Learning from Local Government Research Partnerships in a Fragmented Political Setting

: Research partnerships between scholars and local governments offer promise to advance scholarly understanding of local public administration and to improve the lives of people living and working in local communities. Yet political fragmentation complicates the prospect of broader learning from these partnerships and creates the risk that research partnerships will amplify disparities in local government performance. If scholars and practitioners are attentive to these risks, they can design research to facilitate learning across local government settings. Lessons from policy diffusion, program evaluation, and team science inform a set of recommendations for the conduct of local government research partnerships and the distribution of results.

D irect engagement between scholars and local governments offers many potential benefits on both sides. Governments can learn what works in their community and others; they can anticipate a broader range of policy effects that extend beyond their immediate goal; and they can consider time horizons or externalities that would not have informed their decisions otherwise. Benefits to scholars include introduction to the problems most relevant for local government, even if they are not priorities in the academic literature; exposure to the practical constraints and impacts of local government decisions; and enhanced ability to generalize and build theories of administrative behavior and policy impacts through the firsthand study of applications.
Recognizing these benefits, social scientists have long partnered with local governments on projects that bring together research expertise and local government practice through participant observation (Pressman 1975;Wolfinger 1973), technical and programmatic support (Ferman and Hill 2004), and experimental interventions (Arceneaux and Butler 2016;Linos and Riesch 2020). Partnerships of this type have received heightened attention in recent years because of formalized initiatives in Philadelphia (Philadelphia Behavioral Science Initiative), Washington, DC (The Lab @ DC), and a set of cities affiliated with the MetroLab Network to engage scholars directly in creating more effective policies through place-based interventions informed by social and behavioral science.
Although local governments may enter into research partnerships primarily to address their own policy problems, scholars and funders typically are interested in building knowledge that can transport to other settings. Political fragmentation presents an important challenge to the broader goal of integrating research with practice. A crucial consideration when drawing lessons from research partnerships is their situation in a particular local government context. Because of the large number of local governments in the United States and the wide variation between them in capacities and responsibilities, results obtained in one setting often will not generalize to others. This is particularly true if the setting for the research partnership is unusual, as is often the case. Even if results do generalize, a challenge remains in distributing information from research partnerships to governments that could benefit from the findings. Careful attention is needed to facilitate learning across local government settings.

Fragmentation and Heterogeneity in Local Governance
Fragmentation has long been a topic of interest to scholars of U.S. local governance. Over 90,000 local governments divide responsibility for providing public services including education, police and fire protection, social services, and water and transportation infrastructure.
The boundaries between local government jurisdictions are meaningful: they define rights, obligations, and access to services, and they serve to demarcate and solidify socioeconomic inequalities (Mullin 2014). Fragmentation enables socioeconomic sorting, promoting homogeneity within jurisdictions (Drier, Mollenkopf, and Swanstrom 2001;Weiher 1991). Economic and racial segregation between jurisdictions has been on the rise (Fischer et al. 2004), contributing to growing disparities between communities in demands for public services and fiscal resources to deliver them (Jimenez 2014). Different places are facing different challenges: either rapid population gain or ongoing population loss; skyrocketing housing costs or inability to maintain property values and a robust tax base. Challenges are often geographically situated, including some that are becoming more pronounced as a result of climate change and other large-scale forces (Mullin 2020). Collaboration and other strategies to address these challenges entail risk of failure that can have high financial and political costs (Hansen, Mullin, and Riggs 2020).
In short, there is wide variety in the responsibilities and capacities of U.S. local governments. Although all local governments face common challenges related to fiscal stress, aging facilities and infrastructure, and addressing the problems of their neediest residents, those with more resources are better positioned to confront these challenges. Many of the problems that local governments are tackling have less ideological content than issues on national agenda and would benefit from research that has been designed to support decision-making. Yet fragmentation poses two obstacles to broad learning from research partnerships: the findings from these partnerships might not generalize across settings, and many of the localities that could benefit most are unlikely to access the findings.

The Problem of Generalizing from Research Collaborations
Findings produced from local government research partnerships are context-dependent: they apply to a particular place, at a particular moment, and according to a particular research design (Cartwright and Hardie 2012;Orr et al. 2019). Institutions have an important role in shaping public service behavior and must be considered in research that focuses on micro-level behavior (Bertelli and Riccucci 2020). Of course, the problem of generalizing findings from research conducted in one context to a larger category of cases is endemic in social science. However in efforts to inform practice for local governments, it is important to recognize the ways that context can potentially shape results.
The local governments engaged in research partnerships are not representative of local governments nationwide. Those that have established ongoing policy laboratory programs are large cities with access to expertise from nearby universities or federal government and the capacity to compete successfully for support from private foundations (Wogan 2017). When scholars take the lead in finding partners, they typically select a purposive sample that takes advantage of geographic proximity or personal connections (Butler 2019;Olsen and Orr 2016). College towns, doubtlessly overrepresented as sites for research partnerships, are distinctive places characterized by high levels of education and income, low median age, and unusual diversity in residents' backgrounds (Gumprecht 2008).
Even if they are approached for a partnership, underresourced local governments may not find collaboration to be worthwhile. It is costly in time and staff resources and could be politically risky if constituents are distrustful of observers or research methods such as randomization. Unlike in the developing world where research collaborations with governments often have been facilitated by requirements of the World Bank and other funders, local governments in the United States have no external incentive to engage in research partnerships (Butler 2019).
The distinct characteristics of study sites affect the results. Where studies involve population response to some policy change or intervention, populations that are wealthier or more educated may respond differently on average than populations with lower incomes and education, and contextual conditions including population density and racial diversity are likely to influence response as well. Social science participation in policy design and evaluation can improve the quality of inference in a particular setting, but findings will not necessarily transport to localities with population and contextual conditions that differ. A second source of potential bias that stems from case selection lies in the capacity of the cooperating partner to implement a program that may be under study. This bias could operate in either of two directions in the common scenario that the local government engaged in a research partnership has higher capacity for implementation than other governments that may adopt the program later. One possibility is that estimates of program effectiveness will be higher than what we would see for implementation by governments with less capability. Alternatively, if the baseline level for the outcome measure is already high, estimates of effectiveness might be lower than would be the case in other communities (Allcott 2015).

The Problem of Learning from Research Collaborations
Accuracy in applying results from one local government context to another is only one part of the challenge that fragmentation presents for using research partnerships to build a broader understanding of local governance. A separate part is the ability to spread information about potentially useful results to the thousands of jurisdictions that may benefit from them. Shipan and Volden (2008) outline four mechanisms by which policies spread across localities: learning about successful policies from early adopters; imitation of larger cities; economic competition among proximate cities; and coercion or external incentives from state governments. Ideally, research partnerships with local governments would produce knowledge about the effectiveness of policies or practices in producing desired outcomes that could then be learned by other governments. Studies of policy diffusion point to several factors that might help facilitate this type of learning. One factor is geographic proximity (Berry and Berry 1990;Rubado 2020;Shipan and Volden 2008): although policies do not cluster regionally as much as they once did, information still flows more easily between communities that are near neighbors or within the same state. Informal networks among local officials are an avenue for information sharing, and these networks exist mostly at the local or regional level (Wolman and Page 2002). Intergovernmental professional associations and associations of governments can propel diffusion at a broader geographic scale by facilitating communication among officials responsible for particular areas of public policy (Balla 2001). Innovations may spread through the activities of policy entrepreneurs (Mintrom 1997)-individual advocates for policy change-including ambitious bureaucrats and city managers who may have the incentive to introduce innovations for career advancement (Kwon, Berry, and Feiock 2009;Teodoro 2009).
Research suggests that it is high-capacity governments that are most able to learn and take up policies from other jurisdictions. In state legislatures, professionalism is key to learning (Shipan and Volden 2006). Legislatures that meet part-time and do not have substantial staffs or salaries do not have the time and capacity to process policy information. Capacity is even lower, on average, among local governments (Svara 2003). To make use of information from research partnerships in other jurisdictions, local decision makers need to understand not only the findings in the original setting but also the likelihood that the same findings would apply in their setting. This type of analytic capacity is not evenly distributed; it is more likely to be seen in large or wealthier cities, and where local governments can attract mobile, professional managers or recruit locally from a more highly educated workforce base (Teodoro and Switzer 2016).
Action by independent local governments is not the only way for policy ideas to spread, however; diffusion also can occur through requirements and incentives from state and federal governments (Shipan and Volden 2008;Welch and Thompson 1980). A study on state adoption of accounting requirements for local governments shows that states were more likely to enact a mandate on localities where local governments had less capacity (Kim, McDonald, and Lee 2018). Whether this pattern stems from states' motivation to oversee financially vulnerable local governments or from lower-capacity governments having less ability to resist the state mandate, the result points to an alternative avenue for spreading evidence-based reforms beyond the uptake of reforms by (typically high-capacity) local governments acting on their own.

Recommendations for Using Research Collaborations to Improve Local Governance
The recent popularity of research collaborations with local governments, including those involving the use of randomized controlled trials as well as those using other methods, calls for consideration about how their results can be used to improve local governance more broadly. Because local governance in the United States is highly fragmented and heterogeneous, the convenience selection of study sites and existing mechanisms for policy diffusion may produce bias in who gets to learn from study results. To the extent that research partnerships are able to identify ways to improve policy effectiveness and government performance, the outcome of this bias could be to amplify existing disparities in the quality and capacity of local governance.
Does that mean that we should abandon these research partnerships? Certainly not! Policy and performance improvements in some places are better than none at all, and the tendency for these partnerships to be situated in large cities means that more local residents are able to experience any benefits that result. Furthermore, with attention to the risks of bias outlined here, it is possible to design research that can generalize more broadly and to improve circulation of useful findings.
An important first step is to design research using well-developed theory that elucidates mechanisms and contextual effects (Bowers and Testa 2019). To the extent that studies can identify not just whether a causal relationship exists between some intervention and an outcome but also why or for whom, which helps to build understanding about the conditions that would need to be present in other communities for the same relationship to hold. Scholars and governments then have more information to evaluate the appropriateness of a finding for a particular community and could even engage in proactive search for communities that have conditions favorable for benefiting from the finding.
Second, scholars seeking a local government project partner should consider reaching beyond the partners that are most convenient by virtue of geography or personal connections. In an ambitious recent effort along these lines, Goerger, Mummolo, and Westwood (2020) sent inquiries to approximately 3,000 local police and sheriff 's departments to discuss potential collaborations, following up with all departments that responded affirmatively with offers to develop partnerships at the scholars' expense. Levine (2020b) implemented a similar recruitment model for partnerships with civic organizations. Even with smaller budgets, scholars can be more deliberate in finding jurisdictions dissimilar from those commonly represented in research studies. Distance can be a barrier for reasons of cost, logistics, or student involvement, but a benefit of fragmentation is that wherever a scholar may be located, there are likely to be many local governments nearby. Whether through random selection from within some sampling frame (Olsen and Orr 2016) or a purposive selection method, scholars could engage in more active search and outreach for partners, perhaps working with intermediary organizations such as regional councils of governments.
Third, in order to engage a broader range of project partners, scholars may need to be flexible about the nature of the partnership, allowing local governments a role-and potentially leadershipin setting the agenda. The new initiative research4impact, a matchmaking service using broad search to establish connections between scholars and practitioners, learned in its first year that many organizations seek shorter-duration, lower-cost forms of engagement with scholars rather than full collaboration in an original research study (Levine 2020a). Scholars seeking deeper engagement with a local government might consider co-production, a concept from environmental science by which scholars and practitioners together create knowledge through an iterative process. Practitioners are not treated as the end users of knowledge created by scientists but rather as active participants in shaping research questions and approaches. Co-production is increasingly an expectation or even a requirement  (Lemos et al. 2018). Although the concept shares some underpinnings with the public administration literature's treatment of co-production as citizen participation in the delivery of local services (Nabatchi, Sancino, and Sicilia 2017;Ostrom and Ostrom 1977), the two forms of co-production rarely are treated together (Bremer and Meisch 2017). There would be reward in cross-fertilization, as the growing literature on knowledge co-production is fruitfully grappling with questions about the ethics of engagement with communities, processes for learning across contexts and scales, and the importance of boundary organizations that can facilitate co-production and the sharing of produced knowledge.
Fourth, a range of actors and organizations can be recruited to help spread the word about findings that have the potential to improve local governance. Boundary organizations that bridge between research and practice-including, among others, university extension programs and private data firms that work with local governments-often have the capacity to interpret results and analyze how they might generalize across communities. Associations of governments and public officials have well-established communication networks and information flows. Organizations with a regional basis or that concentrate on rural communities are trusted sources of technical support for low-capacity governments. Finally, as the literature on policy diffusion has shown, states can play a critical role in spurring the spread of innovations. By working directly with lawmakers and relevant state agencies, scholars can help design programs that use the capacities and incentives of the state to advance local government performance. Even there, scholars need to be attentive to generalizability from purposive sampling, as studies tend to occur in states with well-developed, high-quality data systems (Stuart et al. 2017).
Uptake of the recommendations summarized in Table 1 requires flexibility not only from researchers, but from academic institutions as well. For scholars to have latitude in designing projects that respond to the needs of local governments, tenure and promotion guidelines and departmental cultures must recognize the value of applied research. Broader representation of local government partners in the academic literature requires journals and peer reviewers to be receptive to studies conducted in small communities as well as in large cities. Grant agencies and university spending rules must accommodate adequate compensation for the work of boundary organizations and other collaborators in communicating results and spreading innovations.
On the whole, partnerships with local governments offer great promise to advance scholarly understanding of local public service delivery and to improve the lives of people living in local communities. Deliberate attention to where and how we conduct these studies and how we distribute results can help to maximize benefits from partnerships and avoid aggravating disparities in local performance.