Coordinating Non-Profit Children’s Behavioral Health Services: Clique Composition and Relationships

Abstract

Background

Why organizational cliques are associated with better performance in service delivery networks has yet to be explained. Certain properties of cliques may account for the improved performance including the composition of clique members and the quality of their relationships.

Purpose

This study offers insight into how organizations working through cliques improve network performance by exploring the complementarity of services provided by clique members and testing two hypotheses about trust and perceived benefits among clique members.

Methodology

Survey and archival data were collected from a regional network of 36 non-profit children’s mental health agencies that belong to a coalition. First, clique analyses and network visualization were used to identify cohesive subgroups. Second, the complementarity of services provided by the groups was explored by calculating scores for each group to reflect the level of differentiation in services and client population as reported in archival data. Third, ANOVA density models were used to test whether clique relationships are characterized by higher perceived trust and benefits compared to non-clique member relationships.

Findings

Three groups were identified. These groups provide complementary services to similar client populations. Trust within all three cliques was higher than non-clique member relationships. Members of all three cliques perceived greater efficiency and two of the three cliques also perceived greater access to care and service quality.

Practice Implications

Results support selecting clique partners based on service mixes. To gain organizational benefits and improve network performance, partners should offer distinct services relative to one other but to similar clients.

Children and youth with behavioral health problems often have service needs that extend beyond the capacity of a single agency (Burns et al., 1995). Comprehensive care typically entails services delivered by multiple organizations. Properly aligning services across multiple organizations is necessary to help families access care, avoid duplication, and prevent treatment conflicts (Burns, Schoenwald, Burchard, Faw, & Santos, 2000).

To facilitate seamless and efficient delivery of comprehensive services, behavioral health organizations are encouraged to partner. Previous system reforms emphasized building dense partnerships among all agencies in a network (Lehman, Postrado, Roth, McNary, & Goldman, 1994; Rosenheck et al., 2002) but these reforms did not result in better system performance. On the other hand, high performance in human service delivery networks (Provan & Sebastian, 1998) and for-profit industries alike (Rowley, Baum, Shipilov, Greve, & Rao, 2004) has been associated with the presence of multiple, overlapping cliques, or clusters of three or more closely partnered agencies with strong, reciprocal relationships (Wasserman & Faust, 1994). These findings suggest that building close partnerships among small groups of agencies (cliques) may be a more effective strategy for coordinating service systems than building dense networks.

In a seminal study highlighting the link between cliques and human service network performance, Provan and Sebastian (1998) speculated that clique composition along with the quality and stability of their relationships are keys to network performance, especially clique composition. For example, the complementarity of partners’ service arrays almost certainly influences clients’ access to comprehensive services (Selden, Sowa, & Sandfort, 2006), thereby giving clique composition a fundamental role in setting up clique dynamics and explaining performance. Furthermore, the quality of the relationship, especially those based on trust and mutual benefits influence the relationships’ strength, stability and effectiveness (Chen & Graddy, 2010; Jones, Hesterly, & Borgatti, 1997).

In this study of a network of children’s behavioral health organizations, we build on Provan and Sebastian’s (1998) work by examining clique composition and the quality of clique members’ relationships. First, we explore the complementarity of services offered by clique members. Second, we test two hypotheses related to the level of trust and the perceived partnership benefits among clique members compared to relationships among non-clique members in the larger network. Understanding composition and the nature of clique-member relationships is a step toward understanding the link between subgroups and network performance, and has implications for managers selecting partners and strengthening their agencies’ relationships.

The Focus on Relationships

Cohesive subgroups are clusters of tightly connected agencies, and cliques are the most tightly connected subgroups (Wasserman & Faust, 1994). Cliques are comprised of at least three agencies connected by mutual, strong ties. Cliques are more connected to one another than they are to the rest of the network.

Organizational cliques may offer an effective way of coordinating human service delivery networks. Demonstration projects and network studies conducted with mental health and human service agencies in the 1980s and 1990s examined the impact of ties among agencies on network effectiveness for addressing mental health needs. Ties in human service delivery networks typically involve sharing or exchanging resources such as client referrals, or administrative resources like funding, staff expertise, or space (Bolland & Wilson, 1994). As agencies share resources with one another, they align their service delivery and administrative functions and become more closely coordinated. Results from these studies suggested that increasing the number of ties (density), or concentrating ties around a single agency (centralization) made no difference in the way the networks performed (Lambert & Guthrie, 1996; Lehman et al., 1994; Rosenheck et al., 2002). Instead, networks with small cohesive organizational subgroups based on multiple shared resources have been linked to better outcomes (Provan & Sebastian, 1998).

Actors (individuals or agencies) within networks have a natural tendency to cluster into smaller groups because actors are constrained in the number of partnerships they can develop, strengthen, and maintain (Roberts, Dunbar, Pollet, & Kuppens, 2009). Therefore, it is no surprise that organizational cliques based on client referrals or shared administrative resources are common in human service delivery networks (Provan, Milward, & Isett, 2002). However, cliques based on both administrative resources and referrals (multiplex ties) are rare (Provan & Sebastian, 1998). These overlapping clique members share and receive both client referrals and administrative resources with each of the other members, representing a reciprocal pattern of interactions. It is these types of overlapping cliques that have been associated with better performance than networks with dense inter-organizational ties based on one type of resource, such as client referrals (Lemieux-Charles, et al, 2005; Provan, Milward & Issett, 2002).

Provan and Sebastian (1998) note several clique properties that may account for improved performance. First, the composition (service mix) offered by clique members may come closer to an optimal set of well-aligned services for clients. Second, the relationships among clique members are especially strong and therefore may be based on positive social qualities including trust and mutual benefit that are likely to contribute to sustainable relationships and overall positive outcomes for the network. However, little is known empirically about either the composition of cliques, or the stability and social qualities of their relationships. Establishing this foundational knowledge about clique members and their relationships is an essential step toward understanding their role in network performance. Next, we draw on theory and findings from the inter-organizational literature to clarify why clique composition, trust and perceived benefits may be important clique features to consider for network performance.

Composition Matters - Complementarity

Provan & Sebastian (1998) note that the complementarity of services offered by overlapping clique members matters in network performance. Organizations search for partners that serve dissimilar or complementary functions relative to one another (Li & Rowley, 2002). Specifically in human services, when partnering agencies provide functionally distinct services, they expand the range of services available to a common population of clients, facilitating access to more services, and theoretically improving outcomes (Selden, Sowa, & Sandfort, 2006). For example, a clique comprised of a specialized psychiatric treatment agency, substance abuse center, and crisis intervention provider would be expected to serve youth with co-occurring mental health and substance abuse treatment needs better than a group of three agencies that all provide similar psychiatric care. By that logic, if clique members are too similar and provide duplicative services to the same client population, their partnership cannot meaningfully expand the diversity of services available to clients.

Human service agencies tend to form cliques with agencies that serve similar client populations (Bolland & Wilson, 1994; Rivard & Morrisey, 2003). However, the extent to which these cliques unite functionally distinct services to a shared client population has remained unexamined to date. As a step toward understanding whether clique composition could contribute to network performance, our first objective is to explore the complementarity of services provided by clique members.

Relational Quality – Trust and Perceived Benefits

For cliques to serve as effective building blocks of coordinated networks, their relationships must remain stable over time. Stability depends largely on social qualities of relationships, especially trust, and the financial and service-related benefits of partnership (Granovetter, 1985; Provan, Nakama, Veazie, Teufel-Shone, & Huddleston, 2003). Therefore, high levels of perceived trust and benefits among clique members are expected to contribute to strong, stable and effective partnerships that enhance the whole network’s performance. We examine how trust and perceived benefits contribute to stability and may be especially high among clique members.

Trust

Trust is the expectation that partners will behave reliably, fairly and with good will (Rousseau, Sitkin, Burt, & Camerer, 1998). Trust reduces risk and uncertainties associated with partnerships (Gulati & Nickerson, 2008; Jones, Hesterly, & Borgatti, 1997). Partners build and establish norms of trust as they interact with one another, and assess their partners’ behavior (Uzzi, 1997). Agencies are likely to discontinue partnerships with agencies that do not fulfill their commitments or behave opportunistically. When partners provide evidence of their trustworthiness, agencies are likely to maintain these relationships (Rivard & Morrisey, 2003).

Trust is also important for building strong partnerships based on both referral and administrative ties. When developing administrative relationships, agencies are likely to select partners with whom they already have established trust via previous referral ties (Foster-Fishman, Salem, Allen, & Fahrbach, 2001). For example, Impink (2004) describes how a network of domestic violence agencies began by sharing information and referrals. As these agencies learned more about their partners’ trustworthiness, they developed stronger partnerships by sharing space to co-locate services.

We anticipate that opportunities to learn about a partners’ trustworthiness are greater in the context of an overlapping clique. Since cliques are comprised of more than two members, there is both direct and collective monitoring of partners’ behavior, shaping their reputation and accountability to one another (Raub & Weesie, 1990). Also, clique members connected through multiple relationships can observe their partners across different collaborative activities to evaluate their trustworthiness. Therefore, our first hypothesis is: Trust between clique members will be higher than trust among partner agencies that are not involved in cliques.

Perceived Benefits

Partnerships have potential to yield benefits for both the organizations and the clients they serve (Provan & Kenis, 2008). For example, dividing responsibility for providing comprehensive care to a client may enhance an individual agency’s efficiency (Reitan, 1998). Clients benefit when a partnership offers access to a needed array of services (Selden et al., 2006), or improves service quality (Zapka et al., 1992). As a result, agencies often partner as a strategy for improving service delivery as well as their own operations.

Beneficial relationships contribute to strong and stable networks. When agency directors perceive partnership benefits, they are likely to strengthen and sustain their relationships (Chen & Graddy, 2010; Provan, Lamb, & Doyle, 2004) while relationships that yield too few benefits are dissolved (Ring & Van De Ven, 1994). Clique members are joined by multiple types of reciprocal relationships - the strongest relationships in the network. Based on the strength of these relationships, it is expected that they yield benefits, else they would have been dissolved. Therefore, our second hypothesis is: Partnership benefits between clique members will be greater than partnership benefits between partners not involved in cliques.

Method

Network Setting and Population

The context of this study is an urban county coalition of 36 non-profit agencies serving children and adolescents with behavioral health problems. The network boundary was defined as paid coalition membership. The coalition formed in 2007 to coordinate advocacy efforts toward expanding local funding for children’s behavioral health services. Members must be registered as a 501(c) 3 organization and provide direct children’s mental health services. The coalition defined services broadly to include crisis intervention, school and home-based prevention programs, temporary shelter, outpatient psychiatric and substance abuse treatment, individual, group and family counseling, services for pregnant teens, and respite care. Organizations range in size with between three and 730 employees and brought in $35 thousand to $39 million in annual revenue. The majority are small with 61% reporting under $5 million in annual revenue and 59% employing fewer than 200 staff.

Of the 36 members, 89% (n=32) participated. Of the four non-participating agencies, one declined due to time constraints and three provided no reason. All four provided specialized services and may not generate many reciprocal resource exchanges, and thus may not be members of overlapping cliques, although this cannot be determined.

Data Collection and Measures

Data on annual expenditures, revenue, total employees and major service programs for each organization were drawn from 2009 IRS 990 forms, the most recently available reporting year on Guidestar.com. Data about the complementarity of services delivered by each clique was constructed using agency-provided descriptions of the three largest service programs (Part III, Question 4). For each program we recorded the named service type and client population to construct service and client differentiation scores for each clique exactly as they were described in the 990 (Table 1). A service differentiation score reflects the percentage of all programs reported by clique members that provide a unique service type (e.g. mental health counseling, crisis intervention). A client differentiation score is a percentage of all programs reported by clique members that serve unique client populations (e.g. young children, general population). Higher percentages suggest greater differentiation. Since complementarity entails providing functionally distinct services to a similar client population, cliques with high service and low client differentiation scores were considered complementary.

Table 1.

Characteristics of Member Organizations in Each Small Group (extracted from IRS 990s)

#	Org	Revenue	Staff	Service Programs	Clients
1	A	$11,832,406	290	1. Residential treatment	Adolescents
				2. Special education
				3. Independent & transitional living
	B	$7,158,333	192	4. Foster care case management,	Youth
				5. Consultation & family treatment
				6 . Family preservation services
	C	$12,996,617	122	7. Head Start	Young children
				8. Community Youth Dev.	Adolescents
				9. Emergency Shelter
	# Unique:			9	3
	Differentiation:			100%	33%
2	I	$30,391,163	730	1. Day & special education	Youth
				2. Residential treatment
				3. Therapeutic foster care
	J	$2,705,544	60	4. Residential treatment	Young children
	K	$7,141,031	200	5. Transitional Housing	Adolescents
				6. Residential services
				7. Family reunification services	Families
	# Unique:			5	4
	Differentiation:			71%	57%
3	D	$4,021,182	155	1. Mental health/psychiatric counseling, fatherhood teaching, advocacy	General population
				2. Interpretation/translation services	Immigrants & refugees
	E	$3,761,281	50	3. Mental health counseling	General population
				4. Child abuse prevention & detection
				5. Home-maker/In-home assistance	Frail/elderly adults
	F	$7,348,697	247	6. Early Child Education	Young children
				7. Child welfare, adoption & pregnancy counseling	Pregnant Women
				8. Mental health counseling	General population
	G	$39,005,361	425	9. Medical care	General population
				10. Mental health care
	H	$5,625,755	100	11. Mental health & crisis counseling	General population
				12. Community services
				13. Employee counseling
	# Unique:			9	5
	Differentiation:			69%	38%

Data on referral and administrative partnerships, trustworthiness, and perceived benefits were collected via an online survey of coalition agency executive directors from October 2009 to March 2010. Respondents could also complete the survey in hard-copy (9%) or via phone (6%).

The survey used a roster format with eight items related to partnerships, trust, and benefits. Agency directors were given a roster of the other organizations in the network and asked to respond to survey items for each of the other 35 organizations in the network (Wasserman & Faust, 1994). In other words, directors responded to each survey item 35 times. To minimize respondent burden associated with the roster format, multi-item scales were trimmed and only the most relevant items were included in the final instrument, as detailed below.

Referral and administrative partnerships among the 36 agencies were measured using four items from Van de Ven and Ferry’s (1980) Resource Flows scale. Referral partnerships were measured with one item that assessed on an 11-point scale from zero (none) to 10 (100%) the amount of client referrals given within the past six months. Administrative partnerships were measured as the sum of three items on the same 11-point scale assessing the amount of organizational resources (money, staff, and physical space) in the past six months.

Trust was measured using one item from Zaheer, McEvily and Perrone’s (1998) inter-organizational trust scale. Directors rated the trustworthiness of each agency on a scale from zero (not trustworthy at all) to ten (completely trustworthy). The original scale included five items about negotiating contracts, keeping promises, and opportunism. These items were not relevant to informal relationships or new partnerships that have yet to yield opportunities to assess partners’ follow-through, or opportunistic behavior. Because of concerns about relevancy and respondent burden, only the single item related to perceived trustworthiness was retained.

Partnership benefits were measured using three items developed for this study. Each item asked agency directors to rate the extent to which their relationship with each organization in the network benefitted three dimensions of service delivery: 1) enhancing efficiency, 2) client access to services, and 3) quality of care. Directors responded to each item on an 11-point bipolar scale. For example, directors rated efficiency benefits from each relationship on a scale from negative five (relationship is wasteful for our organization) to positive five (relationship is efficient for our organization). The scale anchors for service access benefits ranged from “prevents clients from accessing services” to “facilitates client access to services.” The scale for quality benefits ranged from “detracts from quality care” to “enhances quality care.”

Analysis

Subgroup Identification

Survey responses were used to construct a 36×36 valued matrix for each item, with 1260 dyadic values (not including the diagonal values). To identify cliques, clique-analyses in UCINET 6.0 (Borgatti, Everett, & Freeman, 2002) were conducted for both referral and administrative networks. These analyses identify subgroups of three or more agencies that are maximally connected by reciprocal ties (Wasserman & Faust, 1994).

Next, we adopted Provan and Sebastian’s (1998) definitions and procedures [also replicated by Lemieux-Charles et.al (2005)] to identify “identical cliques” or overlapping sub-groups that meet the criteria of both referral and administrative cliques. We compared the lists of referral and administrative cliques to identify these groups.

Referral and administrative resource networks were visualized together using NetDraw 2.097 in UCINET to confirm identical cliques and identify other tightly connected multiplex groups that did not meet the stringent criteria as a clique. The identical cliques and one strongly connected subgroup identified in these procedures are the focus of subsequent analyses.

Exploring Clique Composition

To examine the complementarity of overlapping clique members, we describe the service and client differentiation scores for each clique.

Examining Clique Relationships

To test hypotheses about trust and perceived benefits within cliques compared to partners that are not clique members, ANOVA density models were estimated in UCINET 6.0. These models use all network data by comparing dyadic relationships within cliques to relationships between non-clique members in the network. Similar to traditional ANOVA tests, group averages are calculated and compared, except this procedure is a permutation test that adjusts for the non-independence of network data; it yields standard errors and p-values based on simulations of the data.

Trust and the three types of perceived benefits were first averaged for the relationships within each clique, among non-clique members, and then between organizations from each group. Next a regression model was run to test whether identical cliques have higher within-group values than between groups. The variable homophily model was fit to the data which allows the values of within-group relationships to vary (Hanneman & Riddle, 2005).

Results

Clique Analysis

Clique analyses and visualizations identified three overlapping, cohesive subgroups. The initial clique analysis identified 19 referral and 14 administrative cliques involving 25 of the 36 coalition members. More organizations were involved in referral (n=21) than administrative cliques (n=15), and 12 (46%) of the agencies involved in a referral clique are also administrative clique members. There was limited identical clique overlap: only two groups (each with three agencies) were both referral and administrative resource cliques.

Using NetDraw, referral and administrative ties were drawn among the 25 organizations belonging to a clique (Figure 1). One sub-group with a high concentration of multiplex ties was identified. This group of five is maximally connected where each organization has a multiplex tie with each of the other four partners. It was not identified in the clique analysis because it was just shy of meeting criteria of reciprocal exchange of both referral and administrative resources. Anecdotal information indicated this group has a 20-year working history and is regionally prominent. Given its influence, this group was included in the analysis.

Figure 1.

Subgroup Identification (n=25)

Composition - Complementarity

The three groups identified in the analysis include eleven organizations that provide 29 programs (Table 2). The data support our first proposition that cliques are comprised of complementary service organizations. All three cliques had relatively high service differentiation scores (69–100%), and lower client differentiation scores (33–57%) suggesting that these cliques are comprised of agencies that provide functionally distinct services to similar client populations.

Table 2.

Model Fit Results and Average Scores (n=1260)

	Model Fit		Averages
	R2	p-value	Whole Network	Within Groups				Between Groups
				Non- Group Members	Group 1	Group 2	Group 3
Trust	.048	.038	5.958	5.881	7.333	6.000	7.650	5.706
Perceived Benefits
Access to Care	.106	.001	8.143	7.56	9.67	7.50	9.20	8.030
Quality of Care	.102	.000	8.074	7.64	9.17	7.67	8.80	7.992
Organizational Efficiency	.116	.000	7.578	7.36	8.20	8.00	8.95	7.395

Bold print denotes model significance at p<.05 level.

Of the three cliques, Clique 1 has the best complementarity. These three agencies have 100% service and 33% client differentiation, meaning that collectively they have no service overlap but serve similar client populations (adolescents specifically, and at-risk youth in general). Clique 3 also serves similar clients (client differentiation = 38%) and offers a diverse set of services (e.g. translation, pregnancy counseling, and in-home assistance), but to a lesser degree than Clique 1. All five Clique 3 agencies provide mental health counseling, creating some duplication that lowers service differentiation (69%). Clique 2 has three agencies that offer a diverse set of services (service differentiation = 71%), although all three offer residential treatment. The services are targeted to multiple client populations yielding higher client differentiation (57%). Clique 2 meets complementarity to a lesser degree than Cliques 1 and 3.

Relational Qualities

ANOVA density model results are addressed below and in Table 2. For each test, the global average value is reported for all network relationships, four within group averages (the three cliques, and non-clique members), and the between group average.

Hypothesis 1: Higher Levels of Perceived Trust among Clique Members

Trust within the cliques is higher than trust between organizations across cliques, or that are not part of the cliques. Cliques 1 and 3 reported higher trust (mean=7.3 and 7.7, respectively) than Group 2 (mean = 6.0), non-clique members (mean = 5.9), and between groups (mean = 5.7).

Hypothesis 2: Higher Levels of Perceived Partnership Benefits among Clique Members

All three partnership benefits (access to care, quality of care, organizational efficiency) varied significantly by group membership. Members of Cliques 1 and 3 perceived greater benefits (than the rest of the network) in access to care (mean = 9.7 and 9.2, respectively), care quality (mean = 9.2 and 8.8), and organizational efficiency (mean = 8.2 and 9.0). Clique 2 was an exception: partnerships were perceived to benefit efficiency (mean = 8.0), but were perceived as less beneficial for facilitating access (mean = 7.5) and improving care quality (mean = 7.7) than other relationships. The differences by clique membership accounts for 11% of variance in access to care, 10% of variance in quality care and 12% of variance in organizational efficiency.

Discussion

This study builds on prior evidence suggesting that organizing systems around cliques and cohesive subgroups may be an effective way to coordinate human service delivery networks. We gained insight into why organizations working through cliques contribute to network performance by exploring clique composition and the quality of their relationships. We found that organizations in cliques offering differentiated service arrays to similar clients trusted and benefited from one another more than other organizations in the network. These findings suggest that cliques may contribute to network performance by uniting complementary services, and establishing sustainable relationships based on trust and mutual benefit.

In this network study of 36 children’s behavioral health organizations, we identified two cliques of three organizations each and one cohesive group of five organizations. Compared to other organizations in the network, these three groups provide complementary services to similar client populations. Trust within two cliques was higher than the trust between organizations across cliques or the trust between organizations not in a clique, supporting our second hypothesis. In addition, the members of all three cliques perceived higher levels of organizational efficiency, and two of the three cliques also perceived greater access to care and quality of service lending partial support to our final hypothesis.

The clique comprised of residential service providers (Clique 2) reported efficiency benefits from partnering, but less trust and benefits related to care access and quality. The services offered within this clique were less complementary than the other two cliques because all three provided residential care. The nature of these services may make close coordination less essential for direct client care. For example, residential care is typically reserved for youth with the most severe behavioral health issues and provides comprehensive services onsite, reducing the need for agencies to facilitate clients’ access to other organizations’ services. In addition, residential care is an expensive service and the members of Clique 2 may be competing with one another for new clients and funding. Our findings are a first step toward understanding the link between organizational subgroups and network performance. When each member within a small organizational group offers a distinct service to similar clients they expand the range of services available, reduce duplication, and are able to coordinate treatment approaches (Selden et al., 2006). Each of these characteristics contributes to improved service delivery (Chen & Graddy, 2010). The cliques in this network shared and received both client referrals and administrative resources with each other. This reciprocal pattern of interaction not only facilitates the alignment of services or operations but also allows continual review and adjustment of the partnership (Doz, 1996). We believe that this type of feedback driven process enhances service delivery.

The higher levels of trust found among clique members helps to stabilize partnerships. Trust builds as partners interact and reflect on each other’s behavior. The cliques in this study had multiple relationships making it possible to observe their partners under varying conditions and thus expand the foundation of trust. In addition to the individual partner evaluations of trust, cliques also offer collective monitoring of reputation and accountability to one another (Raub & Weesie, 1990). The trust underlying clique relationships enables emergence of norms and routines for collaboration (Jones et al., 1997) which may contribute to improved service delivery.

Finally, the benefits of clique membership may also contribute to overall performance and sustainability of networks. Relationships perceived as beneficial are likely to be sustained, while those that fail to yield expected benefits are not (Provan, Lamb & Doyle, 2004). The eleven organizations operating within the three subgroups reported greater efficiency than the other organizations in the network. In addition, two of the three groups also reported better access to care and service quality. The services were more alike in the group reporting lower access to care and care quality, suggesting that differentiated services play an important role in the benefits of clique membership. Differentiation is typical of larger organizations with staff and revenue to support the delivery of multiple services, therefore size may also play a role in the perceived benefits of clique membership. The 11 most connected organizations in this network tended to be large compared to the rest of the network with all but two generating more than $5 million annually.

Implications for Management

Our findings have implications for how managers develop partnerships, specifically the types of partners they select. Participating in cliques or cohesive subgroups may benefit the organizations, their clients, or both. We believe the key to these benefits stems primarily from the composition of services offered. Differentiated service arrays minimize direct competition between organizations and maximize opportunities for cooperation by reducing service duplication and enabling coordinated treatment (Baum & Singh, 1994). In retrospect, it is not surprising that the clique with the most differentiated array of services also had the highest levels of trust, efficiency, access to care, and quality of services. The more differentiated and comprehensive the service array, the more likely the members will benefit each other, and these benefits facilitate the development of trust. Thus, to gain organizational benefits and improve network performance, partners should offer distinct services relative to one other but to similar clients.

The network methods used to identify subgroups can also be used by managers to help build local networks of purposely aligned cliques. Network methods accommodate the full range of resources and services available within networks. This can facilitate potentially effective matches between organizations that work with similar client populations.

Limitations and Future Directions

This study’s findings and implications should be interpreted in light of its limitations. First, only three subgroups were identified in our analysis limiting our ability to examine how complementarity, trust, and perceived benefits vary based on clique size, strategic goals, and other clique features.

Second, the complementarity measures may lack precision due to variability and subjectivity in the way agencies report on their programs with the 990 forms. Agencies may have used different terms to describe their services or their populations, hampering our ability to determine differentiation. Furthermore, differentiation scores are based on the total number of programs reported by clique members, therefore the percentages may be affected by group size; it seems likely that small subgroups will have a smaller denominator than larger subgroups. Third, measures of benefits, trust, and relationships are based on scaled-items. Directors’ may have interpreted the intervals between each scale response differently. Although subgroups reported statistically higher trust and benefits than non-subgroup members, the group averages were only one to two points higher and such differences may have marginal practical significance.

Overall, this study was a preliminary step toward understanding the features of cliques that explain network performance. The findings highlight the importance of selecting complementary partners, and raise new questions about the optimal number of cliques in network, and how to implement purposely designed cliques. While forming small groups of differentiated service organizations may be necessary for coordinating comprehensive services, it is unlikely to be sufficient. Networks are complex structures that exist within and respond to a rapidly changing environment. For example, the shifting constellation of client needs can quickly alter the balance of a once mutually beneficial relationship toward one where agencies differentially benefit, creating unequal power differentials. Understanding the dynamics of cliques, and how they contribute to network performance requires research that moves beyond cross-sectional examinations of a single network to longitudinal studies of multiple networks. The hypotheses supported in this study provide justification for encouraging clique formations to structure human services networks.