Cross-sector collaboration in cities: learning journey or blame game?

Abstract

The success of cross-sector collaborations (CSCs) in cities is mixed, and important questions remain about what distinguishes effective from ineffective collaborations. This comparative case study examined nine CSCs in three US cities, focusing on three public policy areas: education, economic development, and public safety. Nine group interviews, 110 individual interviews, and analysis of archival documents revealed common patterns that allow us to build grounded theory about the roots of CSC success. We propose that how a collaboration responds to setbacks plays a crucial role. Success arises in collaborations that respond to setbacks with a process of mutual learning, in which participants anticipate each other’s actions, devise new ways of apportioning labor, and approach problems collectively. In contrast, failure follows when setbacks lead collaborations into a process of mutual blaming. No single mode of network governance is especially associated with success, but more successful collaborations tend to be characterized by adaptability concerning governance mode. Mutual learning appears to be facilitated by a few key actions: building on prior relationships, relying on trusted key participants, engaging with the community, using data to advantage, and investing in joint problem-solving. Our findings suggest that collaborative leaders in public, private, and nonprofit organizations should emphasize these key actions to enable collaboration and facilitate mutual learning.

Introduction

When tackling complex social and economic problems, public, private, and nonprofit organizations in cities often collaborate across sectoral boundaries (Goldsmith and Coleman 2022; Martínez Orbegozo et al. 2022). City governments alone rarely have enough resources, authority, and expertise to achieve public goals like reducing crime or creating economic opportunity, so they partner with businesses and nonprofits to pursue such aims. Similarly, private and nonprofit organizations often rely on effective collaboration with city hall and with each other to accomplish their goals. Increasingly, scholars and practitioners conclude that addressing complex social issues—often characterized by uncertainty, volatility and value conflict—requires cross-sector collaboration (CSC) (Ambos and Tatarinov 2023; Ferraro, Etzion, and Gehman 2015).

Although common, CSCs in cities vary dramatically in how effective they are (Bryson, Crosby, and Stone 2015; Gray and Purdy 2018). While some scholars have synthesized the lessons of successful cases (Ansell and Gash 2008), other researchers have underlined the insights from failures and uneven results (Andrews and Entwistle 2010; Babiak and Thibault 2009; Martínez Orbegozo et al. 2022). As Gray and Purdy remind us, partnerships are not panaceas, and “simply teaming up with other stakeholders does not offer a magic bullet for tackling an issue” (2018, p. 11).

While CSC outcomes vary widely, research to date has provided little insight regarding what distinguishes successful CSCs from failures. In this paper, we aim to build grounded theory about the roots of such differences. Specifically, we ask: what helps and what hinders the design and management of CSC in cities?

To shed light on this question, we take a distinctive empirical approach. Prior papers on this subject fall into three categories. In the first category, each paper analyzes a single CSC in depth, exploring the dynamics of collaboration in a specific context. It is difficult to generalize insights from such studies beyond individual cases. Studies in a second category focus on comparing multiple CSCs within a single area of policy—for instance, education. These studies generate some conclusions that generalize across local contexts but only within a particular policy area. Finally, some papers present literature reviews that synthesize findings from case studies and suggest integrative models. These papers generalize across locations and policy areas, but their propositions remain tentative as they are based on underlying studies with different goals and methods. In contrast, the research in this article presents systematic, in-depth, and consistent analyses of cases across local contexts and policy areas.

Specifically, our comparative case-study approach employs a three-by-three matrix design: we examine nine CSCs, active in three cities and three policy areas, selected from a lengthy list of CSCs that met predetermined criteria. The matrix design enables us to compare across issues and within cities, across cities and within issues, and across cities and issues. For each collaboration, we collected in-depth data using individual interviews, group interviews, and group exercises. Our data analysis then combined established techniques from the grounded theory approach with iterative rounds of flexible coding.

Our findings suggest that how a collaboration reacts to setbacks plays a key role in its success or failure. All the CSCs we examined, including the successful ones, experienced setbacks—moments of stagnation or failure in the pursuit of public value. Examples include a loss of financial support from philanthropic backers or a loss at the ballot box. In some CSCs, partners responded by engaging in a process of mutual learning that strengthened the trust needed for effective collaboration, allowing these CSCs to spiral upward to success. In contrast, other CSCs triggered a process of mutual blaming in response to setbacks. Trust deteriorated, and a downward spiral toward failure began. Our findings also suggest certain key factors that dispose CSCs toward learning journeys rather than blame games when facing setbacks. Moreover, we propose that a collaboration’s success depends not on its mode of network governance but rather on its ability to adapt its governance mode over time.

The Prior Research section situates the present study within the existing literature and describes its contribution. The Methods section details our three-by-three design. The Findings section presents our findings and formulates three key propositions. The Discussion section concludes the analysis and considers its implications.

Prior Research

Defining cross-sector collaboration

Over the past two decades, scholars in public administration have offered varying definitions of CSC. These definitions consistently refer to organizations from multiple sectors coming together to solve social problems, but they differ regarding the degree of formality of these arrangements. Donahue and Zeckhauser (2011), for instance, underscore the importance of a “carefully structured arrangement” that links different sectors’ capabilities, and Provan and Kenis (2008) and Emerson, Nabatchi, and Balogh (2012) emphasize the importance of structures and processes that signify a formal collaboration. In contrast, Gray and Purdy (2018) characterize a CSC as any form of initiative where three or more types of stakeholders partner to achieve a common purpose toward transforming an institutional field, and Bardach (2001) describes an informal culture in a multiagency arrangement that becomes a “psychological reality” for participants beyond its formal structure. Within this assortment of definitions, some scholars have focused on delimiting the formality of specific types of collaborative arrangements, such as public–private partnerships (Hodge and Greve 2007, 2017), while others have embraced wider definitions that encompass multiple possible formal or informal network combinations ranging from roundtables to transnational accords (Gray 1989; Gray and Purdy 2018).

Crosby and Bryson (2005) reconcile various definitions by plotting how organizations collaborate along a continuum from essentially working independently, as in silos, to working together so closely that they merge into a new entity. CSC occupies the middle ground on this spectrum. Building on this idea, Bryson, Crosby, and Stone (2006) offer a more specific definition that emphasizes the voluntary association of organizations engaged in sharing, for example, information, resources, and capabilities. In line with this work, we define CSC as occurring when organizations from at least two sectors (public, private, and nonprofit) collectively engage in a structured, voluntary effort to produce public value. Purpose (the creation of value that no partner can produce individually), structure (the commitment to some enduring form of governance), and lack of coercion (voluntary participation with ongoing flexibility) are key elements of this definition.

Crosby and Bryson (2005) are not the only scholars who have provided a taxonomy of collaborative arrangements. A growing stream of literature has suggested a continuum of ways to work together that are labeled “cooperation,” “coordination,” or “collaboration” (Castañer and Oliveira 2020; Himmelman 2002; Stout and Keast 2021). For Brown and Keast (2003), for instance, this continuum varies from loosely structured arrangements (cooperation) to fully integrated systems (collaboration). McNamara (2012, 2015) builds on this continuum to suggest more dimensions that differentiate collaboration from cooperation and coordination, such as the level of organizational autonomy, the governance design, or the level of trust. A systematic review of the management literature on inter-organizational relationships (IORs) by Castañer and Oliveira has aimed to address conceptual inconsistencies by proposing that while coordination and cooperation refer to respectively determination and execution of shared goals, collaboration refers to “voluntarily helping others to attain a shared IOR goal or private goal” (2020: 994). As the lines between these concepts in the literature remain unclear, we build on their proposal by defining collaboration in terms of its voluntary nature and structured effort to help others accomplish a shared goal, which, for this study, is public purpose.

Revisiting success and failure

Multiple studies have examined the enablers and challenges of CSCs in different contexts. For instance, some scholars have identified a set of obstacles to success (Gray 1989) and factors for enhancing collaborative advantage (Huxham 1996). Quantitative research has validated dimensions that define successful collaboration (Thomson, Perry, and Miller 2009), while qualitative studies identify enablers from more diverse contexts (Kapucu and Hu 2022).

Yet, important questions remain about what leads CSCs to succeed or fail. These questions persist partly because most prior research has been conducted within the context of a single case study in a specific location or through comparative studies within a single policy domain. While we know of barriers and enablers limited to these contexts, we know less about whether these elements apply across locales and policy domains. Synthesizing literature reviews have compiled these lists of enablers and barriers, but it is unclear how these findings can inform the design and management of specific CSCs. We do not know whether CSC enablers that apply to education are also relevant to economic development. A review of the literature clarifies why we need to revisit the causes of success and failure.

Studies of single CSCs.

Many studies provide a deep understanding of a single collaboration in a particular setting. Examples include in-depth explorations of CSCs involved in employment services in Australia (Baker, Kan, and Teo 2011), corporate citizenship in Austria (Battisti 2009), healthcare information in Europe (Reypens, Lievens, and Blazevic 2016), elder care in Finland (Grudinschi et al. 2013), and regional innovation in Texas and Oklahoma (Bland et al. 2010), among many others. Each of these studies examines the underlying dynamics, characteristics, and evolution of a specific collaboration in depth, with a focus on stakeholders and their interactions. Each study, however, is specific not only to one location but also to a single public challenge. This makes it difficult to generalize the findings of any study beyond its particulars.

Studies of multiple CSCs in a single area.

Other studies compare multiple collaborations that tackle a specific type of public challenge. Examples include research on CSCs in housing (Madden 2017), government services (Donahue and Zeckhauser 2011), transportation policy (Bryson et al. 2011), public health (Douglas and Ansell 2021), and economic development (Agranoff and McGuire 2004). For example, by surveying CSCs in the education space in 100 US cities and comparing them, Henig et al. (2016) identified patterns among education-related CSCs. Similarly, Grossman and Lombard (2015) explored CSCs engaged in educational efforts from a business sector perspective through interviews with business leaders. Thomson, Perry, and Miller (2009) surveyed organizations participating in a national service program to identify common dimensions of collaboration. Such studies can generate hypotheses about CSCs in a particular policy area across geographies. Yet, their findings are hard to generalize beyond a particular policy arena.

Syntheses of frameworks and theories.

In the late 2000s, scholars undertook the task of organizing and synthesizing the literature on collaborative public management, governance, and networks that had been growing over three prior decades. A foundational work in this area is Bryson, Crosby, and Stone's (2006, 2015) comprehensive literature review, which organizes past research into a framework with seven conceptual categories and twenty-two propositions. Another seminal piece is Ansell and Gash’s (2008) meta-analysis of 137 cases of collaborative governance, which led them to propose a contingency model that includes four sets of key variables and a series of factors for the collaborative process. A third widely cited review is Emerson, Nabatchi, and Balogh's (2012) integrative framework, which synthesizes cases and previous conceptual frameworks. Other syntheses that have influenced the field include Kania and Kramer’s (2011) collective impact framework and Forrer’s (2014) continuum of CSC arrangements. Both works offered typologies and proposed conditions for success, such as having a backbone organization and a shared measurement system.

An important synthesis is Provan and Kenis’s (2008) theory of modes of collaborative network governance, which we employ to analyze our data. Provan and Kenis identify three modes. In shared governance, a network of organizations acts as a collective, no single member organization dominates, and decision-making is a shared responsibility. In lead-organization governance, one organization takes the lead—serving as administrator and facilitating activities. In network administrative organization (NAO) governance, an entirely separate entity is created to govern the network and its activities. These authors develop propositions about when each mode is effective and what tensions each creates. Other scholars have built on this model to suggest essential tasks for network managers (e.g., Milward and Provan 2006). As we explain below, we use Provan and Kenis’s theory to suggest how adaptability in network governance plays a role in the success of CSCs.

Other recent literature reviews have underscored the differences in various definitions of collaboration and units of analysis (Amsler and O’Leary 2017; Intindola, Weisinger, and Gomez 2016) or have synthesized literature for particular types of collaborative arrangements or outcomes (Brogaard 2021; Clarke and Fuller 2010; Cristofoli et al. 2022; Innes and Booher 1999). These papers have highlighted the importance of context and the need for more clarity on the contextual factors within which CSCs arise—such as the institutional landscape, the type of public issue, or participating sectors. Although these studies build relevant theories and tools, they do not conduct original empirical research, nor do their syntheses of findings account for differences across institutional settings or types of public challenges. It remains unclear whether and how these inventories from a multitude of studies in a host of contexts can apply to specific settings.

A systematic study across locales and policy domains

Overall, prior research on CSCs lacks a systematic study that spans locales and policy domains. Our research was designed to fill this gap. Our approach occupies a helpful middle ground between the depth and empirical grounding that case studies provide and the generalizability of insights that syntheses promise. We propose a comparative, multiple case research design in which we examine different collaborations across both locations and policy issues.

This design has multiple benefits. First, it allows us to explore the commonalities across locations (within issues) and across issues (within locations). Prior literature has highlighted how multiple elements of the context—including policy and legal frameworks (Bingham 2008), power relations, administrative and regulatory institutions (Ansell and Gash 2008), and socioeconomic and cultural characteristics (Sabatier et al. 2005)—shape collaborations (Emerson, Nabatchi, and Balogh 2012). Our research design aims to control for some of these contextual factors across two dimensions: the characteristics common to a location (i.e., cultural and political) and the features common to the issue (i.e., policy and regulation). Therefore, our methodological approach is more robust for understanding the role of the context. Second and related, the design provides enough variety of settings to start to propose generalizations across locations and policy issues.

Third, our approach allows for a deeper exploration of collaborative learning dynamics. Prior research has focused on specific static factors that hinder or promote collaboration (e.g., agreement on scope in Gray [1989] or unclear responsibilities in Kapucu and Hu [2020]). However, a focus on specific static factors can cause one to overlook the dynamics of success or failure. By piecing together detailed histories of a set of CSCs across sectors and locations, we can bring to the surface important collaborative learning dynamics that prior research has hinted at but not examined fully (Argyris 1982; Woolcott et al. 2023).

The locations we focus on are cities. Cities, as concentrations of people, are places where social issues manifest themselves concretely. City government is closest to these issues and is often expected to tackle them, but typically lacks the resources and authority to do so by itself (Goldsmith and Coleman 2022; Schragger 2016). Seemingly intractable problems therefore often require city-level CSCs (Martínez Orbegozo et al. 2022; Quayle, Grosvold, and Chapple 2019).

We elaborate on our design in the following section.

Methods

Research design

We build grounded theory about the roots of CSC success using a comparative multiple-case embedded design, as defined by Yin (2018). The collaboration was the main unit of analysis. We studied nine collaborations in three cities focused on three different policy issues that were the same for all cities (Table 1).

Table 1.

Three-by-three methodology.

	Issue 1	Issue 2	Issue 3
City 1	Collaboration 1	Collaboration 2	Collaboration 3
City 2	Collaboration 4	Collaboration 5	Collaboration 6
City 3	Collaboration 7	Collaboration 8	Collaboration 9

This approach allows us to compare cases from different institutional contexts (horizontal comparisons in Table 1) and policy areas (vertical comparisons) using standardized data collection across all nine cases. This helps us to explore similarities and differences at three levels of analysis: across cities (issue-specific patterns), across issues (city-specific patterns), and across cities across issues (overall collaborative patterns). Systematic analysis of some dimensions while holding others constant enriched the analytic quality and elevated the validity of the findings (George and Bennett 2005; Yin 2018).

Case selection and participant recruitment

Rigorous screening criteria were established around cities, types of policy issues, and collaboration characteristics. Knowing that our research would require access to and candor from leaders of local CSCs, we started with a list of fifty-nine US cities whose mayors had attended a leadership development program that we helped organize and deliver. We anticipated that the mayors would introduce us to CSC leaders and encourage them to participate in our research, and this turned out to be true. These cities represented a convenience sample with significant diversity in terms of geographical representation, sizes, and government types.

With this list in hand and to avoid false comparisons between cities, we focused on cities with populations above 100,000 and strong mayoral systems. Mayors in strong mayoral systems tend to have greater influence and convening power.¹ Our focus on such mayors had an important benefit: it improved the odds that we would get intimate access to CSC leaders. The focus also had a cost: our findings might not generalize beyond cities with strong mayoral systems. These criteria yielded a short list of thirty-five cities.

We identified six policy challenges commonly addressed using cross-sector approaches—housing, education, public safety, economic development, workforce development, and infrastructure—and screened for cities that were using CSCs to address several of these challenges. We also required collaborations to involve at least two sectors and to have existed for three to seven years. We wanted to focus our analysis on the challenges during the development stage of collaborations (Kapucu and Hu 2020). Younger collaborations (formation stage), we feared, might not have a track record sufficient to judge success or failure beyond overcoming the early-stage turmoil of finding an entry point (Martínez Orbegozo et al. 2022). For older collaborations (resilience stage), we worried that it would be difficult to piece together an accurate history of their origins and early (mis)steps due to recall bias (Fraser, Greene, and Mole 2007; Mezias and Starbuck 2003).

With these criteria in mind, we ordered the cities based on how long they had worked with us, and then in alphabetical order. We went down the list and, in each city, used public sources to identify candidate CSCs in the six policy domains. We then had exploratory calls with city representatives to determine whether each collaboration was still functioning, to ask for possible points of contact inside the CSC, and to identify other possible CSCs. Next, screening calls with points of contact in each collaboration allowed us to determine whether each candidate fulfilled our CSC definition and had been operating for three to seven years.

In the fourth city that we examined, for instance, we identified seven possible collaborations, but only four of them fulfilled our CSC definition and time criterion. The collaboration there in public safety, for example, had eleven member organizations from three different sectors that came together to think creatively about solutions to reduce gun violence (purpose). In this collaboration, the convener was the mayor, who invited different stakeholders to meet regularly to provide recommendations (structure). The CSC had been working for three years (time), and some invited members had freely stopped attending, while other organizations had joined through the years (lack of coercion).

We continued down the list of cities, identifying and screening CSCs until we completed a matrix—that is, found three cities with CSCs that addressed the same three policy issues. This occurred after we had examined seven cities, identified thirty-nine candidate collaborations in those cities across the six issues, and found twenty-two collaborations that fulfilled our CSC definition and time criterion (see Appendix A1).

The policy issues that fulfilled our design were education, public safety, and economic development. The cities must remain anonymous to preserve confidentiality, and we will refer to them as Cities Alpha, Beta, and Gamma. Unintentionally, the cities included one in the northeast United States, one in the mid-Atlantic region, and one in the southeast. Despite the commonality in policy issues and locations, the selected CSCs had very diverse objectives, ranging from building trauma-sensitive schools to promoting a robust innovation ecosystem. Table 2 provides brief descriptions of the selected CSCs. Appendix A2 shows how each fulfills our CSC definition.

Table 2.

Descriptions of the final sample of CSCs.

	Education	Public Safety	Economic Development
City Alpha	Collaboration composed of a group of organizations, residents, and youth working in evidence-based initiatives. Their shared purpose is to ameliorate trauma, build trauma-sensitive schools, and support families in culturally relevant ways	Collaboration composed of community organizations, members of law enforcement, and the mayor’s policy team. Their shared purpose is to reduce gun violence	Collaboration composed of a group of community leaders. Their shared purpose is to create a vision for rehabilitating a river corridor and identifying joint opportunities to reinvigorate economic growth
City Beta	Collaboration composed of a group of partner organizations. Their shared purpose is to improve access to high-quality STEM resources and opportunities for underserved and underrepresented students	Collaboration composed of organizations focused on reducing violent crime in a particular area of the city. Their shared purpose is to alter the perception of youth aged fourteen to seventeen regarding group violence and employing the focused-deterrence model	Collaboration composed of a coalition of coalitions. Their shared purpose is to improve transportation infrastructure and help reduce traffic congestion through implementation of an earmarked sales tax
City Gamma	Collaboration composed of government agencies and organizations. Their shared purpose is to connect childcare providers and families seeking high-quality, affordable early childhood learning opportunities	Collaboration composed of organizations working directly with violence-prone gang members. Their shared purpose is to minimize arrests and incarcerations and foster police-community collaboration through the focused-deterrence model	Collaboration composed of city agencies and local start-ups. Their shared purpose is to provide opportunities for entrepreneurs to test new products and for government agencies to innovate and explore new technologies and services to improve operations

Having selected a set of CSCs, we then made an introductory call to a point of contact for each collaboration to explain the research objective, ask the CSC to participate in the research, and ask the point of contact to introduce us by email to individuals representing the various organizations in the collaboration. We searched for partner organizations and prioritized them based on conversations with the point of contact or through desk research. When lacking a formal email introduction, we used cold emailing and snowball sampling to contact in each collaboration at least ten partners representing all sectors and perspectives. Prior relationships with city officials provided entry points to each city. In previous research and educational programs, we had worked with mayors and city leaders in the cities that were part of our convenience sample for more than two years. The trust built through these programs enabled us to gain access to individuals quickly.

Data collection and analysis

Primary and secondary data were collected for all collaborations using various techniques, including individual interviews, group interviews, collective exercises, and archival documents. We recorded and transcribed 110 semi-structured individual interviews conducted via Zoom. (See Appendix B for the number of interviewees for each CSC.) To learn more about the group dynamics and collective collaborative process, we brought individuals from contributing organizations together for a group interview that included an exercise in which individual members ranked the most challenging barriers to collaboration. Responses were aggregated at the collaboration level and used to encourage participants to share their interpretations of the collective results during the group interviews. When these data had been collected for all collaborations in each city, we conducted a sixty-minute semi-structured interview with the mayor to understand better the institutional environment and overall government involvement in each policy challenge. (See Appendices C1–C4 for protocols.)

Data were analyzed via sequential cycles of analytic exercises following Yin’s (2018) multiple case study procedure and Edmondson and McManus’s (2007) iterative funnel. The analytic process involved four stages. (See Appendix D1 for a graphical explanation.)

• Stage 1: within-case analysis. We employed a grounded theory approach (Glaser and Strauss 1967) to analyze individual cases. Theoretical frameworks from prior literature that came to light as data collection continued during analysis informed the construction of our codebook in terms of enablers of and barriers to collaboration, leading us to shift from an initial codebook with primary codes to flexible coding (Deterding and Waters 2021) that combined deductive and inductive codes. After performing individual coding using Dedoose software, the authors met as a team to discuss emerging patterns in each collaboration. Case findings were summarized in memos capturing the main categories of enablers and barriers and other relevant categories that reflected our initial questions.

• Stage 2: within-issue synthesis. Within each policy issue, we compared the three cases that focused on that issue, revised individual analyses, and collectively identified patterns that emerged from all three cases. We met as a team to discuss aspects of enablers, barriers, and modes of network governance that were common across cases, and we consolidated our preliminary findings in a memo per issue.

• Stage 3: within-city analysis. We compared the cases for each city across policy issues. At this stage, we triangulated our within-city observations with the insights obtained from the interviews with the mayors, aiming to identify city-specific patterns as well as commonalities across cities.

• Stage 4: cross-city, cross-issue synthesis. We focused on identifying overarching patterns across all collaborations, building on patterns identified in stages 2 and 3.

The multiple rounds of iterative analysis required returning to analyses from previous stages to ensure that the evidence supporting each code matched the overall patterns emerging. Thus, we avoided losing track of the granularity of the data at successive levels of aggregation. The multiple cycles involved going back and forth between final constructs and individual cases to develop a theory that felt robust across issues and cases (see Appendix D2). These steps surfaced additional questions and led us to view the data through different lenses at each point.

Defining successful CSCs

Despite having similar lifespans, some of the collaborations were more successful than others. To develop hypotheses about what helps or hinders CSC in cities, we placed each CSC along a spectrum from more successful to less successful.

The success of a CSC can be assessed in a variety of ways. Ultimately, all the CSCs we studied defined success in terms of social outcomes: a positive impact on the lives of residents. While different CSCs used different indicators of success and different quantitative and qualitative measures, they all aimed to deliver public value (Moore 2013). Public value, understood as the “net good” a collaborative effort produces, is a multifaceted concept. For example, CSCs focused on public safety typically aimed to reduce crime but not at the expense of community trust and racial justice. CSCs focused on economic development usually sought to enable growth without increasing inequality. And CSCs focused on education aimed to optimize educational opportunities without infringing upon the free choice of families.

The studied CSCs were all “middle-aged” (between three and seven years old). This meant that none of them were nascent, and none were fully mature either. Some had produced tangible desired outcomes, while others were struggling to develop a sound course of action. No CSC had yet solved a problem fully or created a sustainable impact on the community, but some had made significantly more progress than others toward delivering public value.

Our research yielded multiple data points that indicated how much progress each CSC had made toward delivering public value. The mayors we interviewed gave their frank assessments. We asked each CSC leader to assess the collaboration’s effectiveness. We asked all team members to rate their collaboration’s success on a ten-point scale, identify the outcome they were most proud of, and explain why they considered that outcome a success. In addition, we studied archival sources that gave data reflecting effectiveness—for instance, crime statistics for public safety CSCs, clients served for education CSCs, and programs launched for economic development CSCs. Some of the indicators of success were tied to the ultimate outcomes (reduction of crime and placement of students), while others could be considered proxy indicators (services delivered and programs launched). Because public value is multifaceted and can take decades to materialize, we combined multiple indicators and data from many sources for each CSC to determine the progress it had made toward delivering public value. (Appendix E1 shows indicators of progress for each CSC.)

Using this approach to assess and rank the CSCs, we found stark differences in progress across the nine CSCs. For instance, the Economy-Beta CSC aimed to develop infrastructure to alleviate traffic congestion. By the end of our study period, Economy-Beta had rallied bipartisan support for a sales tax increase to fund infrastructure investments, helped convince voters to approve it, devised a way to prioritize specific projects, and helped get the city council to approve construction. In contrast, the Education-Beta CSC sought to improve the STEM knowledge and skills of underserved students. Despite considerable early momentum, Education-Beta was providing services in just one of its region’s many school districts, had been “evicted” by a larger nonprofit that initially hosted it, and had lost the confidence of once-supportive business leaders (by the end of the study period).

With multiple inputs in mind, each author individually assessed each CSC’s progress toward delivering public value. We then compared and reconciled to arrive at a shared assessment. The inputs, coupled with the stark differences in progress, made agreement among ourselves straightforward (see Appendix E2). This process allowed us to classify our nine collaborations into three tiers of success (Table 3).

Table 3.

Three tiers of CSC success.

Findings

We preface the proposed answers to our question—what helps and what hinders the design and management of CSC in cities?—by noting some of the factors we did not find to be associated with success. No city stood out as having an institutional environment more conducive to collaboration, and in no city did all three collaborations thrive despite all three mayors being favorably disposed to collaboration. The institutional context of a city may be important, but it appears insufficient to explain success or lack thereof. Neither does our evidence support the notion that a particular policy issue is more likely to be addressed successfully by a CSC. Each policy issue included one more successful and one less successful (Table 3).

Propositions 1 and 2 identify factors that distinguished the three more successful collaborations. Proposition 3 identifies five key actions revealed by our synthesis of the analyses of all nine collaborations to recur across all collaborations, not just the more successful ones.

Proposition 1: CSCs that embrace mutual learning rather than mutual blaming are more likely to succeed

As we iterated on the codes that characterized the more successful and contrasted them with the less successful collaborations, a strong pattern emerged in the trajectory that collaborative processes followed. All collaborations required initial trust among the partners of the collaboration to get off the ground. This is consistent with prior literature that recognizes the importance of trust as a pivotal initial ingredient (Emerson, Nabatchi, and Balogh 2012). The more successful collaborations (the ones that delivered more public value) entered a process that reinforced trust, while the less successful ones fell into a cycle that slowly eroded trust. We use the definition of trust in CSCs proposed by Martínez Orbegozo et al. (2022, p. 621), which encompasses the “expression of confidence between collaborating parties and an interpersonal expectation that others’ actions will not be harmful or exploit any perceived vulnerabilities” (Edmondson 2004; Jones and George 1998; Mayer, Davis, and Schoorman 1995). Trust has been identified as a necessary condition for collaboration (Bryson, Crosby, and Stone 2015), and the process of trust-building has been recognized as a requirement for success (Ansell and Gash 2008; Emerson, Nabatchi, and Balogh 2012).

Consistent with prior scholarly work, all the CSCs we examined encountered setbacks in their early days (e.g., personnel departures, funding cuts, ballot losses, and partner defections). We use the term setbacks since these were moments of stagnation or failure in the pursuit of public value and put partners’ trust in the CSC and each other to the test. What distinguished more successful CSCs from less successful was the way they responded to these setbacks. Our findings provide a more nuanced understanding of the process by which collaborations engaged in the trust-building process over time. The more successful collaborations responded to setbacks by engaging in a trust-building process of mutual learning, while the less successful CSCs responded in ways that led to a trust-eroding process of mutual blaming.

Mutual learning loop.

The initial level of trust that brought participants together in pursuit of a common goal (see fig. 1) rapidly transitioned to members’ commitment to do things together, and subsequently to action manifested in activities, programs, or other joint efforts. Successful outcomes of these actions reinforced trust, which encouraged members’ continued commitment and further action. Our evidence suggests that collaborations’ responses began to diverge, however, when action caused a collaboration to hit a setback. Some collaborations then engaged in a process of mutual learning and adopted a joint-problem-solving orientation (Kerrissey, Mayo, and Edmondson 2021), meaning that members saw problems as shared and solutions as requiring co-production. These CSCs learned to create spaces to solve problems together, anticipate each other’s actions, and divide responsibilities among themselves. This joint problem-solving orientation closed the mutual learning loop since, regardless of the intermediate outcomes, the collaboration continued to nurture trust among the members, allowing them to engage further in commitment and collective action. CSCs that engaged in mutual learning wound up among the more successful we observed. Unfortunately, this was hard, and not all CSCs followed this path.

Figure 1.

Mutual learning and mutual blaming loops.

Mutual blaming loop.

Other collaborations started with initial trust that led to preliminary commitment and collective action. Some of these early efforts led to short-term gains and quick wins. Yet, when the collaboration hit a setback, these CSCs—rather than engaging in mutual learning—turned to mutual blaming: each member placed responsibility on another actor rather than the whole group, and infighting and inaction ensued. Trust eroded, commitment declined, collective action decreased, and little public value was created.

Collaborations that gravitated toward a mutual learning loop tended to strengthen while those that gravitated to a mutual blaming loop tended to decay.  Exhibit A illustrates in detail the process of mutual learning that City Beta’s Economic Development CSC experienced, while  Exhibit B walks through the process of mutual blaming that undermined City Beta’s Education CSC. Participants and organizations are anonymized. These two exhibits illustrate extreme cases. We also observed CSCs that engaged sometimes in mutual learning and sometimes in mutual blaming. Appendix F describes one such collaboration: City Gamma’s moderately successful Education CSC, whose setbacks sparked episodes of both mutual learning and mutual blaming.

Proposition 2: Success is associated not with a collaboration’s mode of network governance but with how adaptable a collaboration is in its mode of network governance

Building on Provan and Kenis (2008), we wanted to explore whether a particular governance mode might be associated with collaborative success. Having learned from participants about the history of each collaboration, we also wondered whether the evolution of a CSC’s governance mode might influence its success.

Our evidence (fig. 2) suggests that no single mode of network governance (shared, lead organization, or NAO) is associated with success or failure. The three less successful collaborations used three different modes of network governance, and the three more successful ones also employed three different modes at different times.

Figure 2.

Modes of network governance and adaptability in network governance. Note: Red = less success; yellow = intermediate success; green = more success; the lines from CSC Econ-Beta 1 to 2 and from Safety-Beta 1 to 2 indicate that these CSCs changed their governance structures within the same mode of governance. Modes of network governance are based on Provan and Kenis (2008).

What did differentiate more successful CSCs, however, was their willingness to adapt their governance structures. None of the less successful collaborations switched governance modes, while two of the three more successful collaborations changed modes. Our detailed data suggest that members of the more successful CSCs, recognizing that a structure which may have served their CSC well in its early days was no longer effective, figured out how to reorganize themselves to better fit the collaboration’s purpose. A member of City Alpha’s education CSC—a more successful case—recounted how it had adapted its governance mode while continuing to prioritize the mutual learning loop.

In some ways [the collaboration] has completely changed what its role was while actually remaining totally true to what its original vision was. If you think about the original one, it was a bunch of independent organizations talking about where they wanted to go. Ideally, let’s get people aligned. Well, now it’s the [network administrative organization] that is in some ways setting the path for how those organizations can move forward. (…) But each step along the way, we carried off those transitions in a way that did not alienate or lose support, but in fact, created a way to expand support.

Accordingly, we propose that collaborative success is associated not with a particular mode of network governance but rather with the ability to adapt one’s mode.

Proposition 3: Five key enabling factors can launch a CSC on a path of mutual learning

As we analyzed each collaboration, we identified factors that partners said enabled the success of the collaboration. We coded these enablers for each collaboration and then, in our cross-case analysis, looked across all nine collaborations for recurring ones. In figure 3, we illustrate our process with two of the nine CSCs: we constructed the full list of enablers for each collaboration, sought out common themes, and ultimately distilled them to a set of key enabling factors. We observed that some of the recurring factors were more prevalent in the more successful collaborations. Appendix G gives a more detailed recounting of all enablers for the nine CSCs and Appendix H provides anonymized quotations from interviewees that illustrate common enablers for each CSC.

Figure 3.

Individual and common enabling factors for CSCs

This process pinpointed five key enablers: building on prior relationships, reliance on a trusted participant, engaging the community, using data and evidence, and investing in joint problem-solving. Next, we elaborate on each factor.

Building on prior relationships.

Collaborations that engaged in mutual learning tended to be built on preexisting professional or personal relationships. Individuals from different partner organizations invited other organizations or trusted individuals with whom they had worked before in other settings. Prior relationships fostered the trust that enabled an initial commitment to a CSC (Kapucu and Hu 2020), and relationships encouraged collaborators to learn rather than blame when a setback hit. For instance, an individual in a member organization of City Alpha’s Public Safety CSC recalled that the relationship-building process started long before the formal creation of the initiative.

We have spent a decade building relationships with both rank and file, but also importantly, with city officials and with the former chief [of police], the current chief, and the commissioner, really rigorously building relationships with them. Then we are able to come to this table as equal partners. We know that we carry an amount of responsibility for the city, and we are an integral part of the public safety fabric.

Similarly, a leader in City Alpha’s Economic Development CSC highlighted how repeated encounters with other leaders concerned with the same issues fostered the trust that facilitated the collaboration’s launch.

It is a lot of the same players who have been working together for a really long time. There is a lot of this built-in history and trust. It feels like a safe space in that way. It is a group of people across the board; it is the heads of organizations mostly.

Relying on a trusted individual.

More successful collaborations were also characterized by the presence of an individual or small group of individuals who had gained the trust of actors across sectors and could sustain trust in the face of a setback. Never newcomers, these boundary-spanners (Ryan and O’Malley 2016) were well known in different sectors, had built strong reputations through work in different roles and organizations, and had demonstrated that they genuinely cared about the particular challenge. For example, the individual appointed director of City Gamma’s Public Safety CSC had worked on safety issues for more than a decade alongside community members, advocacy organizations, police officers, convicts, and ex-offenders. A member of City Gamma’s police department observed:

One thing about [the director] is that he had a good relationship with a lot of the probation officers. So instead of the police trying to make that happen, we let him make that happen. So that meant that we got a lot of collaboration from the probation officers through the coordinator (…) And he was able to get that relationship going. It is just a matter of letting the right person drive the bus. We figured that out, and that bus has been driving. The people at the top may not be able to do it. Everybody has a relationship with somebody, and when you can figure that out, you are golden.

Similarly, in City Alpha’s Education CSC, almost every member recognized the importance of the collaboration’s director. A leader in a nonprofit partner acknowledged the director’s role in fostering trust:

The whole time that this has been happening, I can’t even tell you how many times the director [of the collaboration] checks in on everyone. Not because of the work, but because she genuinely cares about the people who are being impacted by what is happening. I think that’s important. I think that is a piece of the puzzle that, besides the work and the dollars, there is this investment in really caring about the relationship between the school district and their organization.

Engaging the community.

Once collaborations were up and running, a key factor that allowed them to translate trust into collective commitment and action was community engagement. Community engagement is the process of involving people benefiting from—or otherwise affected by—policies and programs when developing, delivering, and evaluating these policies and programs. This is done by soliciting input and feedback or mobilizing their capacity to co-produce public value (Arnstein 1969; Gilman et al. 2023). Members of our CSCs were aware that to achieve impact, they needed to include the broader community—not just for input but also for legitimacy and support. Leaders of City Gamma’s Public Safety CSC, for example, discovered that they needed to understand the human networks in their city. Toward that end, they deployed outreach workers and police officers in the community. A nonprofit leader pointed out:

We integrate ourselves in existing spaces. So, it is not often that we call a meeting that is just the members of the collaboration. We go into existing meetings that are already established in the community because the leadership structure is already there. They already have an infrastructure in place, and we just come in as guests. We talk about the work that the city is doing. We talk about the work of our street outreach workers. We talk about what it would look like or what it does look like in the communities and how it has been beneficial, but more importantly what the volunteer opportunities are. How can you get involved in this as well so that we can make sure that we are not keeping this with just law enforcement or just our outreach workers, but making sure that people understand that this is an initiative that impacts the community as a whole? So, it’s important for the community to be involved as well.

A similar approach arose in City Beta’s Economic Development CSC, which aimed to solve transportation problems across the metropolitan area. The mayor opened spaces to speak with the community about problems and possible solutions, and community input proved fundamental in identifying and prioritizing projects. A business leader involved in the collaboration remarked:

I am a firm believer that you can’t do it all by yourself. If you try to do it by yourself, you are probably not going to be very successful. This program was very important to the community. I think the mayor got a lot of buy-in from throughout the community. To me, the most successful part was going to the meetings and listening to the people and what their needs were. They had an opportunity to talk about it, and we had a chance to respond to it, to go into those meetings and have people tell you, “I trust what you’re going to do. You’ve been around a long time and when you say what you are going to do, you do it.”

Using data and evidence.

Leaders of many of our nine CSCs pointed to the collection and analysis of data and evidence as a key enabler of collaboration. A focus on data and evidence (they reported) gave them a better grasp of the problems they were tackling, generated conversations that built mutual understanding, kept discussions grounded in outcomes rather than politics, and centered the group on learning rather than blaming when setbacks arose. One member of an academic organization in City Beta’s Public Safety CSC commented:

We collected all the law enforcement data, but one of the main things we did was that we coordinated monthly meetings with all the representatives of law enforcement (…) We would go through a whole list of every single homicide that happened in the city over the past 30 days and talk about each case, and everybody would have information about that particular case. And all of it was very much aimed at group violence (…) So it was a coordination of all the information, very specific information. That was part of the data we collected. Also, we had to collect information about community activities, who was there, where was the activity, what was the activity for, lots of qualitative data collection on the community, focus groups in the community, lots of interaction with the community.

Members of City Beta’s Economic Development CSC, when seeking community input to prioritize possible infrastructure projects, recognized the need for data that could inform and support their decisions, particularly to gather political support. One of the community leaders explained:

We want to make sure that the roads that you fix are being done all in a good priority and not the priority that every council member says. But, really, that started with the basis of data-driven solutions to prioritize, because politicians historically will do what they believe their local voters want them to do. You really have to take a step back and be willing to do something for the common good. The only way to give a politician, in my opinion, the political courage to do that is to, in fact, have data determine what should be done first. So that is what we advocated for. That was a key piece. We said, “We have to see the data, we have to understand, and we have to support the conclusions of the data.”

Deliberately investing in joint problem-solving.

A joint problem-solving orientation does not emerge spontaneously; rather, it results from conscious dedication by most members of a collaboration. Members of City Alpha’s and City Gamma’s Education CSCs, for instance, invested time and energy in (bi)weekly meetings at which they would troubleshoot problems, fostering constant communication that helped members see problems as shared and solutions as co-produced. An individual in an organization of City Alpha’s CSC offered:

What I am really happy about is the biweekly meetings. Sometimes they meet more often. Sometimes it could be weekly as well, if the need is there, meeting as a team, problem-solving as things happen (…) [L]ast year, we were sort of meeting ad hoc and as things were coming up, we were addressing them, whereas now there’s constant communication between both sides, and it is almost more preventative and action-oriented.

Similarly, an individual in City Gamma’s Education CSC reported:

We got through it because number one, we were committed to the work. Number two, we made sure that we were finding time to come together to talk about what was working, what wasn’t working, and what we needed to change. Then we also had to find time to go and support each one of our teams. It was a little stressful for a while, but I think, you know, really as organizations we were very committed to making sure that we were going to see this through, and we were going to figure it out.

Among these five factors, all appeared in the more successful CSCs (see Appendix G), and some were present in the less successful ones. That said, we do not suggest that these factors are necessary conditions for a successful CSC. Rather, we propose that these factors improve the odds of success in different institutional contexts and for different policy challenges, especially by launching CSCs onto a path of mutual learning. Therefore, practitioners who design and manage CSCs should consider them.

Discussion

City leaders engage frequently in cross-sector collaboration, yet their results are mixed (Andrews and Entwistle 2010; Bryson, Crosby, and Stone 2015; Kale and Singh 2009). In this study, we attempted to build grounded theory about what distinguishes successful from unsuccessful collaborations. The three-by-three design of our comparative case study approach allowed us to make comparisons and identify patterns across cities and across policy areas.

Our main conceptual contribution is to propose that a CSC’s success depends not on the choice of governance structure or the circumstances of a particular city or policy domain, but rather on the learning dynamics of the collaboration. More specifically, we suggest that setbacks in the pursuit of public value are inevitable moments of truth: the way that a collaboration responds to setbacks—mutual learning or mutual blaming—is pivotal to its odds of success. Likewise, we propose that a collaboration’s success hinges not on its choice of a network governance mode but rather on its ability to adapt as circumstances change.

Prior integrative frameworks have suggested three broad dimensions that influence collaborative outcomes: system context, governance structure, and collaborative dynamics (Bryson, Crosby, and Stone 2015; Emerson, Nabatchi, and Balogh 2012). Our research helps disentangle the relevance of some of these elements.

Regarding system context, our evidence did not suggest that some cities have produced institutional conditions that render collaborations successful in all policy areas. On the contrary, all cities in our sample hosted collaborations in different policy areas that varied in their effectiveness. Nor did we find evidence that collaboration tends to be more successful for one policy issue. On the contrary, each policy area included both more and less successful collaborations.

Interestingly, the mode of governance does not appear to be correlated with success or failure, but adaptability concerning governance mode is. The more effective collaborations we observed were able to modify their governance configuration within a given mode or shift their governance structure from one mode to another.

This brings attention to the relevance of collaborative dynamics. We contribute by providing a more granular understanding of how these dynamics operate. Our analysis suggested that collaborations begin with an initial level of trust that brings organizations together around a commitment to engage in collective actions in pursuit of public value. Early wins increase trust and deepen commitment. Collaborations diverge, however, when they hit setbacks. The more successful collaborations respond by adopting a joint-problem-solving orientation characterized by a view of problems as shared and solutions as requiring co-production. The less successful react by assigning responsibility to one actor rather than to the group and devolving into inaction or infighting as trust erodes. We name the former response a mutual learning loop and the latter a mutual blaming loop.

Our evidence regarding collaborative dynamics supports some of the claims in Ansell and Gash’s (2008) and Emerson, Nabatchi, and Balogh's (2012) synthetic frameworks, particularly regarding the presence of virtuous cycles. Yet, our findings are importantly different from prior research, as we observe that the early actions undertaken by CSCs very often lead to “small losses,” not just “small wins” (Ansell and Gash 2008). We propose that a big part of the identified virtuous cycle hinges on what happens after a collaboration experiences a setback. Whether the group embarks on a learning journey or resorts to a blame game can define their likelihood of success. In all collaborations, we observed breakdown moments. What characterizes those who bounced back—we propose—is their learning approach toward these setbacks.

Our focus on learning in the face of setbacks allows our findings to contribute to prior literature that has looked at learning as a feature of collaborative networks and organizations (Argyris 1982; Gray 1989; Woolcott et al. 2023). For example, some scholars have argued that through the back-and-forth interactions among individuals in a collaborative network, memory accumulates within the group. This can lead to a warehouse of cultural capital (i.e., knowledge, skills, and experiences) that is available to those joining (Woolcott et al. 2023). Our findings complement this literature by providing more nuance on what happens in these exchanges. Our evidence suggests that it is not any interaction but those that follow the setbacks—and it is not any exchange but those that foster mutual learning through joint problem solving—that allow a CSC to accumulate cultural capital rather than deplete it.

Our proposition on collaborative dynamics is also closely related to the literature on trust-building. Previous studies have highlighted the relevance of trust and trust-building (Bryson, Crosby, and Stone 2015; Emerson, Nabatchi, and Balogh 2012). Our propositions theorize trust as an enabling factor an individual can bring to a nascent CSC and as an initial asset that builds on prior relationships and brings partners together for joint action. Our study highlights trust-building as a continuous process that builds or erodes trust, depending on the response to setbacks.

Finally, our research pointed to five key factors that enable collaborative success. The factors we identified are not necessarily new to the vast inventory of design elements in collaborative governance. For example, Kapucu and Hu (2020) and other scholars have recognized the influence of past collaborative experiences on trust and legitimacy (Evans et al. 2014). Likewise, prior research has highlighted the reliance on a trusted individual, a factor that some have framed as facilitative leadership (Vangen and Huxham 2003). Our contribution here is to take the long list of factors suggested by existing literature, pass them through the filter of what is relevant for mutual learning across cities and policy domains, and propose five especially salient factors. We suggest that these factors can launch a CSC towards a reinforcing cycle of mutual learning.

Limitations and future research

Our propositions put forward that these five factors, adaptability in governance mode, and mutual learning all cause certain collaborations to succeed. It is important to acknowledge, however, that our observations could be given alternative causal interpretations. Take, for instance, the observation that adaptability in governance mode is correlated with CSC success. This pattern could arise, as we propose, because adaptability causes success. However, an alternative explanation could be that successful collaborations show greater adaptability as a result of their success and the increased insight, trust, and confidence it breeds. Additionally, the people we interviewed and the authors of the documents we analyzed have their own biased perspectives. Their interpretations of facts and events are not necessarily the only ones possible. Further research is needed to test the causal relationships between variables in the propositions we advance.

The range of policy issues considered is a further limitation of the present study. Other challenges addressed through collaboration (e.g., housing and infrastructure) might benefit from our findings but also entail characteristics not accounted for. Ours being the first three-by-three study in this arena, we welcome further comparative research aimed at expanding our findings to support generalization to more sectors and settings. A further limitation on the generalizability of our findings is imposed by particularities of the cities and the sample of selected CSCs. Our sample includes US cities that were similar in population size and mayoral system, and to which we had unique access because their mayors had attended our program. More comparative case studies are needed, including ones of cities in the Global South, the characteristics of which may be quite different in terms of population size, mayoral system, or other selection criteria. Finally, studying the development phase of CSCs limits the generalizability: our findings do not necessarily apply to more mature CSCs. The conceptual and empirical work in this article points the way toward similar systematic research on collaborations in later phases.

Implications for practice

If borne out by future testing, our three propositions suggest clear, practical implications for collaborative leaders in public, private, and nonprofit organizations. First, leaders should emphasize the five key enabling factors at the outset of a collaboration. By capitalizing on existing relationships, rallying behind a trusted leader, engaging with the community, using data to make decisions, and creating spaces and methods to solve problems together, they can help steer a CSC toward mutual learning.

Second, leaders should spend time upfront with partners discussing how they will react to setbacks. After launch, a CSC’s members should be attentive to how their collaboration responds to early setbacks. Members can proactively devise a plan that encourages learning, reinforces trust among members, and creates spaces and structures in which to solve problems jointly.

Third, leaders should avoid the trap of seeking the perfect governance model. Early on, leaders should not focus overly on identifying the most appropriate governance mode. As the collaboration evolves, members should also be attentive to whether the governance mode continues to be the best fit for its goals. If not, they should be adaptable enough to make adjustments or migrate to another mode.

Collaborating across sectoral boundaries to solve complex social issues is both necessary and challenging. Success is never guaranteed, but it is more likely when leaders are thoughtful in designing and managing collaborative efforts, adaptive with regard to structure, and committed to learning together—not blaming others—when they face setbacks.

Acknowledgments

The authors wish to thank Angelo Kalaw, Eva Flavia Martinez Orbegozo, Jennifer Nilsen, Sunghea Khil, Isabel Garcia, Kelly Mallon, Wanjiku Ngare, and Anahide Nahhal for their invaluable research assistance on this project. We also thank Mark H. Moore for his generous feedback and our research interviewees for their time, insights, and candor.

Ethics approval

This manuscript presents data that are subject to human subjects’ approval. This research, with protocol number IRB19-1233, was approved by the Harvard University Institutional Review Board on November 14, 2019.

Funding

This work was supported by the Bloomberg Harvard City Leadership Initiative.

Data availability

The text, exhibits, figures, and online appendices of this article provide extensive quotations, interview guides, diagrams of the analytic process, and detailed case descriptions. We cannot share additional data because the research participants were promised confidentiality as they shared sensitive details of their actions and relationships. It was infeasible to produce and provide comprehensive interview transcripts that were both meaningful and anonymized enough to protect the privacy of individual participants and the relationships among them.

Conflict of Interest

The authors declare no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Footnotes

1

“Mayoral system” refers to the level of political power and administrative authority a mayor exerts within the city administration. In a city with a strong mayoral system, executive power is centralized in a mayor elected by the citizens of the city. In a city with a weak mayoral system, the mayor is selected from among members of the city council, and executive responsibilities are distributed among the council members (League of Cities 2022; Saffelll and Gilbreth 1982).

Exhibit A: Example of a mutual learning loop

Before City Beta’s Mayor was elected, traffic congestion in the city had worsened for over a decade. To tackle this problem, the Mayor put infrastructure improvement atop the agenda. A year after being elected, the Mayor launched a collaboration to spearhead a public campaign to approve a sales tax that would finance projects to improve traffic problems. Yet, the journey of mutual learning for this collaboration had begun years earlier.

Gaining initial trust and commitment

Starting four years before the Mayor was elected, a coalition of three coalitions assembled to support an infrastructure agenda. Coalition 1 included leaders of important industries at the regional level. Coalition 2, led by the local Chamber of Commerce, brought together businesses to promote economic development. Coalition 3 incorporated the first two groups under a broader umbrella that also included nonprofits and community organizations. It focused on advocacy for better regional infrastructure solutions. These three coalitions reached an understanding of their respective roles:

What was good about this arrangement was that Coalitions 1 and 2 were happy to hold up Coalition 3 as the leader on infrastructure. Leaders of Coalitions 1 and 2 had told their members that they were involved with Coalition 3 and that Coalition 3 could advance transportation issues in the city. So they were motivated for Coalition 3 to successfully lead this effort. No group ever tried to get in front of the others. Coalition 3 was the clear leader. Everybody understood the role the other groups were playing.

Taking action and hitting setbacks

The same year the Mayor was elected, Coalition 3 tried to pass a ballot measure to increase property taxes to support infrastructure. Voters rejected the measure by a large margin.

Engaging in joint problem-solving

Rather than being discouraged by the ballot setback, the leaders of Coalitions 1–3 joined with the newly elected Mayor to launch a new Economic Development Collaboration. The Collaboration engaged in joint problem-solving to learn from election failure. The Collaboration’s leaders used the setback to pinpoint three major barriers facing infrastructure investment: (1) the public in Beta City’s region had a general anti-tax sentiment; (2) tensions existed between local and regional stakeholders; and (3) the tax proposal had had no political champion.

To overcome these barriers, the Collaboration’s leaders went through multiple iterations to understand each barrier better and coproduce solutions for them. In response to the anti-tax sentiment, the Collaboration started an extensive community engagement process—more than twenty meetings across the city—to listen to people’s concerns. This showed them that the anti-tax sentiment was rooted in a generalized distrust of government spending. To build trust, the collaboration decided to suggest a tax earmarked to a locked-down list of specific infrastructure projects. This list would also solve some of the misalignments between local and regional levels by making very clear how each part of the region would benefit from a new tax. In following iterations, the Collaboration undertook a careful prioritization process that used data-driven criteria to evaluate each project.

We wanted to prioritize the list ahead of time. We wanted to clarify how the list was going to be prioritized. How do we make sure that things get done in the right order? We worked to get alignment on a metric: dollars spent on congestion relief.

Also, the collaboration secured bipartisan support from top business leaders and brought the Mayor onboard as the project’s political champion. Recognizing the alliance’s importance, the Mayor joined and made it her top priority. Other sectors would later recognize how fundamental her leadership had been:

It took the Mayor doing it, and I think that’s the case anywhere, you know. You’ve got to have a driving force, and it really was the Mayor’s total focus to make that happen. We all know coalitions aren’t always easy to keep together. She had to keep nudging and keep people moving the right way. She had been in elected office—not as mayor but in other offices—so she knew the community and the players well.

The journey of mutual learning

The Collaboration’s iterative engagement in joint problem-solving as they hit barriers led them into a virtuous cycle: a journey of mutual learning about other sectors helped them build trust and engage in further collective actions, resulting in new co-produced solutions. Through this process, individuals from different sectors learned more about other sectors’ interests, resources, and operational capacities. They also learned how different sectors tend to see certain elements, such as implementation speed, differently. A leader from the private sector remarked:

I think the other difficult conversation for us involved the speed at which government moves versus the speed at which we believe private industry can move. You make your data-driven decisions [as the private sector] and you start moving. Well, you then encounter things such as Right of Way. And then if we’re going to tear up the road, we have to put the sewer system in. So for us as the business community, we have to accept the fact that things move at much, much slower rates than we believe they should. We have to feel comfortable that the rate of progress will likely be significantly slower than we would like it to be.

The project proposed by the Collaboration and championed by the Mayor ultimately received bipartisan and cross-sectoral support. Two years after the Mayor's election, more than 60 percent of voters approved a new ballot measure to raise taxes and improve infrastructure. By the time we conducted our research, the resulting infrastructure projects were entering the implementation stage, and the city council had approved the start of construction.

Exhibit B: Example of a mutual blaming loop

Gaining initial trust and commitment

City Beta’s Education Collaboration was launched when the convener, a retired teacher, set up a nonprofit with the local public school system and with financial support from a multinational firm. As envisioned by the convener, the Collaboration aimed to improve the STEM knowledge and skills of underserved students. This vision appealed widely to leaders in business, K-12 education, higher education, and community organizations: everyone agreed that STEM education could create jobs for young people and skilled workers for employers. The convener knew this idea could bring different stakeholders together. As one member of the collaboration pointed out:

Everybody has different objectives. What industry wants is to pull qualified candidates into their workforce. Academia wants to be the one that trains students and prepares people to be good candidates for those jobs. We [the nonprofit sector] want to make sure that we are supporting the education and preparedness of future generations. So while all of us have—in some respects—very different objectives, there are enough similarities to pull us all together.

Taking action and hitting setbacks

Working together over the first months, members of the Collaboration achieved some small wins. Short-term programs such as summer camps and weekend events were implemented. These successes broadened the Collaboration’s legitimacy and support within the community while strengthening trust among partners.

It was only a matter of time before the Collaboration hit setbacks. As external funding became less available, the nonprofit hosting the Collaboration faced difficult decisions regarding the best use of their funds. For the nonprofit, the local school district was a priority since this was its foundational mission. Private-sector leaders, in contrast, wanted to serve the broader region from which their workforce came, which spanned multiple school districts.

This regional-versus-local debate was exacerbated by generalized ambiguity about the nature of the Collaboration. Members wondered what specific objectives the Collaboration was trying to achieve and how the short-term projects aligned with a longer-term strategy. This was particularly relevant for private-sector leaders, who wanted to understand the strategy before providing more funding. Attempts to quell concerns led only to more ambiguity, especially because the convener saw the Collaboration as all-encompassing, saying “yes” to everyone. More and more, different members of the coalition felt that each sector and each organization had a different vision. Members highlighted some of these barriers:

I’ll just be honest. There was no central idea of what [the Collaboration] was. [The convener] was running it and had this abstract thought that it is an all-encompassing [collaboration]. And some of the arguments that I heard from the business community were: ‘But I can’t sell that to a business to get them to buy in.’ ‘Is this a clearing house? Is this just a place for ideas? Are we providing programming?’ [We] struggled with that for a while.

I was banging my head against a brick wall and telling my boss: ‘I’m trying to make this regional and they don’t do it.’ That’s when my boss put her foot down and said: ‘If they don’t make it regional, you’re not wasting your time there anymore. You have work to do. We have a five-year plan. In their five-year plan, each year is a new plan.’

Infighting and inaction

In response, the Collaboration did not engage in joint problem-solving. Rather, partners fell into a spiral of mutual blaming that led to infighting and inaction. Instead of seeing problems as shared, the different sectors focused on what divided their perspectives. One way this materialized was in the regional-local divide:

One of the challenges was regional versus local. That was a very hard discussion that happened over the course of months of going back and forth and back and forth. Are we regional? Are we just one school district? What are we? How are we going to do this?

The inability to prioritize shared objectives over diverging interests led to infighting, which reduced trust. This, instead of opening avenues for co-produced solutions, reinforced inaction. Members recognized that this inaction made the collaboration less effective:

For overall effectiveness, I would probably rate [the Collaboration] as a five [out of ten] because we didn’t get the job done. We didn’t lift off. There was no clear understanding of what it was. Different members thought it was different things. It did not launch. Students have not been impacted very much outside of the [local] school district.

The journey of mutual blaming

Continuous iterations of infighting and inaction led the collaboration to a journey of mutual blaming—a loop that erodes trust and individual commitment.

The entire thought of what the [Collaboration] was is another difficult problem. Ultimately that’s what led to the demise of the group. No one understood what we were and we couldn’t market that or explain it to anyone.

I don’t know what I would change other than maybe my attitude about it. I wish I’d been a little less skeptical and a little more expressive. We needed some more intentional trust-building. I think that’s the hardest part of all of this: trusting.

After several months of inaction, the organization hosting the Collaboration stopped doing so. The convener reached out to different members to see if anyone wanted to become a new backbone organization. Nobody stepped up. By the time we started collecting data for this research, the collaboration was in a stalemate, still waiting for an organization to take the lead.