Evaluating the Role of Government Collaboration in the Perceived Performance of Community-Based Nonprofits: Three Propositions

The topic of government–nonprofit collaboration continues to be much-discussed in the literature. However, there has been little consensus on whether and how collaborating with government is beneficial for the performance of community-based nonprofits. This article examines three dominant theoretical interpretations of the relationship between collaboration and performance: collaboration is necessary for the performance of nonprofits; the absence of collaboration is necessary for the performance of nonprofits; and the effect of collaboration is contingent on the nonprofits’ bridging and bonding network ties. Building on the ideas of governance, nonprofit, and social capital in their respective literature, this article uses set-theoretic methods (fsQCA) to conceptualize and test their relationship. Results show the pivotal role of the nonprofit’s network ties in mitigating the effects of either collaborating or abstaining from collaborating with government. Particularly, the political network ties of nonprofits are crucial to explaining the relationship between collaboration and performance. The evidence demonstrates the value of studying collaboration processes in context.


Introduction
Nonprofits-such as associations, trusts, and cooperatives-play an essential role in providing local community services (Marwell 2004;Milward and Provan 2000). They offer homeless people shelter, give extra food to the poor, or simply provide extra services to their community. The nonprofit literature has established that there is a wide variability among nonprofits regarding mission, function, and the provision of services (Galaskiewicz, Bielefeld, and Dowell 2006;Smith and Lipsky 1994). In this study, we focus on community-based nonprofits in which health and human services are provided to and on behalf of the community (Edelenbos, Van Meerkerk, and Schenk 2018;Marwell 2007). Although community-based nonprofits significantly contribute to addressing poverty and degeneration at the neighborhood level, they are, just like governments, unable to solve these kinds of complex issues in isolation (Halpern 1995;Marwell 2016). Hence, to address pressing social problems, nonprofits are increasingly engaged in cross-sector collaborations (Cornforth, Hayes, and Vangen 2015). By collaboration, we mean the process by which organizations with a stake in a problem seek a mutually determined solution (see Sink 1998). In this collaborative process, government and nonprofit organizations pursue joint objectives by sharing information, exchanging resources, and developing joint activities (Bryson, Crosby, and Stone 2016;Gazley 2008). Despite the growing interest of academics and practitioners in the potentials of government-nonprofit collaboration (Bryson, Crosby, and Stone 2006;Gazley and Guo 2020), it is still unclear whether this kind of collaboration benefits the local community. Systematic insight into the relationship between government-nonprofit collaboration and the performance of communitybased nonprofits, of which serving the local community is their main purpose, is scarce (see Cornforth, Hayes, and Vangen 2015;Stone and Sandfort 2009). Performance in this context is determined by considering the dimensions of "effectiveness," "legitimacy," and "resilience" (see Emerson and Nabatchi 2015;Hood 1991).
While some scholars argue that collaboration and performance go hand-in-hand (e.g., co-creating public value), others argue that collaboration with government poses huge risks for the performance of community-based nonprofits and should be avoided (e.g., protecting public value). Still others argue that the relationship between collaboration and performance depends on the power position of community-based nonprofits in terms of their community network (CN) and political network (PN) ties. This study aims to clarify this debate by unraveling which conditions are necessary and/or sufficient for the perceived performance of community-based nonprofits. Consequently, this study makes two major contributions. First, on a theoretical level, it contributes to the literature by clarifying a core aspect of the debate on government-nonprofit collaboration: does collaboration with government go hand-in-hand with outstanding performance of community-based nonprofits? And what role does the political and community network of community-based nonprofits play in explaining this set-relationship? It assesses these questions by connecting and combining contributions from the literature of three prominent bodies: collaborative governance, nonprofits, and social capital. In doing so, this article responds to the call for scholars to integrate contributions from multiple theories and disciplines in studying cross-sector collaborations (Bryson, Crosby, and Stone 2016;Cornforth, Hayes, and Vangen 2015). Second, on a methodological level, this study innovates the study of cross-sector collaborations by using a set-theoretic configurational comparative approach to unravel the complex and dynamic interplay between necessary and sufficient conditions (Bryson, Crosby, and Stone 2016;Schneider and Wagemann 2012). Contrary to previous studies that have mainly focused on variable-driven case studies or survey designs to study government-nonprofit collaboration (see Gazley and Guo 2020), this study relies on identifying setrelationships to provide critical insight into whether collaboration works only, or mainly, in combination with certain conditions. This study is structured as follows. First, the theoretical section describes the three interpretations of the relationship between government collaboration and perceived performance of community-based nonprofits. After describing the methods, data, and calibration strategy, the results are presented. In the final section, important conclusions and avenues for future research are discussed.

Explaining Performance: Three Interpretations
There is a massive literature on the relationship between collaboration and performance. Within this literature, it is possible to distinguish coherent clusters that share a specific focus on certain elements or values. For the purpose of this study, we have discerned three ideal-typical interpretations that reflect clusters in the collaborative governance, nonprofit, and social capital literature. Each interpretation, and its expectations on the relationship between collaboration and performance, will be discussed briefly. We do not strive toward a definitive clustering of the literature, but rather for a lens that can be used to empirically unravel the role of government collaboration in the performance of community-based nonprofits.

Interpretation 1: Collaboration Is Necessary for Performance
The first interpretation builds upon the idea that collaboration is a prerequisite for achieving outstanding performance. The core assumption that underlies the literature on (collaborative) governance is that tackling complex problems typically requires a combination of various resources that are owned or controlled by different organizations (Berry et al. 2004;Emerson, Nabatchi, and Balogh 2011). Overlapping missions, therefore, make it almost inevitable for governments and community-based nonprofits to engage in some sort of collaborative activity (CA) to accomplish their main objectives (Healey 2015;King and Cruickshank 2012). These CA can consist of sharing information, exchanging resources, and developing joint activities (Bryson, Crosby, and Stone 2016;Howlett, Kekez, and Poocharoen 2017). Pursuing CA with government enables the small and locally organized communitybased nonprofits to attract and acquire more resources for achieving their organization's mission, for example, outstanding performance. For community-based nonprofits, the financial and regulatory resources that governments possess are especially critical as they generally lack these resources (Dale and Newman 2010;Nederhand, Bekkers, and Voorberg 2016). In sum, this interpretation emphasizes the added (and necessary) value of collaboration for the performance of community-based nonprofits. Accordingly, it is Downloaded from https://academic.oup.com/jpart/advance-article/doi/10.1093/jopart/muaa059/6053749 by guest on 11 March 2021 hypothesized that the condition of CA is a necessary but not sufficient condition for perceived outstanding performance (P). The backward arrow "←" means "is necessary for": H1 : CA ← P Interpretation 2: The Absence of Collaboration Is Necessary for Performance The second interpretation argues that the absence of collaboration is necessary for achieving outstanding performance of community-based nonprofits. While the nonprofit literature agrees that CA between the public and nonprofit sector have both practical and political benefits, nevertheless, much of the relevant scholarship also highlights the potential disadvantages of a nonprofit sector, that is, too reliant on government funding and programs (see Brooks 2000;O'Regan and Oster 2002;Smith and Lipsky 1993). Being the weaker actor in relation to government, the small-scale local community-based nonprofits easily run the risk of being overruled and consequently lose some of their distinctive nature and qualities (see Anheier, Toepler, and Wojciech Sokolowski 1997;Brandsen, Trommel, and Verschuere 2017;Brooks 2002;Korosec and Berman 2006). For nonprofits, relying on government has been associated with a loss of managerial autonomy, mission infidelity, and bureaucratization (Eikenberry and Kluver 2004;Jang and Feiock 2007;Minkoff and Powell 2006;Salamon 2006;Suarez 2011). This, in turn, could lead nonprofits to prioritize performance measures related to external ideas of what "performance" looks like (efficiency, equality) at the expense of a more organic definition of performance that might uphold other values (interpersonal connection, responsiveness). These kinds of commitments greatly restrict the freedom of policy and action for nonprofits to be responsive to community needs (Smith and Lipsky 1994). This may even result in the destruction of the self-governance capacity of community-based nonprofits (Brandsen, Trommel, and Verschuere 2017;Korosec and Berman 2006). Moreover, the changedmore rule-bound-character of nonprofits can lead to diminished community support as people (donors and volunteers) are more attracted to community-based nonprofits that appear strong and independent and can maintain control over the organization (Brooks 2000). Or, as Smith and Lipsky (1994) point out, those that "[…] deal with citizens sympathetically and without having to reduce them to a set of official characteristics." As a result, some nonprofits avoid public money altogether out of concern for these threats to their performance (Gazley and Brudney 2007). According to Marwell and Calabrese (2015), the concern about the negative effect of government affiliation turns on a view of community-based nonprofits that privileges self-governance. Hence, it is hypothesized that the absence of CA is a necessary, but not a sufficient condition for perceived outstanding performance. The backward arrow "←" means "is necessary for," and the tilde sign " ~" denotes the absence of a factor: H2 :∼ CA ← P

Interpretation 3: Collaboration Interplays With Political and Community Network Ties
The third interpretation considers that the effect of government collaboration is contingent on its interplay with the CN and PN of community-based nonprofits. Based on social capital literature, this interpretation underlines the strategic importance of bonding and bridging network ties for enhancing the performance and relative power position of organizations (Galaskiewiz et al. 2006;Lin 2001;Szreter 2002). Whereas bonding ties refers to trusting and cooperative relations between people of a network with a shared social identity, bridging ties refers to relations between people who are heterogeneous in the sense of social identity (Putnam 2000;Szreter 2002). With regard to government-nonprofit collaboration, it may be expected that both types of network ties help to smooth the collaboration process with government by increasing the resistance of nonprofits to severe pressures from government. The presence of bridging PN ties has an important symbolic value. According to Lewis (2010), even if the PN is not activated, it can play a role in the background by enhancing the social standing of nonprofits. If they disagree with the way the collaboration is involved, they can try to go "over the heads" of public administrators by lobbying their superiors to overcome or reverse decisions. Therefore, community-based nonprofits with political influence have significant agenda-setting potential which somewhat equalizes the power balance between nonprofits and government. This strategic power resource could, in turn, foster a more careful and deliberate collaborative approach by public officials and a stronger negotiation position for organizations such as nonprofits to resist pressures (Lin 2001). The same goes for bonding CN ties. Community ties enhance the social standing of nonprofits as they increase their legitimacy as a collaborative partner (Edelenbos, Van Meerkerk, and Schenk 2018). Simultaneously, a close-knit CN can act as a buffer to government pressures by reinforcing the community identity and preserving resources. As such, a cohesive group tends to develop and guard group norms to prevent defection and to maintain the status quo (Coleman 1988;Putnam 2000). These network ties thus act as a buffer to protect community-based nonprofits against potential negative collaborative Downloaded from https://academic.oup.com/jpart/advance-article/doi/10.1093/jopart/muaa059/6053749 by guest on 11 March 2021 consequences. Accordingly, hypothesis 3 states that the combination of CA with either bridging PN ties or bonding CN ties is a sufficient condition for perceived outstanding performance.
H3 : CA × (PN + CN) → P When there are no CA between nonprofits and government, PN and CN ties make community-based nonprofits better resistant to a lack of resources. These networks are valuable alternatives for tapping into different capabilities, mobilizing resources, and transferring novel information (Provan et al. 2005;Shrestha 2018). When engaging in frequent contact, community-based nonprofits gain better knowledge on the relevance and timelines of the relevant resources of political officeholders and community grassroot organizations operating in the community. Moreover, greater familiarity with the nonprofit increases the readiness of partners to assist the community they care about (Shrestha 2018). These networks thus act as a buffer to protect community-based nonprofits against a lack of resources. Accordingly, hypothesis 4 states that the combination of the absence of CA with either PN ties or CN ties is a sufficient condition for perceived outstanding performance.
It should be noted that the first and second interpretations represent two different variants of the view on the government-nonprofit relationship. The first interpretation hypothesizes a positive role for CA in the performance of community-based nonprofits, whereas the second interpretation hypothesizes a negative role for CA with government. The third interpretation highlights the decisiveness of PN and CN in the interaction with CA to trigger the performance of communitybased nonprofits. Although these interpretations are compatible, they are not identical. Their compatibility lies in the notion that CA (or their absence) can be necessary for performance (H1 and H2) and, in combination with the network ties of community-based nonprofits, sufficient (H3 and H4). However, these interpretations are not identical as the first and second interpretations imply CA (or their absence) to be a prerequisite for performance (necessity). On the other hand, the third and fourth interpretations assume that CA (or their absence) in situations of strong network relationships typically result in perceived outstanding performance (sufficiency).

Methods
To clarify the questions of whether and how collaboration and network characteristics are necessary and/or sufficient for the perceived performance of nonprofits, 14 community-based nonprofits in the Netherlands were studied. In this section, we first describe the empirical setting of the study. Following that, we elaborate on fsQCA (fuzzy-set qualitative comparative analysis), the analytical tool used in this study. Finally, we turn to the operationalization and calibration of the conditions.

Community-Based Nonprofits in the Netherlands
The data used in this study stems from 14 communitybased health and human services nonprofit organizations in the Netherlands that were examined in the period between September 2017 and April 2018. We selected the nonprofits from the databases of Dutch umbrella organizations LSA, Vilans, and Kracht NL, by using the following three selection criteria. The first criterium was that the nonprofits had to be categorized as established organizations that transited the initiating phase. To ensure that the nonprofits were roughly in the same phase of development, we selected only cases that had been established between 2012 and 2015. We took 2012 as a starting point because this is the year that marks the start of major welfare sector reforms in the Netherlands in which the Dutch government decided to cut-back welfare budgets and delegate responsibilities "back" to communities (see Nederhand and Van Meerkerk 2018). Hence, the community-based nonprofits were developed in the anticipation of facing major welfare reforms and cuts. Given that these nonprofits were not "initiated" by the Dutch government as part of an ambition to contract out services, their development was autonomous. We took 2015 as a cutoff point to ensure that, by the time of data collection, the nonprofits from the sample were all well established. The second criterium was that the nonprofits should be truly community-based. This implies that they are independent, locally based organizations that provide services to residents in a particular geographical place ("community"). It is this requirement of serving a public rather than a private purpose that distinguishes nonprofits from associations. Community members participate in the organization's activities as staff, volunteers, and board members. Services are thus provided to and on behalf of the community. The third criterium for our selection of nonprofits was to consider their financial situation of whether they worked with a mixed revenue model, meaning that they were not solely reliant upon government funding. Moreover, to ensure that our cases formed a balanced reflection of the existing community-based nonprofits in the Netherlands, we included 4 cases located in small municipalities (<50k inhabitants); 3 cases in mediumsmall sized municipalities (50k-100k inhabitants); 4 cases in medium-large sized municipalities (100k-300k inhabitants); and 3 cases in large municipalities (<300k inhabitants).
To obtain systematic and comparable data, we combined two different methods: semi-structured interviews and surveys. In total, we conducted 50 semistructured interviews with nonprofit professionals and public officials. These respondents were selected on the basis of their close and strategic involvement with the community-based nonprofits. To represent the nonprofits, we selected the most active respondents, mostly initiators and board members who were involved in managing the nonprofit and its external contacts. To represent the governmental municipalities, we selected the public officials who had the most contact with a specific nonprofit. Based on these contacts, the selected representatives were able to answer questions about the local role of the nonprofit in the community. During the interviews, respondents were asked to fill out a survey about the cases, and all 50 respondents complied with this request. Additionally, four respondents only filled in the online survey. The respondents are spread evenly over the cases, with each case covered by 3-5 respondents.

Set-Theoretic Methods: A fsQCA
In this article, we are theoretically interested in examining relations between sets. For this reason, we employed the set-theoretic method of fsQCA (software: R packages QCA and SetMethods; Medzihorsky et al. 2016). A fsQCA allows for different degrees of set membership. An iterative dialogue between theoretical and substantive knowledge determines to what degree cases are members of a certain set. Thus, it established qualitative rather than quantitative differences between the cases.
In a fsQCA, relations are discussed in terms of necessity and sufficiency. A condition is necessary if performance cannot be produced without it; a condition is sufficient if it can produce the outcome by itself without the help of other conditions (Rihoux and Ragin 2009;Schneider and Wagemann 2010). The two main parameters of fit used to analyze the results of a fsQCA are coverage and consistency. Coverage states how well the available empirical information is explained by the condition(s). For necessary conditions, coverage expresses relevance in terms of the condition set not being much larger than the outcome set, and the relevance of necessity (RoN) in terms of the condition being close to constant. Low values indicate trivialness, whereas high values indicate relevance. The latter indicates the degree to which empirical evidence is in line with the statements of necessity or sufficiency (minimum of 0.75 for sufficient conditions, and 0.90 for necessary conditions). The proportional reduction in inconsistency (PRI) indicates the degree to which a given configuration is not simultaneously sufficient for both the occurrence and nonoccurrence of the outcome (see Schneider and Wagemann 2012;Thomann, Van Engen, and Tummers 2018).
The models presented in this study have the highest performance regarding the parameters of fit. The truth tables, directional expectations, conservative and parsimonious solutions, and simplifying assumptions are all provided in Appendices B and C.

Calibrating the Conditions
In this article, we study the performance of communitybased nonprofits using CA, the PN, and the CN of nonprofits as conditions. In this section, we elaborate on assigning set-memberships to our cases (see also tables 1-3 and Appendix A).
Each case will receive a score of 0 indicating full non-membership, 0.33 indicating partial non-membership, 0.67 indicating partial membership, or 1 indicating full membership. These scores display the membership of particular cases in each of the three conditions and the outcome.

Outstanding Performance
Calibrating outstanding performance is the first major task of this research. Since performance is an important element in this article, but also an essentially contested concept (see Johnsen 2005; Stewart and Walsh 2009), we first elaborate how we define performance. The academic literature has examined performance and its dimensions in many different ways. Following Provan and Kenis (2008), we argue that measuring performance is a normative task. First, multiple actors have different beliefs about the criteria of performance and, thus, selecting the preferences of one group over another or assigning weights to preferences is a normative decision; and second, the criteria for measuring performance are normative (Kenis and Provan 2009). According to Simon (1976), assessment criteria are elements of value rather than elements of facts. In this article, we focus on the dimensions of "effectiveness," "legitimacy," and "resilience" to determine performance (see Emerson and Nabatchi 2015;Hood 1991). Based on the work of Igalla, Edelenbos, and Van Meerkerk (2020), who translated these performance dimensions to the context of communitybased nonprofits, set membership is determined by the following three statements: "the nonprofit achieves its objectives"; "the nonprofit is considered important by the community"; and "the nonprofit would continue to exist if specific incomes and/or people were omitted." Respondents ranked these statements on a 5-point scale with 1 representing strongly disagree and 5 strongly agree.
The literature on performance further distinguishes between objective and subjective measures to determine the level of performance. In this article, we Downloaded from https://academic.oup.com/jpart/advance-article/doi/10.1093/jopart/muaa059/6053749 by guest on 11 March 2021 focus on subjective measures. Accordingly, we define performance as perceptions of the effectiveness, legitimacy, and resilience of community-based nonprofits. Using perceived outcomes as a measure of performance is a common strategy in the literature (see Klijn, Edelenbos, and Steijn 2010;Nederhand and Klijn 2019). Furthermore, we combine two different kinds of subjective measures: self-evaluations and external evaluations. Combining these measures may help to overcome the limitations that are associated with each of these measures (see Meier and O'Toole 2013;Wang 2016). Whereas self-assessment measures are prone to personal bias, external-assessment measures lack in-depth knowledge and, thus, may capture only the surface. Here, self-evaluations will be based on the assessment of board members and key volunteers of collectives who have a broad oversight of the community-based collectives' organization and services. External evaluations will be based on the assessment of public officials in the municipality who are familiar with the community and the collectives' services. These different evaluations were combined to construct a composite measure. On average, nonprofits and municipalities rank the performance of nonprofits very similarly. With regard to performance resilience that specifically concerns the internal functioning of the nonprofit, however, only self-evaluation scores of nonprofits are used. See table A1 in the Appendix for more specific calibration details.

Collaborative Activities
Set membership of the condition CA is determined by taking the amount of relationship activities between a specific nonprofit and a governmental municipality into account: no relationship activities = 0.00; one relationship activity = 0.33; two relationship activities = 0.67; three relationship activities = 1.00. Following the definition of cross-sector collaboration by Bryson, Crosby, and Stone (2016), the first relationship activity that was measured is dialogue. Dialogue is necessary for collaboration as dialogue enables the development of a shared understanding and commitment to the process. Hence, it is difficult to imagine effective collaboration without face-to-face dialogue and information exchange (see Ansell and Gash 2008). Dialogue is measured by asking community-based nonprofits about the frequency of contact with public officials (on a weekly basis, monthly basis, half-yearly basis, yearly basis, never). The second relationship activity that was measured is developing joint activities. Following the scholarship on co-production and co-creation, we define joint activities as being involved in a process of co-creating policies and policy objectives (see Howlett, Kekez, and Poocharoen 2017). It is this process, which facilitates and contributes to the alignment of different positions that is an inherent part of collaboration (Henry, Lubell, and McCoy 2011;Huxham and Vangen 2005). Hence, joint activity is rated by asking nonprofits whether they were actively involved in jointly drafting relevant municipal policies (1 = totally disagree, 5 = totally agree). The third relationship activity is measured in the presence of a resource exchange relationship. Exchanging resources is a key element of cross-sector collaboration (Klijn and Koppenjan 2016). In this study, this relationship is measured by determining whether community services of the nonprofit are exchanged for financial government resources as laid down in a formal contract. Despite having a mixed revenue model, these nonprofits are either substantially or primarily funded through financial government resources. The presence of such an exchange relationship is indicated by 1.00, the absence of this relationship by 0.00. See

Community Network Ties
Although all community-based nonprofits provide services to and on behalf of the community and, as a result, have frequent contact with residents, some community-based nonprofits are more locally networked with other community organizations than others. Set membership of the condition CN ties was determined by taking the CN ties of nonprofits into account: no CN ties = 0.00; CN ties based on frequent contact with one actor = 0.33; CN ties based on frequent contact with two actors = 0.67; CN ties based on frequent contact with three actors = 1.00. To determine set membership, nonprofits were asked about the frequency of contact with community grassroot organizations (on a weekly basis, monthly basis, half-yearly basis, yearly basis, never). The final set membership score was determined by translating qualitative frequency scores into set membership scores. See table A4 in the Appendix for more specific calibration details.

Results
The results of the analyses are displayed in table 4 by depicting the solution terms for the performance dimensions: effectiveness, legitimacy, and resilience. The analysis shows three possible routes to perceived performance effectiveness. The first configuration consists of the combination of no collaboration and political network ties (~CA × PN). It suggests that when nonprofits do not collaborate with government, a PN Likewise, the analysis shows that the combination of bridging political and bonding community network ties (PN × CN) is sufficient for perceived performance legitimacy. Finally, the analysis shows, based on our three involved conditions, one route to achieve perceived performance resiliency. The configuration consists of the combination of no collaboration and political network ties (~CA × PN). It suggests that when nonprofits do not collaborate with government, a PN is sufficient for performance resilience.

Evaluating the Three Interpretations
The first interpretation expects that collaborating with government is crucial for perceived outstanding performance of community-based nonprofits. Accordingly, hypothesis 1 states that collaborating with government is a necessary condition for outstanding performance. This hypothesis is not supported. Engaging in CA with government is neither necessary nor sufficient, for achieving outstanding performance. The second interpretation states that collaboration with government should be avoided for perceived outstanding performance of community-based nonprofits. Likewise, hypothesis 2 states that the absence of collaboration is a necessary condition for outstanding performance. We, however, found that the absence of CA is neither necessary nor sufficient, for achieving outstanding performance. Hence, this hypothesis is not supported. The third interpretation considers that the effect of collaboration is contingent on its interplay with the PN and CN of community-based nonprofits. In line with this expectation, hypothesis 3 states-as an extension of the first interpretation-that engaging in CA with government when combined with PN ties or with CN ties is sufficient for perceived outstanding performance of nonprofits. This hypothesis is not supported. Hypothesis 4 states that the absence of CA in combination with PN or CN ties is sufficient for outstanding performance of nonprofits. This hypothesis, which is an extension of the second interpretation, is supported for performance effectiveness and resilience.

Qualitative Mechanisms
This study shows that the third perspective, which highlights the importance of PN and CN ties, is the most insightful explanation for the perceived performance of community-based nonprofits. We will illustrate the results by referring to six concrete cases. Table 5 depicts a key case for each specific solution path.

Conclusion and Discussion
Despite the fundamental theoretical debate on the relationship between government collaboration and the perceived performance of community-based nonprofits, to date, there has been little empirical research that systematically assesses the key assumptions underlying this debate. Responding to calls to blend multiple theoretical perspectives in studying government-nonprofit collaboration, this article demonstrates the potential of combining governance, nonprofit, and social capital literature to capture its complexity (see Bryson, Crosby, and Stone 2016;Cornforth, Hayes, and Vangen 2015;Gazley and Guo 2020). This study contributes to the empirical evaluation of the importance of different components of the government-nonprofit relationship by testing three ideal-typical theoretical interpretations of the relationship: one based on the collaborative governance literature, one based on the nonprofit literature, and one based on the social capital literature. This study demonstrates the pivotal role of the network ties of nonprofits in understanding the relationship between collaboration and perceived performance. There are multiple ways to achieve performance  Providing autonomous care services that prevent people from turning to "checklist-oriented" government services is the core vision of the typical case C3GE. To achieve effective performance, the community-based nonprofit needs government to adjust local welfare policies. The nonprofit is politically very well connected. Based on their political experience, the initiators of the nonprofit believe it is more effective to directly approach elected representatives who control most of the government's resources than indirectly negotiate with public officials. As a result, some public officials feel passed over CA × CN C14UT Typical Some community-based nonprofits don't structurally need government resources to achieve their goals. In this typical case, the main goal of the nonprofit C14UT is to facilitate personal connections within the local community and foster the exchange of personal resources, such as grocery shopping and helping with filling in tax forms. Hence, focusing too much on government contacts (which they describe as tiresome and too concerned with municipal policy interests) would distract from the person-oriented core business of the nonprofit instead of improving it. Investing in community ties is more important to achieving effective performance Facilitating connections within the neighborhood is the main aim of the typical case C12RO. This community-based nonprofit uses its vast community and political networks to contribute to a neighborhood in which residents care for each other. They mainly work on a project basis for government and funds instead of having structural financial relationships. This way of working allows them to be an agile and goal-oriented organization that draws its potential from their local ties C1LA

Inconsistent
The contradictory case C1LA, see table 4, shows that political and community ties are not always sufficient for effective performance. Despite the resources, the nonprofit gets from its network, the usage of their community care services must improve. The nonprofit struggles to find volunteers who can give the organization new élan, a vision for the future. Hence, this case demonstrates the importance of internal organizational capacity The main aim of typical case C6AU is providing care services to keep their village livable. This nonprofit shows that in the process of building legitimacy, community and political connections reinforce each other. Strong local connections make the community-based nonprofit a legitimate partner for politicians to invest in. These investments enable the nonprofit to build their position and services, which in turn further improves the community legitimacy: a selfreinforcing effect. The nonprofit also uses their political network to exert pressure on officials to keep them from strict performance monitoring that endangers their autonomy and unique values To provide resilient care services for people with dementia and their relatives, this typical case C10GR needs structural government subsidies. Although public officials are benevolent, the nonprofit wants to avoid close collaboration as it takes considerable time and energy at the expense of the core activities of providing care. Instead, the nonprofit uses its political connections and experience to exert influence to rearrange pre-existing funding priorities to match its needs.
As a result, public officials, in this case, act cautiously and carefully when it comes to the community-based nonprofit as they are being closely monitored by elected representatives effectiveness: via a community route, via a political route, or via a combination of the two. In the last route, in which both types of network ties are present, it does not matter whether or not the nonprofits collaborate with government. What it does depend on however is the specific type and scope of the nonprofits' goals to determine which route is most appropriate. Both types of ties (political and community) are also important for achieving performance legitimacy. Here, also, it does not matter whether or not the nonprofits collaborate with government. When it concerns performance resilience, there is one route that is sufficient: a political one. In sum, this study offers evidence that collaboration is not necessary or sufficient for perceived outstanding performance. Some routes, namely those leading to performance effectiveness and resiliency, even require nonprofits that are in possession of a political network (PN) or community network (CN) to avoid government collaborations. In these cases, the benefits generated by these collaborations do not off-set the costs of maintaining the collaboration. If the time-consuming nature and the costs associated with collaboration and forfeiting autonomy are not off-set by accessing additional resources, collaboration becomes a liability. It would be very interesting to see if these results also hold in future research endeavors that use other, more objective, ways to evaluate performance. Several limitations apply to this study. The first limitation concerns the generalizability of the findings. As the community-based nonprofits in this study's sample are focused on a specific country, a specific sector (e.g., health and human services), and on a specific timeframe for their operations, future research could replicate the analysis to assess whether the results also hold for nonprofits that operate in different countries, different sectors, and that have, for example, been operating for decades (see Gazley and Guo 2020;Guo and Acar 2005;Young 2000). Hence, generalization of this study's contributions should be treated with care. The second limitation concerns the way this study measures network ties. The measure used here, "contact frequency," does not fully capture the specific empirical nature and quality of the contacts. Is the contact, for instance, mainly digital or does it take place in person? Is the contact positively perceived on both sides? Future studies could take a wider array of communication modes into account to assess the types of bridging and bonding ties and how they constitute social capital (see Provan et al. 2005). The third limitation concerns the relatively low coverage of performance resilience. The low coverage is in line with our expectations as the literature indicates that the performance resilience of community-based nonprofits can only be partly explained by network and collaborative conditions. Conditions such as organizational capacity, leadership, and local policy context are also very important in this respect (see Nederhand et al. 2016). As these conditions were neglected in this study, future research could contribute to gaining a fuller understanding of performance resilience.
Despite the limited scope of this study's empirical data, we believe this article can serve as a stepping stone for further scholarship seeking to uncover the potentials and pitfalls of collaboration and, most importantly, under which conditions collaborations thrive (see Bryson, Crosby, and Stone 2016;Douglas et al. 2020;Hall and Battaglio 2018). By using an innovative set-theoretical approach, this article empirically shows that to fully understand and explain the relationship between collaboration and perceived performance, the type of the nonprofits' goals and their network ties should be considered. These settheoretical findings are a first important contribution to the rapidly growing field of nonprofit collaboration research in understanding and explaining the effectiveness of government-nonprofit collaboration (Cornforth, Hayes, and Vangen 2015;Gazley and Guo 2020;Stone and Sandfort 2009). Furthermore, this study demonstrates the importance of the PN ties of community-based nonprofits in achieving performance legitimacy and resiliency. To date, the role of PN ties in contextualizing the nonprofit-government relationship and its performance has been virtually ignored. This is surprising as dynamics in the political environment can strongly affect nonprofit-government collaborations (Bryson, Crosby, and Stone 2016;Stone and Sandfort 2009). These collaborations often imply decisions about deploying or redeploying significant amounts of resources, which is a strongly politicized process (Nielsen and Baekgaard 2015;Perry and Rainey 1988). This study, therefore, lends support to the view that studying collaboration mainly as a managerial challenge, in isolation from political processes, misses the mark (Huxham and Vangen 2005;O'Toole and Meier 2004). By focusing attention on the political aspect of government-nonprofit collaboration, this article provides important new insights for enriching and deepening our knowledge on collaboration processes.

Funding
This work was supported by the Nederlandse Organisatie voor Wetenschappelijk Onderzoek (grant number 409-14-014) and cofinanced by de Nederlandse School voor Openbaar Bestuur, Deltares, Rebel Group, Resetmanagement, Twynstra Gudde, and Rijkswaterstaat. It has been written as a result of the research project "Governance for smartening public-private partnership."

Data Availability Statement
The data underlying this article are available in the article and its supplementary material.  Note I: Performance dimensions are rated by respondents with 1-5 stars (1 star = bad performance, 5 stars = outstanding performance). When respondents within government or nonprofits scored specific performance dimensions differently, the average score is calculated. Based on in-depth knowledge of the cases the following thresholds were set for set membership of the set performance effectiveness and legitimacy: Score of 2.50 or lower = 0.00; between 2.51-3.99 = 0.33; between 4.00-4:49 = 0.67; score of 4.50 or higher = 1.00. Based on in-depth knowledge on the cases the following thresholds were set for set membership of the set performance resilience: Score of 1.99 or lower = 0.00; between 2.00-2.99 = 0.33; between 3:00-3:99 = 0.67; score of 4.00 or higher = 1.00.
Note II: For the case C14UT the public official could not answer the questions about legitimacy (due to a self-indicated lack of insight), therefore the middle score of 3 is used for this dimension based on qualitative interview data, marked by the * sign.
Note II: Joint activities refer to being involved in a process of co-creating policies and policy objectives rated by answering the following question on a 5-point scale: "We as nonprofit are actively involved in drafting relevant municipal policies" (1 = totally disagree, 5 = totally agree). When respondents within nonprofits or government scored shared decision-making differently, the highest score was used. Based on in-depth knowledge of the cases, the following thresholds were set for shared decision-making: between 1.00-2.00 = 0.00; between 2.01-3.00 = 0.33; between 3.01-4.00 = 0.67; between 4.01-5.00 = 1.00.
Note III: Resource exchange relationship: the presence of contractual financial exchange relationship is indicated with 1.00, the absence of this relationship with 0.00.
Note IV: Fuzzy set score collaborative activities (CA) is determined as follows: no relationship activities = 0.00; one relationship activity = 0.33; two relationship activities = 0.67; three relationship activities = 1.00.
Note II: Fuzzy set score of Political Network Ties (PN) is determined as follows: Membership in both sets = 1; no membership in both sets = 0.00. If a case is member of only one set, compute the frequency scores, if the computed score exceeds the cross-over value of 5 than = 0.33 (C4LE and C7HE); if the computed score falls below 5 than = 0.67 (C12RO).
Note III: For the C10GR case, the score is adjusted one point lower on the basis of qualitative interview data, marked by the * sign.
Note IV: For the C6AU case, the score is adjusted one point lower on the basis of qualitative interview data, marked by the * sign. Note: Frequency score of contact between nonprofits and community grassroot organizations: 1 = weekly contact (fuzzy score = 1); 2 = monthly contact (fuzzy score = 0.67); 3 = once a half year contact (fuzzy score = 0.33); 4 = once a year contact (fuzzy score = 0.00); 5 = never contact (fuzzy score = 0.00). When respondents scored the frequency of contact differently, the highest score was used.

Legitimate Performance PER.L
No necessary conditions were found for Legitimate Performance that meet criteria consistency threshold 0.90, coverage threshold 0.60, Relevance of Necessity (RoN) threshold 0.55.

Resilient Performance PER.R
No necessary conditions were found for Resilient Performance that meet criteria consistency threshold 0.90, coverage threshold 0.60, Relevance of Necessity (RoN) threshold 0.55.