion and belonging, parental engagement in schools, and homelessness and destitution among excluded migrants. Every participating city produced their own City Action Plan, authored by local policymakers who worked closely with a dedicated researcher on each theme. The 'Action for Inclusion' activities revealed three characteristics of the knowledge exchange process that might be more likely to lead to meaningful action by municipal representatives: an appropriately long timescale for implementing the action plans (ideally no less than 12 months), involvement of sufficiently senior participants from a mixture of backgrounds, and the creation of a city taskforce that would provide a city-based reference group to which each representative could provide feedback.
These features informed the design of the Inclusive Cities programme, which generated the City Working Group model we present in this paper. 'Inclusive Cities' is a knowledge exchange initiative aiming to achieve a step change in the approach to the inclusion of newcomers at the local level. Funded by independent philanthropy, it has run in two phases: (1) from 2017 to 2019, involving six UK cities and in partnership with Welcoming America, 7 and (2) from 2017 to 2022, involving the original six UK cities as well as seven new members, 8 and as a founding member of Welcoming International. 9 As a condition of joining the programme, each city committed to developing an action plan that was fundamentally shaped by the underpinning research while also supported by briefings and input from researchers. This plan would also detail their work on inclusion across a wide range of policy areas related to migrant integration and inclusion. 10 Each city also committed to convening a city-wide taskforce of stakeholders to support the project, as well as appointing both a senior sponsor who was either an elected official or an otherwise senior officer and an operational lead.
In total, these aspects and design features demonstrate how the City Working Group model aims to co-produce knowledge with academic, policymakers, and politicians by treating them as core participants who are expected to take shared ownership of the process and its outcomes. This contrasts with prior approaches involving municipalities. Although seeking academic expertise is not new within city network models-the CLIP, Intercultural Cities, and EUROCITIES programmes include expert advisors-these have tended to follow the roles described by Russeil and Healy (2015) as 'researcher-entrepreneurs' and 'consultant-entrepreneurs' . In some cases, cities participate in university-led research projects, while in others, researchers contribute to city-led projects. In either instance, relationships among researchers and network members can be distant and primarily instrumental, such as when university-affiliated academics are perceived to add legitimacy to a project.

WHAT MAKES FOR SUCCESSFUL KNOWLEDGE EXCHANGE? LEARNING FROM THE CITY WORKING GROUP MODEL
In this section, we analyse the process and outcomes of knowledge exchange that derived from the City Working Group model. Our analysis, corresponding with the five core features of the model earlier described, is based on an anonymous survey of city participants comprising civil servants and elected officials undertaken in January 2019 (N = 14), as well as observations from the embedded researcher who facilitated the project. 11 We have supplemented these primary sources with reports from the preceding Action Towards Inclusion project, which drew on a survey of participants and through face-to-face discussions with each of the three working groups.
Although the embedded researcher role resembles the type of 'insider research' outlined by Coghlan and Brannick (2007) as a form of 'action research' , it remains distinctive because the researcher remains within the university context while being centrally involved in the set-up and running of the network. A consequence of this is organizational duality that spans active participation in both the network and research. As Coghlan (2019) highlights, this means that research takes place within the 'present tense' . Therefore, the process relies on a high degree of iterative, reflective practice from the researcher in recognition of the co-productive nature of this role.
We argue this approach presents a distinct perspective, whereby universities and university-affiliated researchers become knowledge brokers who contribute to city network socialization. In this model, academics are not merely bystanders who disseminate relevant findings. Rather, following Boswell and Smith (2017), they are actively involved in the co-production of research findings (such as those contained within this paper) and the development of city networks themselves which-while led by municipal representatives-are facilitated and supported by researchers.
However, acknowledging how researchers fulfil different roles from city officials, politicians, and other municipal actors does not imply their objectivity. By contrast, this form of research involves researchers proactively entering the fray of network building and policy implementation which necessarily has political dimensions. For this reason, the researcher's ability to maintain duality and reflect on their own practice is key, which we recognize can be difficult as institutional incentives for doing this kind of work may differ from those held by individual researchers.

Selecting and recruiting complementary and engaged cities
Selecting and recruiting complementary cities are key steps within the model because they involve developing political support for, and engagement with, the subsequent steps. Indeed, one participant stated that 'ensuring political leadership has given the project a higher status, creating more buy-in from partners both within and out with the local authority. It reinforces the importance of creating a welcoming approach within the city.' Recruitment occurred in two phases. First, the researchers developed a longlist of 15 UK cities that had experienced the highest levels of migration over the previous 10 years, as well as met several other population-based criteria based on existing migration datasets. 12 At this stage, having research-based colleagues to provide the expertise necessary for accessing and analysing existing migration data was crucial. Second, six cities from these 15 UK cities were selected to create a set that varied along demographic, geographic, and political lines. Moreover, the programme prioritized cities that had demonstrated an interest in the policy agenda as well as willingness to commit resources.
Specifically, the programme required member cities to supply two sponsors (corresponding with operations and strategy aspects) as active participants. Moreover, these sponsors had to be relatively senior, operating at the levels of In practice, these requirements presented advantages and disadvantages for participating cities. On the one hand, cities observed how they generated buy-in. On the other hand, achieving more active participation by senior staff during events sometimes remained difficult. One respondent stated 'it would be useful to have greater involvement from elected members . . . it will likely not be possible to get more senior staff to consistently attend meetings around the UK, but if methods of sustaining their engagement throughout the project could be developed, this would be useful' .
Another noted that 'the level of seniority wasn't always the most important thing, as this doesn't necessarily equate to the ability to influence. I found having Member 13 input very useful, and although we weren't able to sustain our Member attendance on all the learning exchanges, exposure at the first meeting in City X has meant that our lead member has continued to be a strong advocate for the programme within the city council' . In response, the programme incorporated shorter events to encourage engagement by elected members while being realistic about the time that senior staff could commit outside their core operational duties.

Providing an interface between research evidence and practice
Another key contribution of the City Working Group model compared to other networks is how it provides a deliberately two-way interface for knowledge exchange among university researchers and municipal practitioners. Specifically, in addition to giving researchers an understanding of practitioners' perspectives, the programme allows policymakers to learn from research outputs in accessible forms (e.g., Spencer, 2016). Research came in various forms: background papers, presentations from academics, briefings, and bespoke research to support local authorities' specific initiatives. However, all of these materials were selected and synthesized for policy audiences, which made them more likely to be used by these groups. For example, one participant observed 'I often use them as quick-reference guides when drafting documents or preparing for local meetings' . Another participant described how 'the academic input has been very important. It has helped shape our critical thinking and provided information on scope and context' .
Beyond making research accessible, the academics' contribution also involved selecting and curating the set of resources for policymakers, which added value to the programme. One participant stating that, 'I think without the academic approach we would be wading through examples of other projects and feeling overwhelmed about where we would get resources to make a change.' This illustrates how universities' brokerage role goes beyond mere facilitation to include taking an active role in identifying high-quality and relevant materials using academic knowledge and skills.

Facilitating and framing discussions
Academic researchers also had roles in facilitating the programme. This took two main forms: creating spaces for discussion (which we will cover later in this section) and modelling new ways of thinking and working within those discussions. As the Action towards Inclusion learning paper identified, participants stressed not only the instrumental value of the evidence presented but that the process itself had led them to think differently about what they were doing: 'we have to make it clear that this kind of process is really useful -to make it a bit more academic than I'm used to, to think a bit more, to ask constantly ''Why?'' Why am I doing this in this way?' (Spencer, 2016, p. 6) Municipal approaches that had seemed normal were now interrogated: 'doing it in this more academic way wouldn't have happened if we'd just done this within our administrations' . One Action toward Inclusion participant noted that 'the fact that the facilitator is an academic is valuable; it takes you beyond just comparing experiences, which has value but is limited. The added value is more evident; the academics have put it in a framework; you understand why something is exceptional or typical, why it is effective or not' . Moreover, participants in the Inclusive Cities programme consistently noted how academic involvement added credibility to the programme and moved the experience outside the usual parameters of city networking initiatives-although measuring whether this substantially contributed to institutional support for the programme is outside the scope of our evidence.
Here, the embedded researcher plays a role of built-in learning partner as opposed to an external evaluator.
Retrenchments in UK local government budgets have affected not only operational capacity (Eckersley & Tobin, 2019) but also the (understandably contested) concept of policy capacity, reducing the resources available to make clear choices about strategic alternatives (Howlett & Lindquist, 2004). Consequently, this has reduced both the capacity and appetite for evaluation. By contrast, providing a 'sounding board' through the project represents an important resource for policymakers, though it is distinct from independent evaluation.

Providing a structured programme to enable progress
Besides the university-and academic-specific aspects of the City Working Group model, we also observe how the structured nature of the programme distinguishes it from other kinds of networks. For example, the primary 'product' of the Inclusive Cities process was a city action plan led by the local authority or municipality. Throughout, there were opportunities for reflections, updates, and support from the academic facilitation described earlier. Yet the action planning model itself developed over iterations of the project. As one participant in the Action Towards Inclusion project noted, 'the timescale of the project [initially 12 months] was too short: not enough time to develop and implement a plan and to garner lessons learnt. One and a half years for the cities working together was suggested, rather than the ten months that the project funding in effect allowed. While the tasks were appropriate as well as manageable with other workloads, the timescale was not' .
This informed the design of Inclusive Cities, with one participant noting that 'the methodology and process of IC has been very useful . . . [City X] has lots of experience in delivering cross-sectoral action plans, however this has had a significant impact very quickly as it has been incorporated into a City-wide strategy with strong political leadership in less than 2 years of it being established' . Another participant noted the role of 'support during and between meetings [which] has ensured that we keep up the pace, that deadlines are met and that the task group and wider Council can see that we are making progress and have objectives' . Of course, this reported sense of forward momentum and progress needs to be interpreted against a realistic view of local authorities' capacity that is often severely constrainedespecially in the United Kingdom where local government has seen a severe retrenchment of its resources since 2010.
To raise the likelihood of success, the City Working Group model also required participants to establish stakeholder taskforces. These comprised representatives from across the city-including in public, private, and voluntary sectorsand were tailored to the needs of each city. Generally, they involved groups such as local employers, community groups, universities, local press, and trade unions. Taskforces intended to widen the range of partners around the table: one participant stating that 'bringing various partners together under the Inclusive Cities banner has encouraged more joined up working and promoted a city-wide view as opposed to a city council view. There is a need for this kind of catalyst to encourage partners to bring ideas to the table in a cooperative spirit -and not "leave it to the Council"' .
Another participant highlighted the role of the programme in growing these horizontal networks, stating that 'it has supported my practice by giving me a wide network of contacts across the city in different organisations with common concerns.' Although partnership and collaboration are obviously not new for city networks, this structured programme aims to embed this approach as an example of 'networked governance' (de Graauw, 2015), by contrast to ad hoc and sporadic engagement.
Yet this approach displayed some limitations in practice. Notably, despite aiming to widen engagement beyond usual partners, taskforces were not entirely successful at achieving this. Broadhead (2020, p. 14) noted that 'all cities have found it difficult to gain sustained employer engagement and few have chosen to include community groups representing longer standing communities (alongside migrant groups)' . Although these limitations may be 'indicative of the complex factors at play in making horizontal, place based networks functional and supportive' (Broadhead, 2020, p. 14), they also hint towards the limits of academic facilitation. As such, programmes such as the City Working Group will require local leadership, political will, and operational capacity as they attempt to implement (or revise) structures and policies.

Peer learning among cities and researchers
The final element of the model was peer learning among cities and researchers. Having a structured space for learning was clearly important to city participants. Yet while being undoubtedly valuable (and valued) by policymakers, this was perhaps the least distinctive aspect compared to existing city networking programmes such as EUROCITIES, Intercultural Cities, and others (Haupt et al., 2020). Nevertheless, in the UK context, Inclusive Cities remains one of the few arenas for UK-based city networking on inclusion: other networks, such as City of Sanctuary, Belong, Core Cities, and Intercultural Cities UK have different point of emphasis (see e.g., Broadhead (forthcoming)).

Outputs and outcomes
The previous section focused on explaining how the City Working Group model functions-its processes and ways of working. Next, we consider its outputs and outcomes. There were six main types: (1) tangible action plans, (2) outcomes from the action plans, (3) municipal guidance, (4) changes in ways of working and policy framing, (5) research outcomes and influencing of research agendas, and (6) policy advocacy and developing a city leadership voice.

Tangible action plans
Fostering tangible policy change through a structured programme marks a key difference between the City Working Group model and other networks which focus on sharing experiences among participating cities to indirectly steering policy change (although we acknowledge how some networks' programmes, such as within EUROCITIES, use action planning processes). By contrast, in this model, cities develop policies directly as a pre-requisite and consequence of participation in the programme. This constitutes more direct influence over the policy-making process than seen in other city networks.
Outcomes from the action plans Yet developing and implementing an action plan, at best, demonstrates activities rather than impact: it highlights processes, not necessarily outcomes. The problem of how to measure performance-either generally in public sector administration or specifically to immigrant integration outcomes-has bedevilled both researchers and policymakers.
Establishing a credible 'golden thread' from policy goals to activity and impact within a single organization, let alone through multi-layered networks, is not only difficult but also can underestimate how constituent parts impact each other by prioritizing prescriptive measures and management (Micheli & Neely, 2010). What is more, specific to the domain of immigrant integration, no common measures exist (Harder et al., 2018)-and, even if they did, these measures would likely mainly capture national outcomes using available observational data.
Instead, where experimental evidence is not available or feasible to collect as a means of assessing causal impacts, a more productive approach would analyse the extent to which interventions create the conditions for change as a proxy measure of impact. This avoids the problem of relying on flawed outcome measurements or contexts in which it is not possible to craft credible treatment interventions. In these circumstances, intensive modes of engagement such as the ones used in Inclusive Cities and the City Working Group model may provide stronger links between interventions and outcomes (Lowe & Plimmer, 2019).

Municipal guidance
This process is most concretely demonstrated by examining how the Inclusive Cities framework-the guidance document setting out the five ways of working and five core thematic areas that contained sets of expectations that participating cities should work towards-was developed. While some aspects of the guidance were informed by research (e.g., defining integration following Spencer & Charsley, 2016), it was primarily produced in a two-way process involving researchers and city representatives at the end of the first phase of the programme. Moreover, cities' experiences up to that point fed into the content. The framework provides a degree of standardization for the programme, yet retains aspects of co-production and proactive reflection rather than attempts to set explicit performance targets. In this sense, it does not establish formal criteria for accreditation as, for example, a 'welcoming city' which would require participants to demonstrate they are meeting a number of standards. 14 As a result, the network acts in an enabling and facilitating manner, rather than a prescribed and controlling one.
Changes in ways of working and policy framing Shifts in thinking and reframing policy questions are other important outcomes of university-facilitated networks.
This is particularly noticeable in low-salience policy issues like immigrant integration that lack statutory duties and resources (Broadhead & Kierans, 2019). Participants reflected that 'being part of the programme has raised the profile across the City Council and across a wider group of stakeholders of the challenges and opportunities brought about through significant growth in the city in the last ten years, largely through European migration into the city' . Another noticed that participation had sparked further discussion that may lead to change: 'being part of the programme has acted as a catalyst for a wider conversation about difficult but critical issues for [City Y], which are leading to longer [term] and more systemic change' .
These shifts suggest how the City Working Group model can support agenda-setting processes (e.g., Cairney, 2018) despite the limited capacity of policymakers to focus on new issues. First, the structured nature of the programme provides a roadmap that fosters engagement and maintains momentum. Second, facilitation by researchers creates a space for extended reflection and discussion. Third, participation in the network itself raises the salience of the policy topic within member local authorities, and allows policymakers to focus on it, when otherwise they may not have time or institutional support that would allow them to pursue the agenda above other pressing commitments.

Research outcomes and influencing research agendas
While these networks primarily aim to foster changes in the participating cities, they can also influence research and conceptions of impact. Knowledge exchange programmes can generate research material, disseminate existing research findings, and identify new avenues for study. Specifically, the Inclusive Cities programme both illustrates and develops a view of knowledge production that is co-productive in nature and style. Here, we take co-production to mean that 'scientific and expert knowledge contribute to the construction of political reality' and that the 'relationship between social science and politics in this example is one of continuous mutual influence and reinforcement' (Boswell & Smith, 2017). As such, it is impossible to separate social scientific research from the domains of politics and policy as they are mutually constitutive. Instead, this interlinked nature can inform research as a way of providing solutions to policy dilemmas, but which also potentially creates new policy problems.

Policy advocacy and developing a city leadership voice
The concept of 'city leadership' remains central to many city networks' theories of change. For example, in the context of climate change, cities join networks for peer learning as well as a desire to influence politics (Haupt et al., 2020). This echoes the experience of the Inclusive Cities programme: participating cities had a clear appetite for policy advocacy on areas which affected local government, but over which they had limited control. After all, although the programme focused on areas that sit within the existing competencies of local government, cities must either develop policy advocacy positions regarding central UK government policies or opt-out altogether of these large policy areas. This was because of the centralized nature of migration governance, as well as the limited salience and resourcing of work on immigrant integration.
Of course, it is important to avoid overstating how, or whether, local 'place leadership' can generate its own power (see e.g., Sandford, 2019 in the context of English Metro Mayors; Hambleton, 2015). In a highly centralized system such as the United Kingdom, policy advocacy and developing a singular 'voice' of city leadership remains secondary to the practical outputs and outcomes listed above. Nevertheless, the fact that the programme includes the four nations of the United Kingdom opens up regional and devolved administrations as additional-and alternative-sites for policy advocacy. 15 by foregrounding universities as key players within these networks, we acknowledge how trends towards civic engagement on the part of academia-seen in various forms around the world-partly motivate and sustain these activities.

DISCUSSION
In response, we draw attention to three specific conclusions.
First, peer learning characterizes the objectives and operations of this network. This is an aspect that has not been explicitly described within existing models of knowledge exchange, yet plays a key role in helping partners achieve their goals. Why is this the case? Examining integration as a policy area in the United Kingdom, as well as considering the political geography of a country comprising devolved nations, provides some clues. The highly centralized nature of migration governance in the United Kingdom does not confer cities either policy space in which they can take decisions or agency to advocate for their priorities with central government-though some options may exist at devolved administration levels. Meanwhile, integration is a low-profile policy area in the United Kingdom (Katwala et al., 2017) with no national integration strategy, little available funding, and few resources for local authorities. These factors contribute to city-level policymakers' willingness to engage in peer learning: it may be a better use of limited resources that is more likely to result in sustainable change. By contrast, further investment in advocacy may struggle to be effective when there is a significant decoupling between the aims and objectives of central and local government (Scholten, 2016).
Second, peer learning is more than a mechanism for exchanging information: it also provides a way for network members to collectively identify and assert their priorities-effectively socializing them into a more cohesive and effective group with common interests. Besides achieving the instrumental learning on substantive topics that obviously occurs via peer learning, this is an important symbolic outcome, too. As such, we view city networks as examples of policy entrepreneurship (Cairney, 2018), where city-level policymakers gather attention around their preferred issues and transmit this onwards with the aim to set wider policy agendas (Baumgartner & Jones, 1993). To be clear, although we acknowledge how there may be more scope for entrepreneurship activities in low-salience issue areas like immigrant integration, we also caution that their actual effectiveness and impacts may be more marginal than commonly believed by supporters of city-led initiatives (Sandford, 2019).
Third, universities provide a distinctive role in developing and sustaining two-way processes of peer learning. They serve as an 'honest broker' among city network members by contributing organizational capacity, technical skills, and institutional credibility that facilitate interactions. Of course, this does not deny how universities have their own interests and agendas as they fulfil this role: funding pressures on academic departments, as well as financial and professional incentives for greater engagement with public stakeholders, increasingly motivate academic researchers to seek out policymakers (see, e.g., Smith & Stewart, 2017). Similarly, the motivations and incentives for universities institutionally may be very different from individual researchers and departments. As a result, there is a legitimate risk of research being led and defined by policy concerns (Bakewell, 2008). Yet as the experience of the Working Group demonstrates, an alternative model involving knowledge co-production among universities and city members is possible. This approach, closer to established models of brokerage (Knight & Lyall, 2013), casts researchers sitting within universities as more than fonts of expertise (Russeil & Healy, 2015). Instead, there are uniquely valuable links between academic debates and outputs on the one hand and policy-making on the other hand. Moreover, as university academics engage in these two-way brokerage activities, they contribute to city members' peer learning-and subsequent socialization.
Focusing attention away from evaluating the outcomes of networked activities that merely happen to involve universities, our conclusions add to current understandings and received wisdom about city networks by providing a grounded example of how universities specifically play distinctive roles in these networks' formation and operation.

ACKNOWLEDGMENTS
The empirical work reported in this paper was funded by grants from the Open Society Foundations (OSIFEgrid.453407) and the Paul Hamlyn Foundation (PHF-grid.481059.1). Additional support for William Allen came from the British Academy (grant number PF21∖210066).