The Effect of Transformational Leadership on Network Performance: A Study of Continuum of Care Homeless Networks

The trend toward using collaborative networks has increased in recent years—creating a need to understand the unique leadership skills and qualities that are necessary of managers to effectively function within this new normal. This article examines the relationship between transformational leadership and network performance in Continuum of Care homeless service networks. We hypothesize that transformational leadership behaviors of network managers contribute to the effective management of a homeless service network. We test this proposal using survey data from 237 respondents who lead federally funded Continuum of Care homeless service networks. Findings indicate that transformational leadership behaviors have a positive and statistically significant effect on the performance of the homeless service networks.

In the United States (US), health and human services policies often use cross sector networks as a form of service implementation with the expectation of positive outcomes generated from effective network performance. The Homeless Emergency Assistance and Rapid Transition to Housing Act (or HEARTH Act) is one example of US federal policy that requires communities to create cross sector service networks. Under this law, services for individuals experiencing homelessness have been designed and offered by a service unit of entities organized into what are known as Continuum of Care (CoC) networks. These networks are locally organized by diverse cross sector service entities; and, members of the networks are expected to engage in collective decision-making, resource development and distribution, and strategies to reduce homelessness.
Although the scholarly literature on the demand for collaborative governance is extensive (e.g., Ansell & Gash, 2008;Emerson, Nabatchi, & Balogh, 2012;Marwell & Calabrese, 2014;Purdy, 2012), there is a dearth of knowledge about how CoC networks operate in practice and the factors associated with effective operation of these cross sector service networks. This study, therefore, explores whether public service managers engage in specific leadership activities and how these activities influence the performance of CoC networks.
We focus on the leadership behaviors of network managers, recognizing that leadership and management are often overlapping, but at the same time different in concepts. Research indicates that not all managers are effective leaders and vice versa (Dukakis & Portz, 2010;Trottier, Van Wart, & Wang, 2008;Van Wart, 2013). Managers are responsible of daily operations of service delivery and the allocation of resources. They are also responsible for oversight of personnel and performance. However, individuals who manage networks effectively and efficiently also have to demonstrate certain leadership capacities (e.g., looking at the big picture, bringing people together to support the vision, and developing innovative strategies for effective goal achievement (Agranoff & McGuire, 2001;McGuire & Silvia, 2009).
While these leadership behaviors may not be visible in daily managerial responsibilities, managers who are effective in achieving the goals of networks tend to be effective leaders (Crosby & Bryson, 2010;Kotter, 1990;McGuire & Silvia, 2009;O'Leary, Choi, & Gerard, 2012;Van Wart, 2005). Indeed, acknowledging differences in managerial tasks and leadership behaviors, studies in network management often present a perspective that leadership qualities are expected managerial capacities in the effective management of public service networks (Agranoff, 2012;Crosby & Bryson, 2008;Linden, 2010;McGuire & Silvia, 2009;Milward & Provan, 2006). For example, the challenges in developing and overseeing the interactions of multiple agencies in a network context require not only key management tasks of managing financial resources, but also effective leadership activities that may offer strategies for motivation, information flows, building interdependencies, accountability, and guidance for better achievement of policy outputs and outcomes (Forrer, Kee, & Boyer, 2014;McGuire & Silva, 2009;Milward & Provan, 2006).
A study by McGuire and Silvia (2009) also tested leadership behaviors and manager's perceptions of network effectiveness in local emergency management networks. They found that there was a significant association between leadership behaviors such as framing, mobilizing, and synthesizing and the perceived network effectiveness of managers.
In the present study, we examine key leadership activities of mangers in homeless service networks. We explore the impact of their leadership behaviors on network performance by building on transformational leadership theory. We focus on the impact of leadership behaviors in explaining network effectiveness since network managers, in representing their home organizations in the collaboration process, engage in leadership behaviors that influence the behaviors of diverse agencies for the benefit of the service network and community.
CoC networks are expected to bring diverse voices together, to challenge members to think outside the box to identify solutions to the needs of those experiencing homelessness in their community, and to be creative in identifying funding and other resources to achieve their collective missions (HEARTH Act, 2009). We theorize that building effective networks requires individuals who care and are passionate about the policy issues at stake and who can coalesce actors from various sectors to join a collective vision for change. Specifically, we argue that transformational leadership behaviors may generate positive relationships among members from multiple organizations by inspiring a collective vision, motivating member efforts to be aligned with network mission, and empowering members to facilitate changes of status quo (Bass & Avolio 2002). Through this study, we hope to expand the study and application of transformational leadership theory, which has been extensively studied in different organizational fields, by applying it to a public service network context.
Since little is known about the degree to which network managers engage in transformational leadership behaviors and whether this style of leadership matters in network performance, we seek to answer the following research questions: 1) What are the key transformational leadership behaviors exercised by network leaders? And, 2) Does a manager's transformational leadership style matter in explaining a network's performance? To answer these questions, we constructed and administered a nationwide survey that captured the leadership behaviors of individuals managing CoC homeless service networks. We used this survey, as well as secondary sources of data, to test the impact of transformational leadership on network effectiveness. In the next section, we present a review of the literature on effective public service networks and leadership. We then develop testable hypotheses; and, the research context, data, and methods are presented followed by the findings. The last section comprises the conclusion, discussion of limitations, and implications for future research and practice.

Effective Public Service Networks
Public service networks refer to structures of organizations working together to co-produce and implement public programs that they would otherwise be unable to accomplish alone (Agranoff & McGuire, 2001;Jang, Valero, Kim, & Cramb, 2015;O'Toole, 1997;Provan & Milward, 1995: Valero, Lee, & Jang, 2020. A public service network, therefore, refers to the horizontal communication and decision-making structures formed by autonomous and interdependent member agencies. Since collaboration and collaborative governance are more of a direction and process, rather than a form or structure of organizational arrangement, an effective network will aim to achieve collaborative governance by pursuing common goals and generating collective outcomes (Ansell & Gash, 2008;Emerson, Nabatchi, & Balogh, 2012;Klijn, 2005;Selden, Sowa, & Sandfort, 2006).
Public service networks are a visible form of organizational structure that engages cross sector organizations (that are participating in the search for comprehensive solutions) with network members having roles, responsibilities, and other mechanisms in place to move ideas forward. The way diverse organizations communicate, make decisions, and implement actions in a collaborative nature are key conditions of an effective public service network.
A central theme explored in previous research on public service networks has focused on why organizations collaborate. However, factors explaining effective networks have been understudied. This is not surprising since it is difficult to observe complex interactions of cross sector actors participating in the multiple stages of the collaboration process. It can also be a daunting task to identify collective goals shared among network members in order to measure network effectiveness.
In seminal work assessing network effectiveness, Provan and Milward (2001) suggested that network effectiveness research can be conducted at three levels of analysis: organization, network, and community. At the organizational level, the focus is on assessing the degree to which organizations are able to accumulate individual benefits as a result of their collaborative participation. For example, are organizations able to better serve their client base as a result of collaborating with other organizations? Other effectiveness criteria at this level of analysis include resource acquisition, agency survival, and enhanced reputation.
At the network level, effectiveness is measured by the degree to which the network as a whole is able to achieve collective benefits. The effectiveness criteria may include increased network membership, range of services provided, member commitment, and integration and/or coordination of services. At the community level, the focus is on investigating whether the network is able to contribute value to the community it serves. The effectiveness criteria at this level of analysis may include reduction in the problem, public perception that problem is being tackled, and cost to the community.
Using Provan and Milward's (2001) framework for evaluating network effectiveness, we find that most of the research in the public and nonprofit management field has focused on organizational level analysis by exploring the conditions and/or factors that may help organizations accumulate individual benefits by participating in collaborative efforts (e.g., Andrews & Entwistle, 2010;Babiak & Thibault, 2009;Chen & Graddy, 2010;Gazley & Brudney, 2007;Provan & Milward, 1995;Selden et al., 2006). Andrews and Entwistle (2010), for example, explored the impact of different types of cross sectoral partnership arrangements on benefits to participating organizations. They found that public-public partnerships were positively associated with effectiveness when compared to the impact of public-private partnerships. Some research has explored network effectiveness at the network level. These studies tend to use subjective measures of effectiveness and adopt a small n case study or qualitative approaches (Chen, 2008;Nolte & Boenigk, 2013). In a case study of family and children services in Los Angeles County, for instance, Chen (2008) analyzed the impact of collaboration processes on perceived network level effectiveness measures such as the quality of working relationships, increasing partner interactions, and goal achievement. In general, the study found that resource sharing and building trust mattered in explaining perceived collaboration outcomes at the network level.
Studies on community level effectiveness are rare. In a case study of three multisectoral workforce development networks, Herranz (2010) measured community level performance by using indicators such as job placement rate and service integration. However, their study did not establish any causal relationships. Instead, they provided an initial exploration of Provan and Milward's (2001) theoretical framework.
In this study, we seek to expand the literature on network level effectiveness by measuring network performance of subjective and objective dimensions we then test for factors that affect the degree of network effectiveness by focusing on transformational leadership exercised by homeless service managers. By using a national survey of CoC network managers, we assess how leadership and other contextual variables are associated with two measures of the dependent variable (i.e., network effectiveness): 1) Perceived network effectiveness and 2) achievement of government funding for CoC networks. We discuss these measures in further detail in the research design section. In the next section, we discuss transformational leadership theory and behaviors that can be expected to impact network effectiveness.

Transformational Leadership in Public Service Networks
The scholarly literature on leadership in networks has grown in recent years. Scholars have noted that today's problems require collective action; and, integrative or collaborative leadership can help cross sector entities overcome collective action dilemmas for the common good (Bono, Shen, & Snyder, 2010;Crosby & Bryson, 2010;Silvia & McGuire, 2009).
Whether leadership makes a difference for effective collaboration, however, is unknown. This is partly because much of the scholarship in this area has focused on organizational leadership and on assessing the conditions that influence organizations to engage in interorganizational collaboration. , for example, calls "for a more nuanced look at the characteristics of the public managers who make collaborative decisions" (p. 669). In the present study, then, we focus on those who lead public service networks by assessing their style of leadership and the potential impact of their leadership on the ability of organizations to work well together in a network.
The extant literature suggests key leadership skills are expected of network service managers. These include nurturing trust, rallying multiple perspectives toward a common mission and objectives, negotiating differences, maintaining commitment of key partners, engaging in mutual learning, and constantly improving deliberative decision-making processes (Agranoff, 2006;Agranoff & McGuire, 2001;Agranoff, 2017;Kickert, Klijn, & Koppenjan, 1997;O'Leary & Bingham, 2008: Van Slyke, 2008. One study reported that homeless service network managers engage in both the day-to-day administration of public funding as well as in leadership behaviors that create incentives for new members to join and resolve conflicts among network members, many of whom often bring their own interests to decision-making processes (Jang, Valero, & Jung, 2016).
Network managers also require the use of leadership behaviors in tasks that are unique to the network development process, such as identifying resources, securing the participation of organizations, and establishing a shared vision and objectives (Ansell & Gash, 2008;Milward & Provan, 2006). In fact, the role of leadership in the service network context is likely more necessary and challenging due to lack of authority given to the network managers to manage partner agencies when no clear accountability measures exist to hold partner member organizations accountable. A report about CoC management suggests that one of the key leadership tasks in public service networks is related to being a positive role model and inspiring other members, while at the same time caring about individual members' interests and assisting them overcome their own challenges (Jang, Valero, & Jung, 2016;Jang, Valero, & Jeong, 2020). Thus, the assumption is that the networks led by managers who conduct these leadership tasks successfully will be more likely to generate positive network outcomes.
Previous work on leadership within single organizational settings has reported that public managers' leadership largely varies from transformational to transactional (Jensen et al., 2016;Sun & Henderson, 2016;Van Wart, 2013;; and, in some cases, transformational leadership is perceived by public employees to be more effective (Trottier, Van Wart, & Wang, 2008). Van Wart (2013) notes, "Effective leaders not only ensure that things get done and that employees are appropriately empowered in the present but also take the organization into the future" (p. 558). Transformational leaders, in particular, help to facilitate change by inspiring a collective vision and motivating employees (Belle & Sanzo, 2014;Bronkhorst, Steijn, & Vermeeren, 2015). This is different from transactional leaders who tend to place heavy emphasis on managing employee affairs through rewards and sanctions (Jensen et al., 2016).
In this study, we adopt dimensions of transformational leadership that were developed by Bass and Avolio (2004). We modify their model to understand the effects of leadership on network effectiveness in homeless services since the original framework was developed for organization level leadership. Bass and Avolio (2004) proposed that individuals can achieve transformational leadership through behaviors organized in four dimensions: idealized influence, inspirational motivation, individualized consideration, and intellectual stimulation.
Idealized influence refers to a leader who is a strong role model and whose behavior is led by strong ethical and moral standards. Inspirational motivation refers to leaders who motivate others by inspiring them to achieve mutual goals and who effectively link individual values and beliefs to the mission of the organization. Individualized consideration refers to leaders who take an interest in the individual needs of others. Transformational leaders foster an environment of innovation and creativity through intellectual stimulation. In this type of environment, leaders and followers are able to exchange ideas, thoughts, and solutions to the ever changing needs of an organization. Followers are also enabled to challenge not only their values and beliefs, but also those of their leaders (and vice versa). Overall, transformational leaders are able to tap into the potential and motivations of others; and, by doing so, they ae able to help followers and/or team members perform above and beyond their own expectations.
Although transformational leadership has been widely studied in the for-profit sector, scholarly work on transformational leadership in the public and nonprofit sectors has lagged behind. Within for-profit organizations, transformational leadership has been linked to innovation (Gumusluoğlu & Ilsev, 2009), organizational performance (Garcia-Morales, Jimenez-Barrionuevo, & Gutierrez-Gutierrez, 2012; Zhu & Akhtar, 2014;Zhu, Newman, Miao, & Hooke, 2013), employee citizenship behaviors (Song, Kang, Shin, & Kim, 2009), employee engagement (Tims, Baker, & Xanthopoulou, 2011), and team performance (Lehmann-Willenbrock, Meinecke, Rowold, & Kauffeld, 2015;Wang & Howell, 2012). There are also studies that link transformational leadership to team building in unitary organizational settings. Lehmann-Willenbrock et al. (2015), for example, studied the interactions between leaders and their teams during regular team meetings in an automotive supply industry. They found that transformational leadership explained functional problem-solving by team members-a relationship that was mediated by the use of solution focused communication by transformational leaders.
In a different study, Wang and Howell (2012) found that transformational leadership was linked to the collective efficacy of teams, which was mediated by group identification. Thus, the process by which transformational leaders affect the effectiveness of teams likely occurs as a result of creating a group identity and engaging in effective communication (e.g., identifying a collective vision and helping members understand their role and purpose in the team).
The application of transformational leadership to the context of public service networks is important since this theory can be used to explain the dynamic interactions among network participants and the ability of leaders to affect real change in the community through transformational leadership behaviors. This is because transformational leaders ultimately help create an environment of shared leadership by building relationships among participants from diverse organizations and developing a common vision for collective benefit (Bass & Avolio, 1994).
A leader in a public service network must maintain high ethical standards and be a strong role model in order for network members to accept the network's vision and goals through his or her idealized influence. The network process requires that leaders be stewards of the collaborative work, inspire others, build consensus, consider the needs of network members and act as good faith mediators, and open to new solutions and change when necessary (Ansell & Gash, 2007;Chrislip & Larson, 1994;Milward & Provan, 2006). Transformational leaders can help increase the number of network members and the commitment of members by communicating a compelling and clear vision that effectively links the interests of each organizational member to the purpose and mission of the network.
Organizations are less likely to participate in and commit to the efforts of the network when there is no salience as to the purpose of the network. In their study of senior managers of US local governments, for example, Wright, Moynihan, and Pandey (2012) explored the relationship between transformational leadership, public service motivation, and mission valence. Mission valence refers to an individual's attraction to the goals and mission of an organization (Caillier, 2014). Ultimately, Wright et al. (2012) found that transformational leadership had an indirect effect on mission valence through its effect on public service motivation and goal clarity. In other words, the process by which transformational leaders were able to increase the attractiveness of an organization's mission was by being clear of goals and building individuals' motivation to engage in public service. Ashikali and Groeneveld (2015) similarly found that transformational leadership had an impact on affective commitment; and, this relationship was mediated by creating an inclusive culture. In the present study, we, therefore, predict that transformational leaders can likewise leverage their ability to motivate and be visionary in order to effectively attract and retain network members. That is, they sell a vision that is worthy of collaboration and inclusive of the needs of all stakeholders involved.
It has been noted that transformational leaders in public and nonprofit organizations engage in innovation; and, they help improve employee performance (Belle & Sanzo, 2014;Caillier, 2014;Dwyer, Bono, Snyder, Nov, & Berson, 2013;Jaskyte, 2011). In her study of human services nonprofit organizations, Jaskyte (2011) considered the impact of transformational leadership on two types of innovation, administrative and technological. The former refers to the implementation of a new administrative procedure or policy, whereas the latter refers to the introduction of a product or service that is new to the organization. The results of this study indicated that transformational leadership was indeed a significant predictor of both types of innovation.
Organizations in network collaboration are also expected to think in innovative ways; after all, one of the purposes of collaborating is to strategize ways to co-produce when a single entity is unable to do it alone (Gray & Gray, 1985;Weber & Khademian, 2008). Within networks, transformational leaders help members engage in innovation by revisiting the repertoire of services offered in the community by network members and by developing ways to reduce service duplication and increase service range. Transformational leaders can also lead network innovation in the process of pursuing resources for network efforts. This may be particularly the case when innovativeness is a criterion of grant awards.
Transformational leaders can leverage their ability to engage in intellectual stimulation and motivation to help members organize fund development strategies that are innovative and cohesive and, ultimately, competitive. Research on employee teams in the for-profit sector, for instance, has found that transformational leadership has an impact on collective efficacy and is mediated by group identification (Wang & Howell, 2012). Thus, we hypothesize that H 1 : A higher level of transformational leadership is associated with an increase in perceived network effectiveness. H 2 : A higher level of transformational leadership is associated with an increase in network funding.

Professional Network Manager
Much has also been written about the role and importance of the professional manager in leading public and nonprofit organizations. In the case of a network, managers must also have certain skills in areas such as organizing, identifying financial and human resources, and solving conflicts between members-among others (Agranoff & McGuire, 2001;Milward & Provan, 2006). Agranoff and McGuire (2001), for example, asked whether a comparable "Planning, Organizing, Staffing, Directing, Coordinating, Reporting, and Budgeting" (POSDCORB) for network management exists. They argue that network management tasks, such as activating, framing, synthesizing, and mobilizing, are important. In this study, we conceptualize the professional network manager in two ways: 1) their years of experience in managing the network, and 2) their level of education.
We predict that the more experience an individual has in managing a network, the more effective s/he will be in leading the network to positive outcomes. This is because over time the individual likely gains valuable knowledge about key resources available in the community to support the collaborative process, understand policy expectations of how to implement network programs, and strengthen relationships with network members. , for example, found that public managers with nonprofit experience, or those in a government with volunteer experience, were more likely to report a higher perceived effectiveness of partnerships with nonprofit organizations. In other words, having prior nonprofit experience allowed these public managers to develop an understanding of how nonprofits function. Thus, they were more likely to understand how to build partnerships with nonprofit groups. We, therefore, hypothesize that the more experience network managers have on the job, the more likely they will perceive the network to be effective and are able to help the network secure financial resources.
H 3 : An increase in the years of experience managing service network will be positively associated with perceived network effectiveness.
H 4 : An increase in the years of experience managing service network will be positively associated with an increase in network funding.
We, likewise, predict that managers with higher levels of education have skills and training that may prove helpful in their management of the day-to-day affairs of the network. In his study of emergency management networks in the US, McGuire (2008) found that managers with postgraduate education and other types of specialized training (e.g., State and Federal Emergency Management Agency (FEMA) training) were statistically related to the level of collaborative activity.
Public managers with higher levels of education are more likely to report engaging in collaborative efforts than those with less education. This can be due to the growing emphasis on network, collaboration, and/or partnerships that are often found in postgraduate programs. These include programs of public affairs (e.g., the Masters of Public Administration (MPA) degree) (DeHoog, 2015). In this study, we anticipate that higher levels of education will be associated with perceived levels of effectiveness and ability to secure network resources. Accordingly, we hypothesize that postgraduate education, as a proxy of professionalization, will result in increased network effectiveness.
H 5 : Individuals with higher levels of education will be more likely to perceive higher levels of network effectiveness than individuals with lower education levels.
H 6 : Individuals with higher levels of education will be more likely to secure network funding than individuals with lower education levels.

Research Context
In this study, we explore the relationship between leadership style and network effectiveness within the context of homeless services. Since 1994, the U.S. Department of Housing and Urban Development (HUD) has encouraged communities to tackle the incidence of homelessness through network collaboration. The assumption in this is that the pooling of local resources and expertise will best serve the needs of each community since issues of homelessness are likely to vary from community-to-community (Homelessness, 2010). This approach was codified into law in 2009 with the adoption of the HEARTH Act.
According to this Act, these networks are responsible for identifying their own system of governance, holding membership meetings, and designing and operating a Homeless Management Information System (HMIS) that tracks homeless services and population (Homeless, 2012;Introductory Guide, 2012). Member agencies engage in collaborative activities, such as yearly counts of individuals experiencing homelessness within their community as well as regular meetings to update one another and seek better approaches to homelessness. Counts of homelessness, for example, take place on a given night and require that the CoC coordinate its efforts with volunteer groups, nonprofits, local government entities, and other partners in order to successfully identify locations where individuals experiencing homelessness congregate and to gather. Homeless networks are expected to be comprised of a variety of cross sector actors, including public entities (e.g., local government, county departments, and education providers), nonprofit organizations (e.g., human services nonprofits and faith-based organizations), and private enterprises (e.g., local businesses and housing providers) (Homeless, 2012; .
To support their community efforts, CoCs are eligible to compete for the limited HUD CoC funding. Those scoring higher and meeting the standards and priorities of HUD are more likely to win these competitive grants. Overall, this context presents an ideal laboratory to explore the role of leadership in public service networks when these networks are selforganized at the local level, which allows for the organic selection and development of network leadership.

Data and Method
This study is based on data collected from HUD, US Census Bureau, and a nationwide online survey titled Effective Leadership in Public Service Collaboration, which we developed and administered in October 2015. In 2014, through the use of HUD's Exchange website (https://www.hudexchange.info), we identified a total of 382 Continuum of Care networks and the collaborative applicant of each respective CoC network. Collaborative applicant is the term used by HUD to refer to the lead agency responsible for developing and submitting a consolidated grant application and overseeing the allocation and administration of HUD funding. The lead agency then is responsible for identifying a network manager within their organization. The network manager will take responsibility for representing their organization within the networks structure and will function as the lead agent of the CoC.
As noted earlier in the literature review, we acknowledge that conceptual differences exist between leaders and managers and that managers may not necessarily exercise leadership behaviors. In this study, therefore, we do not assume that the CoC network managers will indeed engage in transformational leadership. Instead, we seek to understand the extent to which they actually exhibit transformational leadership. We then test whether transformational leadership matters in explaining network performance.
The HUD Exchange website publishes the names of CoCs, the jurisdictions of CoCs, and the contact managers of CoCs. We used this published contact information to send a prenotice email regarding the forthcoming survey to collaborative applicants of CoC networks. In the email, we explained the purpose of the survey. We verified, through this prenotice email, whether the contacts were accurate. From the same HUD Exchange source, we also collected key information such as the total homeless population reported by each network and the amount of yearly HUD funding awarded to each network.
The Effective Leadership in Public Service Collaboration survey that we developed was sent to the collaborative applicant of each of the 382 CoC networks. Respondents were asked a series of questions to assess their perceived level of network performance in diverse areas and the extent to which they engaged in activities associated with transformational leadership. We measured these concepts in multidimensional form by creating indices comprised of multiple survey items. we received completed surveys from 237 networks, representing a 62% response rate.

Dependent Variables
Following the network effectiveness model proposed by Provan and Milward (2001), we measured the network level effectiveness in two ways: 1) Perceived effectiveness of the collaborative network (Data source: Effective Leadership in Public Service Collaboration Survey), and 2) Dollar amount of competitive HUD funding won by the network per capita (Data source: HUD). The use of multiple dependent variables to measure the same concept should strengthen the conclusions that can be drawn from the results about the effectiveness of these homeless networks-particularly when comparing subjective and objective indicators . The goal here is to assess the performance of the network as a collective unit rather than measuring the success of individual members or the impact of the network's effort on the community it serves.
The first dependent variable, perceived effectiveness, is an index measure of respondents' assessment of the extent to which the network as a whole achieved collective benefits in the following areas: • Increasing network membership • Increasing range of services • Reducing duplication of services • Increasing member commitment Respondents rated these areas of network performance using a five point Likert scale ranging from 1) Did not experience success at all to 5) Experienced success to a very great extent. For each respondent, their answers to the four items were summed and then divided by 20 (i.e., the sum of the total possible score for all items in the index) and multiplied by 100 to create an index of network effectiveness (Cronbach's α=0.68).
The second dependent variable is network funding, which is measured using an objective indicator (i.e., dollar amount of HUD funding awarded to each network per capita). This is an appropriate proxy for network performance when a network "must become a viable interorganizational entity if it is to survive" by securing financial resources (Provan & Milward, 2001, p. 417). We consider this to be also an appropriate measure of network level effectiveness when each network is required to submit one grant application to HUD by consolidating the requests of organizations in network.

Independent Variables
The key independent variable in this study is the manager's self-rated leadership style. We rely on Bass and Avolio's (2004) Multifactor Leadership Questionnaire (MLQ, Form 5X-Short) to assess the extent to which network managers exhibit transformational leadership. The MLQ is a standard tool used to measure transformational leadership and there is strong evidence of the validity and reliability of this instrument (Avolio, Bass, & Jung, 1999;Bass & Avolio, 1994;Bass, 1998;Judge & Bono, 2000;Trautmann, Maher, & Motley, 2007;Valero, Jung, & Andrew, 2015).
Although the original MLQ questionnaire includes 45 questions, we used a condensed version of the questionnaire that includes 16 survey items. We focused specifically on items that addressed the four dimensions of transformational leadership: idealized influence, inspirational motivation, intellectual stimulation, and individualized consideration. Consistent with previous research that has relied on condensed versions of questionnaires to assess transformational leadership Valero, Jung, & Andrew, 2015), this 16 item questionnaire allowed us to maintain size of survey at an appropriate level for completion by network managers.
We also modified some of the items since the tool was originally developed for measuring leadership behaviors within an organization and not within a multi-actor service network. For each item, respondents were asked to assess statements using a five point Likert scale ranging from 1) Never to 5) Very often. We created an index by summing the scores for each question and then dividing by 80 (i.e., the sum of the total possible score for all items in the index) and multiplying by 100 (Conbrach's α=0.92). The list of indicators for each dimension are outlined in Table 1.
Professionalization of the network manager is operationalized in two ways, education level and experience in leading the network. We asked each network manager to identify their professional education, and we measured this variable in dichotomous form (1=postgraduate degree, 0=bachelor's degree or less). We assessed education in this way since previous research has shown that postgraduate education has an effect on collaborative activity (McGuire, 2008). We also asked network managers to identify the number of years they had been in the position of 'collaborative applicant,' which is the label that HUD uses to identify the lead organization within the network.

Control Variables
We included a series of control variables important for allowing us to make meaningful comparisons across the homeless service networks in the study. We anticipated differences in the number of homeless populations they served, the characteristics of their networks, and the demographics of the network managers. Thus, we controlled for individual, network, and community attributes in order to test the hypotheses. Women, for instance, may have different interpersonal skills than men; and, as such, they may be more likely to exhibit certain transformational leadership behaviors. Kark, Waismel-Manor, and Shamir (2012) found that a leaders' 'femininity' was strongly associated with effective leadership. We, therefore, measured leader gender in dichotomous form (1=male, 0=female).
The characteristics of a network may also influence the relationship between managerial characteristics and network effectiveness; therefore, we controlled for the size and age of networks (e.g., how long they have been in existence). A large network membership could be an indication of a network that is resourceful, inclusive of the community, and engages diverse stakeholders (LeRoux, Brandenburger, & Pandey, 2010;MacIndoe, 2013). The number of service years that the network has been operating in the community may indicate how solid the network is and capture network capacity. It is, therefore, the expectation that longstanding networks will enjoy a comparative advantage in the competition for resources (Jang, Valero, & Jung, 2016). We measured size of network in terms of the number of organizations that are members of the network and network age as the number of years that the network reports being in existence in their community.
Finally, every community that a network serves will likely vary. Therefore, we also controlled for the homelessness rate 1 (homeless per capita) and the average household income of the population within the network's jurisdiction. 2

Results
The first research question asks: What are the key transformational leadership behaviors exercised by network leaders? To answer this question, we asked homeless service network managers to self-assess the degree to which they engage in transformational leadership activities. Table 1 provides descriptive statistics of the four dimensions of transformational leadership examined in this study. We rank indicators based on mean response for each item.
The results indicate that network managers place a focus on both respecting partner differences and cultivating an environment where sharing ideas and open dialogue is encouraged when intellectual stimulation is the highest rated dimension of transformational leadership (mean of 4.1). In addition, "seeking the counsel of key stakeholders of the network" was the most frequently reported transformational leadership behavior by network managers. This could indicate that network managers focus their efforts on gauging the interests and buy-in from key stakeholders.
We also found that "being open to the ideas and suggestions of network members" was the second most frequently reported transformational leadership behavior. This confirms the expectation that managers make efforts to balance the vision of the collaborative with that of participating organizations.
Results also indicate that network managers have, with frequency, established a fair process • Expressing the need to adhere to ethical standards among members of the network 4.0 7 • Focusing efforts in building future leadership of network 3.3 16

Inspirational Motivation
• Inspiring network members to work cohesively for common purpose 3.9 4.1 4 • Expressing confidence in network members' ability to achieve network vision 3.9 8 • Making an effort to build a network vision to internal and external stakeholders of the network 3.8 10 • Helping each member of the network understand their unique role in network mission 3.6 13 Intellectual Stimulation • Seeking the counsel of key stakeholders of the network 4.1

1
• Being open to the ideas and suggestions of network members 4.3 2 • Helping network members look at issues from different perspectives 4.0 5 • Creating opportunities for network members to engage in creativity and innovation 3.7 11 Individualized Consideration • Providing assistance to network members so that they are able to overcome challenges they encounter 3.7 3.9 9 • Paying special attention to the individual needs and challenges of network members 3.7 12 • Teaching and coaching network members 3.6 14 • Helping assimilate new network members 3.6 15 in managing resources and considering the individual needs of partner organizations. This was the third most frequently reported transformational leadership behavior. This means that network managers understand the importance of maintaining a fair process in managing the limited resources of a network and they also understand that every individual partner is different in their needs. Table 2 and 3 present descriptive statistics of the variables included in the regression models as well as the intercorrelations of dependent, independent, and control variables. The results  Table 3) indicate that a range of weak and strong as well as positive and negative relationships between various independent variables and the measures of network effectiveness. For example, transformational leadership has a strong and positive relationship with perceived effectiveness, which lends initial support to hypothesis 1.
Correlation results also lend initial support for the professional network manager hypothesis, as the results show that having postgraduate education is positively associated with both perceived effectiveness and network funding. While some of the correlations are statistically significant, none exceed 0.50. Thus, issues of multicollinearity are not present (Vigoda, 2000). In addition, tolerance values for all variables are well above the standard threshold; and, the variance inflation factor for all variables is below five.
Next, we estimated the predicted impact of individual, network, and community attributes on network performance using Ordinary Least Squares (OLS) regression. We report standardized coefficients in order to answer the second research question (See Table 4). The overall strength of each model varies, with model 1 exhibiting greater explanatory power than model 2.
In model 1, we consider the impact of leadership behaviors and the professionalization of network managers on the perceived effectiveness of network performance. An R 2 of 0.33 suggests that the individual, network, and community attributes included in our model explain 33% of the variance in perceived network effectiveness. The results confirm that transformational leadership is an important predictor of the ability of networks to achieve collective benefits, such as increasing the number of network members and increasing the commitment of members as perceived by network managers (β=0.50, p<0.01). This suggests that on average, a respondent's perceived level of network effectiveness is predicted to increase by 0.50 standard deviations for every one standard deviation increase in a respondent's level of transformational leadership.
In addition, the findings provide support for the importance of a professional manager leading a network's efforts. Network managers with higher levels of education, specifically a postgraduate education, are more likely to have a greater perceived impact on the performance of a network than those managers without postgraduate education. For instance, a respondent's perceived network effectiveness is predicted to increase by 0.17 standard deviations when having a postgraduate education compared to those respondents with lower levels of education. Other individual characteristics, such as a respondent's years of experience in managing the network and gender, did not have a statistically significant impact on the perceived performance of the network.  With regard to network attributes, the results indicate that while the age of the network may not matter, network size certainly does (β=0.12, p<0.05). Specifically, the findings suggest that perceived effective network performance is positively affected by the size of network membership. In other words, the larger the network (by having more member agencies), the more likely that a manager will perceive that the network is effective in recruiting members, increasing the range of services, and reducing duplication of services. Out of the community attributes, perceived effectiveness of network is associated with average household income of the community that the network serves.
In model 2, in the attempt to understand effectiveness with an objective lens, we assess the impact of leadership behaviors on the dollar amount of HUD funding won by networks. Per capita HUD funding is a useful measure to evaluate how the network secures resources to address the incidence of homelessness within their community. The R 2 of 0.18 suggests that the individual, network, and community attributes included in the model explain 18% of the variance in HUD funding per capita. Unlike the first model where transformational leadership has a statistically significant relationship with perceived network effectiveness, transformational leadership no longer has an impact on a network's performance in terms of securing funding resources in model 2. This suggests that exhibiting transformational leadership does not directly translate into networks increasing their likelihood of winning a competitive HUD grant.
The level of education of the network manager continues to have a positive and statistically significant relationship with the amount of HUD funding per capita (β=0.14, p<0.05). This suggests that leaders with higher levels of education (specifically, in the form of a postgraduate education) better understand government funding opportunities and are more successful in securing these types of grants.
Of the network attributes, age of network has a statistically significant relationship with HUD funding per capita (β=0.21, p<0.01). That is, HUD funding per capita is predicted to increase R 2 by 0.21 standard deviations for every one standard deviation increase in the age of the network. Homelessness rate, as a community attribute, also yields a statistically significant result (β=0.35, p<0.01).

Discussion
First, we sought to expand a leadership theory that has been extensively studied within organizational boundaries by testing its relevance to cross sector service network within the context of public homeless services. The results indicate that network managers perceive that they are exhibiting various levels of transformational leadership-with managers paying more attention to some leadership dimensions than others. For example, network managers perceive that they exhibit greater levels of intellectual stimulation (average mean of 4.1) when compared to the other three dimensions of transformational leadership. One interpretation of this finding is that network managers understand that in order to build an effective team of organizations working together, they must build a culture of exchange and innovation that welcomes and actively seeks a variety of perspectives. This is congruent with previous research showing that transformational leadership influences affective commitment by creating an inclusive culture (Ashikali & Groeneveld, 2015).
We acknowledge that there are important differences between actual and perceived effectiveness. Therefore, in this study we attempted to test the relationship between leadership behaviors and, both, objective and subjective measures of network effectiveness. As predicted, respondents who perceive that they exhibit higher levels of transformational leadership also perceive that their network is effective. Among the explanatory variables in model 1, transformational leadership has the strongest effect on perceived network effectiveness. Thus, there is evidence to support the first hypothesis which states that a higher level of transformational leadership will be associated with an increase in perceived network effectiveness.
This finding confirms the theoretical argument about the relevance of transformational leadership for network effectiveness. This finding also adds to the empirical literature on the impact of transformational leadership on perceptions of effective management (Belle & Sanzo, 2014;Caillier, 2014;Dwyer et al., 2013). As such, it may be beneficial for network managers to engage in the various dimensions of transformational leadership activities (e.g., identifying a vision for the network to pursue collective goals, motivating members that may come from diverse agencies and other stakeholders through inspiration to achieve the various goals of the network, and being open to the ideas and suggestions of network members).
The second hypothesis proposes that transformational leaders will have an impact on the effectiveness of networks by securing HUD funding for their operations and programs. However, the results indicate that the relationship between transformational leadership and HUD funding is not statistically significant. Thus, we do not find support for the second hypothesis.
Third, we analyzed the relationship between the professionalization of network leaders and positive collaboration outcomes. The results indicate that the number of years of experience as a network manager (hypotheses 3 and 4) is not significantly associated with perceived effectiveness and HUD funding per capita. On the other hand, we did find that higher levels of education (hypotheses 5 and 6) yield significant results in both models. This suggests that having a postgraduate education makes a difference in perceived network effectiveness and in securing HUD funding for their network. This finding is consistent with previous work showing that postgraduate education has an impact on the successful network activity of public managers (McGuire, 2008).
Lastly, out of the control variables, we are intrigued by the findings that both network size and age matter in explaining perceived network effectiveness and HUD funding per capita, respectively. For example, the larger the network, the more likely that network managers perceive that their network is effective. One interpretation of this finding is that resources matter in effective service networks-i.e., larger networks are able to leverage the increase in resources that occur when member organizations join the network. With regard to network age, experience is likely a byproduct of network age-i.e., the longer the network is in operation, the more experience the network has in building a relationship with HUD and securing financial resources for the network. Thus, more recently established networks may be at a disadvantage when competing for HUD funding.

Conclusion
The purpose of this study was to explore the relationship between managerial characteristicsnamely, transformational leadership, education, and experience-and their impact on the performance of cross sector networks working to address homelessness within their community. The findings confirm that transformational leadership matters in explaining network effectiveness (as perceived by network managers), which we measured in multidimensional form by incorporating network membership, member commitment, range of services, and duplication of services. In addition, the results indicate that the advanced education of the individual charged with managing the affairs of the network matters. Taken together, these results suggest that public service networks should carefully consider the important leadership activities and qualifications of individuals that are appointed or selected to lead cross sector service networks.
From a theoretical perspective, this study fills a gap in the current literature on public and nonprofit management by providing empirical evidence of the link between transformational leadership and network effectiveness. More specifically, we expand upon a well-studied leadership theory in organizational settings by applying it to interorganizational collaboration. In addition, we provide evidence of the impact that leadership has in network performance in a different policy context-homeless service provision. Previous work has explored the relationship between leadership and network performance in the emergency management context (McGuire & Silvia, 2009).
The results have several implications for practice. First, when establishing a collaborative network and beginning discussions on who should be charged with leading the process, networks should take a close look at the leadership and educational competencies of candidates. For networks that are already established, it may be useful to create opportunities that allow network managers to develop transformational leadership and/or pursue continuing education to acquire the professional skills necessary to lead a service network. Second, the results indicate that the age of the network is a key predictor of the competitiveness of networks in the HUD grant process. Thus, HUD should consider developing supportive programs that allow newly established networks to develop expertise to even the competitive playing field-which ultimately, becomes an issue of equity to help ensure that all communities have the same potential in accessing federal resources and addressing homelessness effectively.
This research is not without limitations. First, we rely on subjective, self-reported measures of transformational leadership. As a result, there is the potential for social desirability bias when network managers over report their transformational leadership behaviors. In addition, the data does not take into account the perspective or perception of network members who may have varying responses about the leadership network managers exhibit. Furthermore, the network effectiveness model proposed here has some key omitted variables; and, governance structure of the network is the one. Therefore, future research should consider the governance structure that networks adopt as a predictor of network performance. Lastly, the work did not consider the impact that leadership has on a network's ability to make a difference in the community-in this case, a reduction in homelessness. Thus, future research should consider the relationship between transformational leadership and positive community level network outcomes.

Notes
1. Homeless population data was collected from HUD Point-in-Time (PIT) Count, a count of "sheltered and unsheltered homeless persons on a single night in January" of each year (HUD Exchange, 2016). We calculated the per capita rate by dividing the homeless population by total population and then multiplied it by 1,000. 2. Household income data was collected from US Census based on the jurisdiction that each network covers.