How Does Policy Funding Context Matter to Networks? Resource Dependence, Advocacy Mobilization, and Network Structures

How Does Policy Funding Context Matter to Networks? Resource Dependence, Advocacy Mobilization,... Abstract This study explores how policy funding context—defined as whether funding for a social service policy domain is discretionary or mandated—affects network structures in social service domains. We present comparative findings from two social service policy networks which differ with respect to funding context: A 47-actor adult basic education policy network that is funded discretionarily and a 40-actor mental health policy network where spending is mandated. Both are located in a US state we pseudonymed “Newstatia.” Using an exponential random graph model, we found that policy funding contexts affect how the locus of resource dependence interacts with the nature of client groups to determine the array of interest organizations engaged in the networks, which leads to differentials in network structure across these domains. We suggest that policy funding contexts are before resource dependence and client factors when explaining network structure. This opens space for reconsideration of the causal claims between policy funding contexts, resource dependence, advocacy mobilization, and network structures. Introduction Network theory and social network analysis have become primary tools used to analyze the multisector, multi-entity processes that lead to policy decisions, both grand and small. These processes go by different monikers, including policy networks (Laumann and Knoke 1987; Kenis and Raab 2007), governance networks (Klijn and Skelcher 2007; Rhodes 1997; Sørensen and Torfing 2005), collaborative governance regimes (Ansell and Gash 2008; Emerson, Nabatchi, and Balogh 2012), cross-sector collaborations (Bryson, Crosby, and Stone 2006), and intergovernmental cooperative arrangements (Agranoff and McGuire 2004). These studies have evolved into three primary streams: policy networks, collaborative networks, and governance networks (Isett et al. 2011). Policy networks focus on policy decision-making processes more than policy implementation processes, whereas collaborative networks emphasize policy implementation more than policy decision making. Work on governance networks considers both decision making and policy implementation jointly (Rethemeyer and Hatmaker 2008). Despite these conceptual differences, all three share a common assumption: policy stakeholders see decision making and implementation processes as a web of strong and weak connections between and among primary organizations with an interest in policy outcomes (Knoke 1990). The iterative exchange of material and social resources among primary organizations generates self-organizing, complex network structures for functional coordination, knowledge sharing, and/or advocacy mobilization. Understanding these interorganizational network structures provides insights into (1) where power resides in networks and how the distribution of power shapes the perceptions, motives, and actions of organizations involved (Knoke 1990) and (2) why and how certain policy actions or agenda items are (or are not) collectively debated, negotiated, designed, and implemented (Bodin and Crona 2009; Ingold 2011; Ingold and Leifeld 2016). However, uncovering latent deep structures (i.e., identifying factors that govern how networks are formed and take shape) is often ignored or obscured by our customary focus on the interactions that facilitate the work of a network but which are themselves governed by broader contextual factors (Knoke 1990). Little is known about how networks become structures for purposes of advocacy while interacting with broader policy funding contexts. Interactions and exchanges are the actions that define a network and bring it into being. But what helps to shape the pattern of interaction—that is, the structure of the network? Much of the literature uses resource dependence theory (RDT) (Pfeffer 1987; Pfeffer and Salancik 1978) to explain the structure of networks (Park and Rethemeyer 2014). For example, organizations participate in networks to access other organizations’ material and social resources, such as funding, regulatory assistance, knowledge, obligations, or legitimacy (Huang and Provan 2007; Oliver 1990, 1991). Constituent organizations in a network build structures through these multilateral and complex interdependencies (Rethemeyer and Hatmaker 2008). In particular, many empirical analyses have focused on the structure of networks in social service policy domains due to the increasing privatization of local, state, and federal provision of public goods and services (Agranoff and McGuire 2001; Graddy and Chen 2006; Huang and Provan 2007; Johnston and Romzek 2008; Knoke 1990; Milward and Provan 2000; Provan and Milward 1995). This work has used RDT as a primary theoretical framework to explain why organizations relate and interact and how “social structural resources” that arise from network configurations are deployed by actors in these networks. However, previous studies have not paid enough attention to variations in network structure that stems, we believe, from differing policy funding contexts. The current conceptual and empirical research on policy networks has not provided a sufficient basis for understanding the interplay between funding environments and network structures. We examine whether the nature of the funding stream itself biases the structure of policy networks. That is, are there systematic differences between service domains that are funded discretionarily by yearly legislative action and those where expenditures are mandated by law and thus subject to legal challenge and executive agency processes as well as legislative allocations. The two funding contexts that we focus on—discretionary versus mandated—are the primary divide in how public resources in social service domains are distributed in the United States. By shifting our attention to the policy funding contexts that define resource flows for entire policy networks and away from the resources available for the survival of individual policy actors in the network, we aim to draw inferences about policy network structures generically using social network analysis. Specifically, we claim that policy funding context affects (1) how the resource allocation mechanism interacts with (2) client factors to determine (3) interest groups’ advocacy involvement and thus (4) the choice of “advocacy technologies” network participants’ use, which then leads to differences in network structures across social service policy domains. This study will improve our understanding of the structure of policy networks by identifying how policy funding contexts alter the locus of resource dependence and shape political mobilization. This study proceeds in four sections. The first discusses why social service funding contexts help to determine resource dependence and mobilization processes within a network. The second reviews our cases and research methods. The third section presents findings and propositions from our comparative studies of (1) a mental health policy network and (2) an adult basic education policy network collected in 2001 and 2005, respectively, in a state we have pseudonymed “Newstatia.” Focusing on differences in the policy funding contexts in which these two policy networks are embedded, we present a mechanism that explains the formation of policy network structure: a causal path between policy funding context, the locus of resource dependence, advocacy mobilization, and network structure. The last section concludes with a discussion of study limitations and implication for future research. Theoretical Discussion Remaining Puzzles Regarding Resource Dependence, Political Mobilization, and Networks in Social Service Policy Domains Previous work on policy networks has discussed the centrality of resources as an environmental factor that shapes the political behavior of organizations involved in policymaking. In these models, policy decisions are a social product that results from the interplay between (1) the level of resources and (2) the political dynamics between resource holders who distribute public funds (e.g., legislators, legislative committees, or state agencies) and resource seekers (e.g., private or nonprofit service providers, advocacy groups) who must acquire resources to provide public goods and services and (often but not always) to survive (c.f., Milward and Provan 2000; Mosley 2012; Saidel 1991; Salamon 1987; Smith and Lipsky 1993; Park and Rethemeyer 2014). Resource interdependence weaves organizations together, and a complicated and persistent set of relational configurations emerge as a policy network. In particular, some policy actors may gain “social structural resources” (SSRs) from their informal roles and structural positions that are mutually agreed upon by members of the network. SSRs differ from “material institutional resources” (MIRs) in that they grant certain actors influential positions from which to mobilize resources from other members—particularly preferential access to actors with state authority. Previous work suggests that SSRs endure over time unless the relational configuration of a policy network substantially changes (Hatmaker and Rethemeyer 2008). However, previous studies have not sufficiently discussed the question of funding mechanisms in explaining what causes variation in the structure of policy networks across social service areas. The missing factor is policy funding context. Although resource exchanges (money, personnel, etc.) are typically voluntary, governmental mandates and their funding flows impose and enforce certain interorganizational connections and even organizational forms (Knoke 2014). The critical issue is that the locus of resource dependence in a network can be shaped by the nature of the policy funding stream. If all funding is discretionary, then the locus of resource dependence resides in the legislative and executive processes that allocate the resources. Here, legislators, legislative committees, and executive officials define the political landscape. These actors seek reliable information to evaluate policy proposals and decide funding allocations (Knoke 2011; Park and Rethemeyer 2014). Alternatively, if funds are allocated by mandate, new avenues for policy change open and thus the locus of resource dependence shifts. One could advocate for new mandates in the legislature; one could advocate for new interpretations of existing mandates in executive agencies; or one could seek legal redress in the courts (administrative and civil). Similarly, policy funding context also helps to shape how advocacy organizations mobilize their resources in networks by determining which organizations and advocacy tactics are considered legitimate. At the most basic level, policy funding context may make traditional interest-based advocacy more or less useful and thus determine whether advocacy organizations form at all. If advocacy organizations do form, policy funding context can also help to determine what tactics such organizations use and which spectrum of policymakers the advocacy tactic is usually directed toward. If funding levels are discretionary, then legislative tactics are clearly useful as legislators and legislative committees have a greater role in the allocation of resources to social services. If funding levels are mandatory, then legal and administrative tactics may come to the fore. Therefore, we focus on the following questions in an effort to explore factors affecting policy network structures: How do differences in policy funding context (discretionary versus mandated) shape the locus of the resource dependence between organizations operating in a given social service domain? How do differences in policy funding contexts influence advocacy mobilization processes in a network? Discretionary Versus Mandated Social Service Policy Funding Contexts Under the banner of “the welfare state,” government makes many promises to socially vulnerable people with respect to their living standards and social well-being. In current federal, state, and local policy decision making and implementation processes, social service policy is generally executed through either discretionary or mandatory budget allocations. Discretionary spending refers to “expenditure that is governed by annual or other periodic appropriations rather than by formulas or criteria set forth in authorizing legislation” (OECD 2012, 3). Examples include health care spending for American Indians and Alaska Natives by the Federal government; early childhood education, health, and nutrition programs for low-income children and families; Head Start; scientific research through the National Institutes of Health and National Science Foundation; and food assistance for Women, Infants, and Children. Specifically, when a social service program is authorized on a multi-year basis with general policy guidelines and maximum spending, funds are annually appropriated by the legislature. Legislators may choose to appropriate amounts that range from zero to the maximum of the authorization. Thus, advocates for a given social service program must, at least annually, seek political support from lawmakers for continuation of the program, or the programs will die. Since there are often no alternative markets for the goods and services that the social service providers produce, providers often fail without annual state contracts or grants (Park and Rethemeyer 2014). In this sense, we expect that resource seekers in a social service area where state resources are distributed discretionarily will actively engage in legislative political action and coalition building (e.g., organizing an industry association) to secure the resources necessary for survival. Legislative advocacy, such as writing letters, sending emails, or making phone calls to members or creating coalitions to collectively lobby for legislation, is often required to maintain state effort when services are discretionarily funded. Additionally, policy research can become an advocacy tool for nonprofits and other policy stakeholders—but a tool that can cut both ways. For organizations that seek change, policy research can be an aggressive advocacy tactic, as it can embarrass legislators and executive agency leaders by demonstrating that current policies and programs are ineffective, inefficient, or both (Berry 2005; Knoke 2011). However, supportive policy research can help to cement political relationships with key legislators and agencies heads by providing compelling evidence that policy is working and that agencies are effective. Indeed, there is some evidence that in unstable funding environments, legislative, and executive decision makers look favorably on supportive advocacy activities such as policy research (Gais and Walker 1991). By contrast, mandatory spending refers to expenditures administered through enacted law. In advance of annual appropriations, the law guarantees necessary money for mandated programs (Westmoreland and Watson 2006). If there are no explicit changes in formulas or criteria written into law, the previous year’s budget bill is applied to the current year (Bowen, Chen, and Eraslan 2014). Mandatory spending is generally characterized as either open-ended or capped. For example, Medicaid and Medicare are open-ended mandatory spending; thus, the federal government must provide guaranteed health services to targeted individuals. The State Children’s Health Insurance Program is an example of capped mandatory spending. In many cases, growth in mandatory spending automatically occurs without legislative intervention (Westmoreland and Watson 2006). Thus, resource seekers in a social service area where state resources are distributed via mandatory allocations may not have strong incentives to engage in legislative politics. Previous work has already demonstrated that once mandatory programs come into being, the organizational landscape around a policy begins to change. Organizations develop that take as their mission defense of the mandatory service. For instance, Gais and Walker (1991) chronicle how a number of occupational associations (for health care providers, social workers, and social service professionals) were actually started with public funds to help support mandatory programs. Once created, the leaders of these groups remained in close contact with administrative agencies that implemented these mandatory social services (Gais and Walker 1991). As interest groups based upon occupational or commercial communities came to support themselves through membership dues or private foundations, they increasingly employed “outside” strategies of political influence, such as mobilizing public opinion through the media, instead of overt “inside” strategies such as legislative advocacy (Gais and Walker 1991). Additionally, a different set of advocacy tools become available in a mandatory environment. Most importantly, judicial processes to assure steady streams of funding (Scheppele and Walker 1991) become viable. Once mandated by law, legal client advocacy organizations may use the courts to (1) provide explicit legal remedies for unfavorable government action with respect to individuals as well as served populations and (2) pursue their policy agendas concerning the passage of favorable legislation and the expansion of public services (Scheppele and Walker 1991). The existing literature on discretionary and mandatory policy contexts provides ample clues that the organizational environment and advocacy strategies that interested parties use differ depending on whether the social service is mandatory or discretionary. Yet the existing work on policy networks does not account for these differences when examining the determinants of network structure. We now turn to filling in that gap. Study Background To examine the relationship between policy funding context and network structure in social service domains, we examine two extreme cases: one where the services provided are funded discretionarily and another where services are largely (though not in all cases) funded through mandates. The discretionary case focuses on policy related to adult basic education. This area of social service policy is characterized by relatively small budgets and low levels of salience beyond provider and user constituencies, making it difficult to receive media attention. Advocacy organizations are more difficult to organize because the user community is generally very disadvantaged. Adult learners tend to be socioeconomically disadvantaged, and English as a Second Language (ESL) communities are further inhibited from political participation through lack of legal or political standing since many service users are immigrants, both legal and illegal. Advocacy organizations tend to focus on government resource flows because it is a discretionary program (Park and Rethemeyer 2014). Policy advocacy is often carried out through research publications for the legitimization of policy (Biesta 2007) by highlighting problems and the return to public investment in literacy. However, educational service providers also engage in more traditional forms of political mobilization by engaging affected communities in letter-writing campaigns, by directly mobilizing their clients, and through efforts to build relationships with the major legislative decision makers. The second case is mandatory in nature and focuses on policymaking for the severely mentally ill. Mental health care services are mostly mandatory at the state level. The policy funding context for mental health care is similar to that of general health policymaking. Health policymaking is “highly politicized, with significant lobbying by pharmaceutical and medical technology industries as well as by health providers and consumer groups” (Child and Grønbjerg 2007, 262). Health policy is controversial and receives extensive media attention (Woolf 2009). Mental health policymaking is characterized by multiple advocacy interests because (1) there are multiple constituencies (e.g., patients, patients’ families, employers, insurance companies, hospitals, pharmaceutical and medical companies, outpatient service providers, and taxpayers), (2) the policy area is heavily regulated, and (3) the funding levels are very substantial (Child and Grønbjerg 2007; Woolf 2009). A previous study suggested that advocacy organizations participate more heavily in health-related policy than in policymaking on education, public benefits, or religion (Child and Grønbjerg 2007). In particular, policy advocacy in health policy is often carried out through litigation. Since diverse constituencies can be mobilized around rights and winning in court secures a more permanent victory, litigation can be an attractive option for advocacy organizations to support public health policy (Scheppele and Walker 1991). Therefore, we expect that social service policy networks in areas of mandated spending are more likely to attract participation by legal client advocacy organizations than when spending is discretionary. Methods Data Collection and Data Sets This study draws on two policy networks collected from (1) a mental health policy network and (2) an adult basic education policy network. The mental health policy network data (40 actors; 37 respondents) was collected in 2001; the adult basic education policy network data (47 actors; 41 respondents) was collected in 2005. Both data were collected using semi-structured interviews in the same state eastern US state (Newstatia) that has more than 10 cities with at least 50,000 inhabitants. Newstatia is a compelling research site for social policy networks as there is both a great deal of demand and large expenditures in both areas of social service. Newstatia has served adult basic learners and the mentally ill for more than a century. At the time of the study, Newstatia ranked in the top quintile in immigrant population per capita, per-learner expenditures, and total expenditures on adult basic education. Newstatia also ranked in the second-highest quintile regarding severely mentally ill patients served and in the top quintile in terms of total and per capita state spending on the severely mentally ill. ABE policy is defined as those decisions that affect the funding or regulation of organizations that provide educational services to individuals 16 years of age or older who are seeking to raise their reading, writing, and/or computational skills to a level closer to that of a high-school graduate (US Congress Office of Technology Assessment 1993). Mental health policy is defined as decisions that affect the quality and quantity of services available from public and private sector sources to children and adults who have severe mental disorders that interfere with some area of social functioning (US Department of Health and Human Services 1999). Policy Funding Contexts and Network Selection While the data for this study was originally collected for other purposes, the networks studied here conform to a theoretical sampling strategy (Eisenhardt and Graebner 2007) to understand how network structures differ by policy funding contexts. The key differentiator is the nature of the funding streams in these social policy domains: services to the severely mentally ill are legally mandated while support for low-literate adults is at the discretion of state decision makers in the legislature. Additionally, the level of resources devoted to the policy areas is substantially different: support for the severely mentally ill is measured in billions of dollars while support for low-literate adults is measured in millions of dollars. Moreover, the two cases differ with respect to resource dependence on state funding: for-profit and nonprofit ABE providers are much more dependent on state contracts and grants than are the for-profits and nonprofits that provide MH services (Rethemeyer 2007). Network Specification This study used a three-step network specification method in the “realist” tradition outlined by Laumann, Marsden, and Prensky (1989). Rather than using researcher-defined criteria (the “nominalist” approach), the realist approach uses informants to identify network members. The first step was to develop a “naïve” universe of potential policy network members through searches of the Internet, newspaper reports, and policy documents. We then asked three informants from the naïve universe who were clearly central to policymaking in these domains to vet and expand the naïve list. Next, we gave seven members of each network the master list and asked the respondents to rate each organization on a scale of 1–3 with respect to their policy influence. As a backstop to our initial specification, we asked each informant we interviewed during the main data collections if any organizations were omitted from the list. The final specifications included 40 actors in the mental health network and 47 actors in the ABE network. Our informant checks found no organization was listed as “missing” by more than 20% of the network members during the data collection, which suggested high convergence on the membership of both networks. Most of the informants were CEOs; a few were government relations specialists. Measurements Network Ties Ties within a policy network can be measured in various ways. We focus on communications that are foundational to the outputs of policy networks: policy decisions (executive, legislative, and/or judicial) that are built on shared understandings of political and policy preferences, knowledge of preference intensity and power distributions, and communal beliefs about policy cause and effect. Common understandings are the product of communication. Thus, network ties were measured by the extent to which a pair of policy actors maintain routine and/or confidential communication relationships (Knoke 1990; Laumann and Knoke 1987). These measures have been widely used to capture structural relationships that affect policy decisions, outputs, failures, and changes that cannot be explained solely by reference to formal, task-based structural design. More importantly, routine communication is used to scan political environments, while confidential communication is used both to establish meaning and to discuss distribution and allocation of resources (Laumann and Knoke 1987; Raab 2002). We presented each informant two rosters that contained the name and contact person for the organizations identified through the network specification procedure. For each roster, we asked informants to estimate the monthly frequency of the two types of communication—routine and confidential—using a zero to seven scale that was anchored to cues for the number of contacts per month. For this study, we used the data on confidential communication ties as the dependent network because confidential communication contains high-value political and policy material (Park and Rethemeyer 2014). The dependent networks were dichotomized for the ERGM analyses, with a “1” assigned to any dyad with communications frequency of at least a few times a year. This cutoff value was chosen because our qualitative data and knowledge of the domain actors suggested that actors that communicated fewer than a few times a year were not regularly engaged in policy discussions sufficiently to influence outcomes (Park and Rethemeyer 2014). Resource Dependence Resource dependence has been measured in a variety of ways (Casciaro and Piskorski 2005; Hillman, Withers, and Collins 2009; Provan, Beyer, and Kruytbosch 1980). We used both objective and subjective measures for resource dependence to capture how MIRs and SSRs may affect tie choice in a policy network. Regarding objective measures of resource dependence, we collected information on the nature of financial flows and regulation—two measures typically used in a resource-scarce environment (Cho and Wright 2004)—through the operationalization of the capacity to control others’ behavior legitimately (Park and Rethemeyer 2014; Provan, Beyer, and Kruytbosch 1980). Using a roster of network members, we asked informants to note (1) which organizations provide them funding and (2) which organizations regulate their activity. Additionally, we considered whether or not each focal actor has formal authority within the public sector conferred by an elected or appointed office or formal position in the civil service. Since formal authority is a scarce resource in highly institutionalized settings like policy networks, other actors without formal authority, such as service providers and advocacy organizations, attempt to gain access to public actors directly or through intermediaries (Knoke et al. 1996). For this reason, informants were asked to indicate whether they were (1) executive actors or (2) legislative actors. As a subjective measure for focal actor’s dependence, we used policy actors’ evaluation of each network member’s influence on decision making in the ABE and MH policy networks. Organizations seek to mold their resource environments through political activity (Emerson 1962; Hillman, Withers, and Collins 2009; Pfeffer and Salancik 1978). To change the resource environment within a policy network, actors evaluate resources—both MIRs and SSRs—that they and other players have and then select their political actions. Since power resides in and is constrained by relationships (Emerson 1962), the pattern of exchange in a network is different from that of dyadic exchange between two actors (Bonacich and Bienenstock 2009). Actors’ perceptions of the influence of others on resource allocations in a network may greatly affect their choice of relational strategies to acquire resources. It is important to include a subjective measure of dependence—perceived influence—to understand how resource dependence (and network members’ perceptions of it) differentially constrains tie selections across social policy domains. To measure influence, we asked each informant to rate the “policy influence” of each member of the roster on a scale from zero to seven. Influence was defined as follows: the ability to get others to believe, think, or act as one prefers with respect to a given policy issue. The more influence an actor has, the more likely it is that policies in line with their preferences will be implemented. To purge the resulting influence scores of raters’ tendency to anchor on high or low values, we transformed each raters’ responses into a Z-score using the mean and standard deviation of each rater’s responses. Organizational Advocacy Information To examine the differential structural participation of advocacy organizations, we first had to identify the nature of all organizations in the network. We used an existing typology from earlier work (Park and Rethemeyer 2014) but refined the advocacy categories to include (1) client advocacy organizations, (2) legal client advocacy organizations, (3) industrial/professional associations and foundations, (4) research and technical support organizations, and (5) service delivery organizations. Network Structural Variables A number of structural variables such as reciprocity, three-cycles, two-paths, and other higher-order structures (e.g., preference for higher outdegree nodes, preference for higher indegree nodes, alternating k-triangles, and alternating two-paths) are included in our model to control for generic social pressures on organizations’ relational choices. For instance, if Actor A is friends with Actor B, and Actor B is friends with Actor C, it is very rare that Actor A is not friends with Actor C because such an instance would violate transitivity (i.e., a tendency for a friend of a friend to be a friend) as a generic social pressure (Park and Rethemeyer 2014). The “microstructures” representing generic social pressures have significant effects on global network structures (Pattison and Robins 2002). Previous studies on policy network have shown that generic social processes pressure policy actors to forge ties that may not be formed otherwise (Ingold and Leifeld 2016; Park and Rethemeyer 2014). Analysis We used an exponential random graph model (ERGM) to analyze our sociometric data. ERGM’s were developed to account for dependence between dyads (Frank and Strauss 1986; Pattison and Wasserman 1999; Robins et al. 2007a; Robins, Pattison, and Wasserman 1999). Standard linear regression models of dyadic data are inherently biased due to these dependencies. Additionally, the dependencies themselves are interesting, as they represent patterns of social interaction that can be analyzed—for instance, reciprocity, the tendency for closed social structures (a friend of a friend is a friend), and so forth. ERGMs can easily accommodate relationships, attributes of an actor, and network structural variables as predictors of the dependent network—here meaning the confidential communications network (Snijders et al. 2006). ERGMs assume that a network may be represented by a random graph that follows a Markov change process. That is, the probability of a tie between any pair of actors is assumed to be random (i.e., independent) given that the rest of the network consists of certain conditionally dependent ties. A tie from actor i to actor j is conditionally dependent only on other possible ties involving i and/or j and other features that systematically make some ties more likely than others—for instance, the well-known preference for homophily in many social interactions (Robins et al. 2007a). For an introduction to stochastic social network analysis, see Robins et al. (2007b); Snijders (2001, 2002a); Snijders and Duijn (2002); for a fuller discussion of modeling procedures, see Park and Rethemeyer (2014). Appendix discusses the assumptions and mechanics of ERGMs in detail. We executed our ERGM using the Simulation Investigation for Empirical Network Analysis (also known as SIENA, see Snijders 2002b) package in StOCNET (Boer et al. 2006) that estimates parameters based on simulation techniques (Markov Chain Monte Carlo maximum likelihood estimation (MCMCMLE) models—see Robins et al. 2007b). We used NetDraw (Borgatti 2002) to create network visualizations and UCINET (Borgatti, Everett, and Freeman 2002) to calculate deterministic network statistics. ERGM models are said to adequately represent the underlying data when the features included in the model (e.g., the number of reciprocal dyads, the prevalence of ties between organizations of a similar age, etc.) may be reproduced through simulations with an average value that is statistically indistinguishable from the measured values in the data. To determine whether the model has “converged,” the average value of the modeled features from a large set of simulations (5,000 in this case) is compared to the measured value using a t-statistic. Convergence is said to occur when the t-value is very near zero—by convention, less than 0.10 for all included features. Both models reported below are fully converged by this standard. Interpretation of results in ERGM is similar to logit parameter estimates—the sign and significance may be directly interpreted, but the magnitudes may not be. However, ERGM creates more complex outputs about attributes that may affect tie formation. One must consider three effects of attributes: ego, alter, and same (or similarity). Ego effects capture whether certain attributes make an actor more “outgoing” (more likely to make connections) if the coefficient is positive and significant. Alter effects capture whether an attribute makes an actor more “popular” (more likely to be sought out by others) if the coefficient is positive and significant. “Same” effects capture homophily when the variable is binary, while “similarity” effects capture homophily when the variable is continuous. Do actors prefer others with the same (or similar) attribute if the coefficient is positive and significant? For an in-depth introduction to ERGM parameters and the interpretation of the estimates, see the SIENA Web site and manual (Ripley and Snijders 2009; Snijders 2002b). Findings The Structure of Discretionary and Mandatory Social Service Policy Networks By taking an inductive stance, our findings help us to develop a series of propositions about the general nature of policy network structure in light of differences in policy funding context. Table 3 reports the final models, including a reduced set of variables found to have significant coefficients through joint goodness-of-fit tests for both ABE and MH policy networks. Our findings indicate that differences in policy funding contexts (discretionary versus mandated) play a major role in determining the locus of the resources dependence, the nature of advocacy mobilization processes, how dependence and advocacy interact with one another, and thus network structures. Table 1. Actor Typology of Mental Health in Newstatia, US Description  Newstatia, US (2001)  State agencies  7  Legislators  8  Legislative committees  4  Industry associations  3  Professional associations/unions  3  Service providers  6  Legal client advocacy organizations  4  Client or family advocacy organizations  5  Newspapers  2  Insurers  3  Researchers/research teams  3  Description  Newstatia, US (2001)  State agencies  7  Legislators  8  Legislative committees  4  Industry associations  3  Professional associations/unions  3  Service providers  6  Legal client advocacy organizations  4  Client or family advocacy organizations  5  Newspapers  2  Insurers  3  Researchers/research teams  3  Note: Total will not add to 40; two advocacy organizations are also government agencies; one research team is also state agencies; five industry/professional associations are also service providers. View Large Table 1. Actor Typology of Mental Health in Newstatia, US Description  Newstatia, US (2001)  State agencies  7  Legislators  8  Legislative committees  4  Industry associations  3  Professional associations/unions  3  Service providers  6  Legal client advocacy organizations  4  Client or family advocacy organizations  5  Newspapers  2  Insurers  3  Researchers/research teams  3  Description  Newstatia, US (2001)  State agencies  7  Legislators  8  Legislative committees  4  Industry associations  3  Professional associations/unions  3  Service providers  6  Legal client advocacy organizations  4  Client or family advocacy organizations  5  Newspapers  2  Insurers  3  Researchers/research teams  3  Note: Total will not add to 40; two advocacy organizations are also government agencies; one research team is also state agencies; five industry/professional associations are also service providers. View Large Table 2. Actor Typology of Adult Basic Education in Newstatia, US Description  Newstatia, US (2005)  State agencies  7  Legislators  7  Legislative committees  5  Industry associations  1  State-funded service providers  12   Community-based organizations  6   Community colleges  0   Municipal agencies  3   School districts  3   Union  0  State-funded technical assistance units  9  Foundations  1  Research/advocacy organizations  3  Learner advocates  1  Consultants  1  Description  Newstatia, US (2005)  State agencies  7  Legislators  7  Legislative committees  5  Industry associations  1  State-funded service providers  12   Community-based organizations  6   Community colleges  0   Municipal agencies  3   School districts  3   Union  0  State-funded technical assistance units  9  Foundations  1  Research/advocacy organizations  3  Learner advocates  1  Consultants  1  View Large Table 2. Actor Typology of Adult Basic Education in Newstatia, US Description  Newstatia, US (2005)  State agencies  7  Legislators  7  Legislative committees  5  Industry associations  1  State-funded service providers  12   Community-based organizations  6   Community colleges  0   Municipal agencies  3   School districts  3   Union  0  State-funded technical assistance units  9  Foundations  1  Research/advocacy organizations  3  Learner advocates  1  Consultants  1  Description  Newstatia, US (2005)  State agencies  7  Legislators  7  Legislative committees  5  Industry associations  1  State-funded service providers  12   Community-based organizations  6   Community colleges  0   Municipal agencies  3   School districts  3   Union  0  State-funded technical assistance units  9  Foundations  1  Research/advocacy organizations  3  Learner advocates  1  Consultants  1  View Large Table 3. MCMCMLE of Policy Networks (ABE and MH)   Parameters  ABE  MH  Structural configuration  Reciprocity  1.3558** (0.1838)  1.1542** (0.2518)    3-cycles  0.0562 (0.0525)  −0.166* (0.0738)  2-paths  0.0126 (0.0132)    Preference for higher outdegree nodes  −0.2684 (0.3907)  −0.2877 (0.3212)  Preference for higher indegree nodes  −8.8049* (4.4680)  −1.0063* (0.4016)  Alternating k-triangles  1.2885** (0.2367)  1.1402** (0.1698)  Alternating 2-paths    −0.0664* (0.031)  Regulators  Regulators ego  −0.1717 (0.4110)    Regulators alter    −0.3289 (0.2404)  Same regulators  −0.2413 (0.2574)    Funders  Funders ego  0.8034* (0.3418)  −0.2713* (0.1331)  Same funders  0.4139* (0.2048)    Regulatory/funding dependence by public actors  Regulated and funded by public actors ego  0.8023** (0.1739)    Regulated and funded by public actors alter  −0.3396† (0.1898)    Same regulated and funded by public actors  0.2550** (0.0984)    Public actors  State actors alter  −0.5488* (0.2753)  0.4142* (0.1758)  Same state actor  0.0266 (0.1757)    Legislator and legislative committee ego  0.6218** (0.2015)    Legislator and legislative committee alter  −0.4122† (0.2255)  0.479** (0.1308)  Same legislator and legislative committee  0.6441** (0.13080  0.4818** (0.1044)  Influence  Influence ego    0.0039 (0.0037)  Influence alter  0.0124** (0.0033)  0.0187** (0.0041)  Influence similarity    1.0373** (0.3084)  NonprofitOrgs.         Legal client advocacy organizations  Legal client advocacy organizations ego    0.3388* (0.1695)   Industrial/professional associations  Industrial & professional associations and foundations ego  −0.0588 (0.2466)  0.5225* (0.2451)   Research and technical support organizations  Same research and technical supports organizations  0.4398** (0.1132)     Service providers  Service providers ego  0.0132 (0.1014)  −0.0448 (0.237)    Parameters  ABE  MH  Structural configuration  Reciprocity  1.3558** (0.1838)  1.1542** (0.2518)    3-cycles  0.0562 (0.0525)  −0.166* (0.0738)  2-paths  0.0126 (0.0132)    Preference for higher outdegree nodes  −0.2684 (0.3907)  −0.2877 (0.3212)  Preference for higher indegree nodes  −8.8049* (4.4680)  −1.0063* (0.4016)  Alternating k-triangles  1.2885** (0.2367)  1.1402** (0.1698)  Alternating 2-paths    −0.0664* (0.031)  Regulators  Regulators ego  −0.1717 (0.4110)    Regulators alter    −0.3289 (0.2404)  Same regulators  −0.2413 (0.2574)    Funders  Funders ego  0.8034* (0.3418)  −0.2713* (0.1331)  Same funders  0.4139* (0.2048)    Regulatory/funding dependence by public actors  Regulated and funded by public actors ego  0.8023** (0.1739)    Regulated and funded by public actors alter  −0.3396† (0.1898)    Same regulated and funded by public actors  0.2550** (0.0984)    Public actors  State actors alter  −0.5488* (0.2753)  0.4142* (0.1758)  Same state actor  0.0266 (0.1757)    Legislator and legislative committee ego  0.6218** (0.2015)    Legislator and legislative committee alter  −0.4122† (0.2255)  0.479** (0.1308)  Same legislator and legislative committee  0.6441** (0.13080  0.4818** (0.1044)  Influence  Influence ego    0.0039 (0.0037)  Influence alter  0.0124** (0.0033)  0.0187** (0.0041)  Influence similarity    1.0373** (0.3084)  NonprofitOrgs.         Legal client advocacy organizations  Legal client advocacy organizations ego    0.3388* (0.1695)   Industrial/professional associations  Industrial & professional associations and foundations ego  −0.0588 (0.2466)  0.5225* (0.2451)   Research and technical support organizations  Same research and technical supports organizations  0.4398** (0.1132)     Service providers  Service providers ego  0.0132 (0.1014)  −0.0448 (0.237)  †p < 0.1; *p < .05; **p < .01. View Large Table 3. MCMCMLE of Policy Networks (ABE and MH)   Parameters  ABE  MH  Structural configuration  Reciprocity  1.3558** (0.1838)  1.1542** (0.2518)    3-cycles  0.0562 (0.0525)  −0.166* (0.0738)  2-paths  0.0126 (0.0132)    Preference for higher outdegree nodes  −0.2684 (0.3907)  −0.2877 (0.3212)  Preference for higher indegree nodes  −8.8049* (4.4680)  −1.0063* (0.4016)  Alternating k-triangles  1.2885** (0.2367)  1.1402** (0.1698)  Alternating 2-paths    −0.0664* (0.031)  Regulators  Regulators ego  −0.1717 (0.4110)    Regulators alter    −0.3289 (0.2404)  Same regulators  −0.2413 (0.2574)    Funders  Funders ego  0.8034* (0.3418)  −0.2713* (0.1331)  Same funders  0.4139* (0.2048)    Regulatory/funding dependence by public actors  Regulated and funded by public actors ego  0.8023** (0.1739)    Regulated and funded by public actors alter  −0.3396† (0.1898)    Same regulated and funded by public actors  0.2550** (0.0984)    Public actors  State actors alter  −0.5488* (0.2753)  0.4142* (0.1758)  Same state actor  0.0266 (0.1757)    Legislator and legislative committee ego  0.6218** (0.2015)    Legislator and legislative committee alter  −0.4122† (0.2255)  0.479** (0.1308)  Same legislator and legislative committee  0.6441** (0.13080  0.4818** (0.1044)  Influence  Influence ego    0.0039 (0.0037)  Influence alter  0.0124** (0.0033)  0.0187** (0.0041)  Influence similarity    1.0373** (0.3084)  NonprofitOrgs.         Legal client advocacy organizations  Legal client advocacy organizations ego    0.3388* (0.1695)   Industrial/professional associations  Industrial & professional associations and foundations ego  −0.0588 (0.2466)  0.5225* (0.2451)   Research and technical support organizations  Same research and technical supports organizations  0.4398** (0.1132)     Service providers  Service providers ego  0.0132 (0.1014)  −0.0448 (0.237)    Parameters  ABE  MH  Structural configuration  Reciprocity  1.3558** (0.1838)  1.1542** (0.2518)    3-cycles  0.0562 (0.0525)  −0.166* (0.0738)  2-paths  0.0126 (0.0132)    Preference for higher outdegree nodes  −0.2684 (0.3907)  −0.2877 (0.3212)  Preference for higher indegree nodes  −8.8049* (4.4680)  −1.0063* (0.4016)  Alternating k-triangles  1.2885** (0.2367)  1.1402** (0.1698)  Alternating 2-paths    −0.0664* (0.031)  Regulators  Regulators ego  −0.1717 (0.4110)    Regulators alter    −0.3289 (0.2404)  Same regulators  −0.2413 (0.2574)    Funders  Funders ego  0.8034* (0.3418)  −0.2713* (0.1331)  Same funders  0.4139* (0.2048)    Regulatory/funding dependence by public actors  Regulated and funded by public actors ego  0.8023** (0.1739)    Regulated and funded by public actors alter  −0.3396† (0.1898)    Same regulated and funded by public actors  0.2550** (0.0984)    Public actors  State actors alter  −0.5488* (0.2753)  0.4142* (0.1758)  Same state actor  0.0266 (0.1757)    Legislator and legislative committee ego  0.6218** (0.2015)    Legislator and legislative committee alter  −0.4122† (0.2255)  0.479** (0.1308)  Same legislator and legislative committee  0.6441** (0.13080  0.4818** (0.1044)  Influence  Influence ego    0.0039 (0.0037)  Influence alter  0.0124** (0.0033)  0.0187** (0.0041)  Influence similarity    1.0373** (0.3084)  NonprofitOrgs.         Legal client advocacy organizations  Legal client advocacy organizations ego    0.3388* (0.1695)   Industrial/professional associations  Industrial & professional associations and foundations ego  −0.0588 (0.2466)  0.5225* (0.2451)   Research and technical support organizations  Same research and technical supports organizations  0.4398** (0.1132)     Service providers  Service providers ego  0.0132 (0.1014)  −0.0448 (0.237)  †p < 0.1; *p < .05; **p < .01. View Large Finding 1: In both ABE and MH policy networks, regulators (who have oversight of policy actors in the network) are not important in structuring policy connections. Finding 2: In both ABE and MH policy networks, funders (who provide funding to policy actors in the network) are important in structuring policy connections. Regarding resource holders’ dynamics, both the ABE and MH networks are affected by funders’ network formation activities—but in different ways—while neither network is structured by relationships built to or from regulators. Turning first to the findings regarding regulation, regulators do not play a role in structuring the network. For both the ABE and MH networks, the coefficients in the Regulators section of the table are statistically insignificant. Regulators do not make connections at a higher rate than other actors in both networks; they are not sought as communications partners in either network; and they do not tend to communicate with each other differentially. While regulators hold a key resource for service providers—authorization to engage in a practice or service—that resource is necessary but not sufficient. Providers can often find ways to satisfy regulatory obligations, but they cease to exist without resources. Policy networks exist primarily to seek material resources, not authorization. Thus, regulators do not play a significant part in creating the structure. However, the structure of both networks is affected by funders, but the ABE network is far more thoroughly affected by relationships driven by the imperative of funding. In most social service policy domains, state funding is an indispensable resource that helps service providers assist socially disadvantaged people (Park and Rethemeyer 2014). Thus state funders are imperative. Previous studies have demonstrated that financial resource dependence is an important shaper of social policy network structure (Park and Rethemeyer 2014). We also find that both networks are affected by funding relationships, but quite differently. Specifically, funding flows (i.e., who funds whom) are a strong predictor of the creation of confidential policy communication ties within the discretionary ABE policy network. Funders ego and Same Funders in ABE are positive and significant at the 5% level. Funders in the ABE network actively shape policy communication, and they have a strong tendency to talk about sensitive political information with each other. Resource holders such as the State Department of Education and the industry association proactively gather reliable tacit, technical, or proprietary information through personal contact in order to make better decisions about resource allocation (Gulati and Singh 1998; Powell 1990; Powell, Koput, and Smith-Doerr 1996). Moreover, because these funding streams are discretionary, the process of collecting data to guide funding must be continuous, intensive, and timed to the legislative calendar. However, in the mandated MH network, our results suggest a different story: MH funders are significant policy actors in structuring the network, but they are not active in the creation of ties with other policy actors (see negative and significant coefficient on Funders ego in MH in table 3) even though there are very substantial funding relationships in the MH network. In fact, funders in the MH network tend to be less likely to create ties to other network members. They do not engage in network formation actively. We will discuss how these findings relate to differences in policy funding context below. Finding 3: In the ABE policy network, financial and regulatory resource seekers (who are given funding and regulated by public sector actors in the network) are important in structuring policy connections. Resource seekers are active and homophilous in forming ties but are less popular as partners in the network. Resource holders and resource seekers make contacts in a policy network differently. Specifically, resource seekers’ relational choices are affected by their evaluation of their dependence on resource holders (Park and Rethemeyer 2014). We examined how relational behaviors of resource seekers differ by policy funding context and found that funding and regulatory dependence is significant in structuring policy connections only within the discretionary ABE policy network. By contrast, even though the resource dependence related variables were included in the model, these variables are not significant predictors of connections within the mandated MH network. The coefficients in table 3 on Regulated & funded by public actors ego and Same funded & regulated by public actors are positive and significant at the 1% level. Also, the coefficient on Funded & regulated by public actors alter is negative and significant at 10% level. In the ABE network, the funding and regulatory dependence relationships are highly correlated; thus, we included only one of these relationships in the model and interpreted the results from a combined funding/regulation dependence perspective. Our funding/regulatory dependence variables are “1” if policy actors are regulated/funded by a public actor within the networks. Thus, policy actors regulated and funded by public actors tend to create ties with others in the ABE policy network. At the same time, they are strongly homophilous: those organizations that are subject to regulation and funding dependence tend to communicate with one another. However, not surprisingly, they are not popular within the ABE network because they do not hold the key financial resources in the network. Our findings and the ABE interview data suggest that those organizations funded and regulated by public actors (mainly ABE service providers and technical assistance units) actively responded to the post-9/11 economic crash by extending their brokerage positions between the provider community and legislative actors (Park and Rethemeyer 2014). They realized that they may not be able to “live off the state” in a hostile budget environment. They were maintaining relationships with resource holders and seeking new relationships with organizations that built successful state funding portfolios (Park and Rethemeyer 2014). Also, as seen in figure 1, resource seekers, such as state-funded ABE service providers (circle in a box) and technical assistance units (box), created a coalition of resource seekers within their own subgroups. In this way, they may be attempting to offset their dependence on resource holders to ensure their survival and cope with uncertainty about state funding. With the data at hand we cannot conclusively demonstrate how much “more” the ABE network structure is the product of resource seekers building relationships as compared to the role of resource seekers in the MH network. However, we know that the ABE policy funding context grants state authorities more discretion regarding financial resource allocations than the MH policy funding context. The power imbalance between resource holders and resource seekers is greater in the discretionary ABE domain than in mandated MH domain. We can say that there is no evidence in the ERGM results that the structure of the MH policy network is driven by resource seekers: every variable in the “Regulatory/Funding Dependence By Public Actors” section of table 3 is insignificant. This difference in structural determinants highlights the relationship between network structure and policy funding context in a social service policy domain. From these findings about resource holders and seekers, we state the following research propositions: Figure 1. View largeDownload slide ABE Policy Network, MDS Layout Figure 1. View largeDownload slide ABE Policy Network, MDS Layout Proposition 1: Being a financial resource holder, not a regulatory resource holder, is a stronger driver of relational structures within social service policy networks. Proposition 2: Within social service networks, the extent to which resource dependence affects relational structures differs by policy funding context. Proposition 3: Resource interdependent relationships are more important in structuring policy network in a policy context where expenditures are discretionary than in a policy context where expenditures are mandatory. Proposition 3a: When fewer resources are to be discretionarily distributed, financial and regulatory resource seekers actively create ties and develop internal communication with resource seekers. Finding 4: In the mandatory MH network, executive actors (who hold public authority) are popular as policy partners; however, they are not popular in the discretionary ABE network. Finding 5: In the discretionary ABE network, legislators and legislative committee members tend to actively form ties. Finding 6: In both the discretionary ABE and mandated MH policy networks, legislators and legislative committee members tend to form ties with each other. Based on Knoke et al.’s (1996) study, we examined how public actors (i.e., state agencies, legislators, and legislative committees) affect network structures in both the ABE and MH networks because these actors’ primary resource is public authority. We expected that public actors shape network structures differently depending on the degree of discretion over the level and continuation of funding (discretionary versus mandated). As shown in table 3, we found that state actors are not sought by other actors in ABE network (see the negative and significant coefficient on State actors alter in the ABE network); in fact, the negative coefficient suggests that state alters are unpopular as relationship partners. By contrast, MH policy actors seek state actors as their networking partners within the MH policy network (see the positive and significant coefficient on State actors alter in the MH network). Turning to the legislative actors within the networks, legislators and legislative committees in the ABE network have a strong preference for making more connections (see the positive and significant coefficient on Legislator & legislative committee ego at 1% level in table 3). However, ABE legislators are not sought by other policy actors (see the negative and significant coefficient on Legislator & legislative committee alter at 10% level) and may be avoided (as the negative coefficient implies). By contrast, legislative actors in the MH network are highly sought by other policy actors within the network (see the positive and significant coefficient on Legislator & legislative committee alter at 1% level in table 3). There is also a tendency for the legislative actors to form ties with each other in both the ABE and MH networks. This is evidenced by the coefficients on Same legislator & legislative committee in the ABE and MH networks, which are both positive and significant at the 1% level. Taken together, in the face of differences in funding context, public resource holders are active in the ABE network while they are passive in MH. In the ABE network, legislative actors who control financial flows tend to create direct ties with others, while both executive and legislative actors are not structurally “popular.” This finding suggests that when social policy is discretionary and the financial pie is not particularly large, network structure is differentiated by legislative members’ activities. Legislators and resource holders can increase political and policy monitoring capacity by intensively communicating with legislative members (see the right-side clusters created by up-triangles and “overlapping triangles” in figure 1). Our ABE interview data also suggests that the budget debacle of 2001–2002 drove network structure: A large set of legislative actors became actively involved in the ABE policy network as resource pressures mounted (see Park and Rethemeyer 2014). Legislative actors reach out within the network and especially to peers in order to better coordinate yearly appropriations. This coordination became critical as the 2001–2002 budget crisis drove deep cuts to discretionary funding. However, in the MH network, both executive and legislative actors shape the network by receiving ties from others, not by sending ties. Public sector organizations are endowed with public authority; they depend on other organizations’ information to inform their decision making (Knoke et al. 1996). Walker (1991) demonstrated that public interest groups are mobilized to serve as reliable sources of information by executive agencies as part of the policy process. Simply put, in a social service policy domain, executive actors tend to receive incoming ties from others without developing outgoing ties. Information flows toward executive agencies in this scheme. Moreover, in the MH context, (1) the social service in question—care of the severely mentally ill—is an entitlement rather than a discretionary expenditure, (2) substantially more public resources are at stake (more than $1 billion versus about $100 million), and (3) alternative financial resources exist inside and outside the network in the form of private insurers who must cover some insured individuals with severe mental illnesses. These aspects of the policy funding context make legislators in the MH network less driven to assert control through information gathering via extensive ties structures. A large portion of the mental health budget is on “autopilot” through mandates. Instead, legislative actors may selectively choose among those actors who are actively seeking ties with them to address more narrowly defined issues—for instance, the creation and content of a formulary, which was a major issue at the time this data was collected. While the formulary was of deep interest to some organizations, it was not an existential threat to the overall service structure. The structure is different in ABE, where yearly budget decisions can have sweeping and even existential implications. Legislators are more outgoing but must contend with constraints from powerful nonlegislative actors who wish to assert control over interactions with lawmakers. As seen in figure 2, the MH network is less differentiated but more structurally complex than the ABE network. From these findings, we have derived the following propositions: Figure 2. View largeDownload slide MH Policy Network, MDS Layout Figure 2. View largeDownload slide MH Policy Network, MDS Layout Proposition 4: Legislators and legislative committee members monitor policy environments by creating ties and intensively communicating among themselves when (a) social services are discretionary and (b) financial resources are relatively scarce. Proposition 5: Public actors monitor policy environments by receiving information from others when (a) a social service is mandatory, (b) large amounts of public resources are at stake, and (c) alternative financial sources exist outside of the network. Proposition 6: Legislators and legislative committee members are strongly homophilous in shaping social service policy networks. Finding 7: In both the ABE and MH policy networks, policy actors with high influence scores are popular. As we expected, influential actors in both networks are sought out by other policy actors since both coefficients on Influence alter are positive and significant at the 1% level. However, perceived influence constrains actors’ behavior in choosing instrumental ties in the ABE and MH networks differently. In ABE, influential actors, such as the State Department of Education, legislative committees (especially the Senate and House Ways and Means Committees), legislators, research and advocacy organizations, service providers (literacy volunteer organizations), the industry association (which we pseudonymed NAABE—Newstatia Alliance for Adult Basic Education), and research and technical assistance units are extensively sought by other policy actors. In the case of the MH network, influential actors, such as the state agency (Department of Mental Health), committees (Senate and House Ways and Means Committees), legislators, newspapers, industry association (which we pseudonymed NAMH—Newstatia Alliances for Mental Health), insurers, and client and family advocacy organizations are sought for ties within the MH network. Also, these influential actors tend to restrict their interactions to a set of powerful peers within the MH network (see the positive and significant coefficients on influence alter and influence similarity at 1% significance level in MH in table 3). Thus, information is more closely held in the MH network. Proposition 7: Influential policy actors are strongly sought by other policy actors across social service policy networks despite differences in policy funding context. Finding 8: In the MH network, legal client advocacy organizations and industrial and professional associations are more outgoing. Finding 9: In the ABE network, research and technical support organizations prefer homophilous policy ties. As noted, we focused on five types of nonprofits: (1) client advocacy organizations, (2) legal client advocacy organizations, (3) industrial/professional associations and foundations, (4) research and technical support organizations, and (5) service delivery organizations. We dropped one type—client advocacy organizations—from our ABE and MH models due to the lack of statistical significance.1 Thus, only four types of nonprofit organizations are used for the analysis. In our categorization, advocacy means promoting a policy agenda on behalf of socially marginalized groups through the courts, the legislature, administrative agencies, and/or the public at large and service delivery implies the provision of services directly to individuals and families (Chetkovich and Kunreuther 2006). However, our categorization is not mutually exclusive: service providers advocate, and some advocacy groups provide services. We found that MH and ABE policy networks are shaped differently by these advocacy and service delivery organizations. The coefficients on both Legal client advocacy organizations ego and Industrial/professional associations and foundations ego in the MH network are positive and significant at the 5% significance level (table 3). This finding suggests that legal client advocacy organizations and professional associations and foundations tend to actively engage in creating ties with other policy actors in the MH network. In MH, there are a number of specialized advocacy organizations, including interest groups for family members of people with mental disorders, professional associations that represent those who work with the severely mentally ill, and organizations that use legal means to (1) advocate for policies and laws against discrimination, (2) improve services and secure the just treatment of mentally ill people, and (3) assist clients seeking access to services for those who are severely mentally ill (Funk et al. 2006). These organizations exist because statutes mandate MH services for those with qualifying conditions. Newstatia also financially supports some legal advocacy activities in our case. Figure 3. View largeDownload slide Conceptual Model Linking Policy Funding Context to Network Structure Figure 3. View largeDownload slide Conceptual Model Linking Policy Funding Context to Network Structure Thus, the MH network includes both advocacy organizations that can facilitate mass mobilization for legislative action and organizations that can seek policy change through enforcement of legal rights of citizens with severe mental illnesses. Here, legal client advocacy organizations, representing the formally intended beneficiaries by MH law and policy, play a pivotal role in structuring the MH network. Also, since the health field is highly politicized and institutionalized (Child and Grønbjerg 2007), professional and industrial associations in MH (those that represent inpatient mental health facilities, mental health clubhouses, outpatient clinics, and a union of MH workers) invested in advocating with state actors who have the power to modify regulations and MH programs. For example, the industry association (NAMH) advocates in legislative and executive processes for both institutional members as well as individuals receiving services and their families. The professional lobbyists in these industry associations give testimony in legislative processes and try to educate policymakers about the impact of MH issues. Since government programs are a key source of revenue for the members of the MH industrial and professional associations (Child and Grønbjerg 2007), these associations tend to actively engage in the MH network by seeking new ties with other policy actors. MH’s mandatory policy funding context combined with the existence of secondary interest groups that represent client and industry interests mitigate toward systematic advocacy mobilization processes (e.g., institutionalized lobbying and litigation). In the discretionary ABE context, client advocacy organizations are not a significant factor shaping the ABE network. However, activities by research and technical support organizations were significant. In our first ERGM of the ABE network, we included the adult learners’ advocacy organizations as a variable, but the coefficient was statistically insignificant. This result was not surprising: the learner’s advocacy organization was small, poorly funded, and not well known in Newstatia. Most ABE learners are socioeconomically disadvantaged and thus tend to lack the resources needed to support a robust advocacy organization. Additionally, ABE learners do not have secondary interested parties (like family members in the case of the mentally ill) who are financially and politically positioned to advocate for them. Finally, in the absence of mandates, there is no way to use the courts to promote ABE learner interests and thus no legal client advocacy organizations engaged in this social service policy domain. These policy funding context differences have restricted the opportunities for client advocacy to systematic mobilization of clients by providers. Taken together, our data suggests that client and industry advocacy organizations are more active when legal rights are granted to protect groups of citizens, whereas these activities are lessened when there are no rights to enforce. However, in the ABE network, research and technical assistance organizations emerged as major advocates. The coefficient on Same research and technical support organizations is positive and significant at 1% significance level. This result implies that research and technical support units in the ABE network tend to communicate intensely among themselves. Our field research confirmed that these organizations provided critical infrastructure and coordination for political and policymaking activities by private sector ABE providers. While these organizations certainly consider the interests of their clients—indeed, they exist to serve clients—it is also indisputable that their advocacy was deeply tied to the perspectives of service providers and other funded units. ABE learners’ advocacy organizations do not play a central leadership role in ABE policymaking; only the provider organizations and industry association do. From this analysis, we state the following research propositions: Proposition 8: In policy contexts where services are legally mandated and secondary interested parties exist, legal client advocacy organizations and industrial & professional associations and foundations play an essential role in structuring social service policy networks. Proposition 9: In policy contexts where services are provided using discretionary funds, providers, research & technical support organizations, and other funded entities are important in shaping social service policy networks. Discussion and Conclusion Networks have become an increasingly common feature of both policymaking and service delivery in a wide range of social safety net programs (Kettl 1996; Milward, Provan, and Else 1993; Salamon 1981, 1995). While much of the research has focused on how network structure affects access, participation, and eventually outcomes, few studies examine the relationship between policy funding contexts and structures in social service policy networks. We seek to close this gap by using comparative quantitative analysis of two policy networks—one focused on a discretionary context (ABE) and another on a mandatory context (MH)—to induce a set of propositions about the relationship between funding context and network structure. Here, we have explicitly examined the interaction between policy funding contexts and dominant theoretical traditions that have informed previous work on social policy networks—resource dependence and advocacy mobilization. Our analyses suggest that the nature of the policy funding stream—discretionary or mandatory—affects the nature of dependence between network actors and the type of advocacy mobilization that occurs. In a discretionary social service policy context, funding depends on annual allocations from the legislature. In this context, resource seekers create advocacy coalitions within the policy networks that focus on legislative advocacy. The network structure flows from the need to keep the legislative spigot open. Highly influential intermediary policy actors—actors that do not necessarily fund providers—also play a role in structuring policy networks in a discretionary context. In the ABE example, these actors include the State Board of Education, a legislative committee focused on social services, a state-based educational research center, an industry association, and several state-funded technical assistance units. These organizations intermediate and control interactions between resources holders—legislators—and resource seekers—funded programs and support organizations. It is through interactions between resource seekers and intermediary actors that resource holders coordinate discretionary funding flows. The structure of the discretionary social service policy network is tied to intermediary actors’ efforts and developed to assure there was a financial “pie” of any size to divide. When much of the funding for a social service policy is programmed and mandated, legislative success is not necessarily about survival. Mandates put “guide rails” around the degree to which legislators may choose winners and losers in the absence of legislation to revoke or greatly modify an existing mandate. Revising a mandate is usually (a) politically difficult—if not impossible—given established coalitions and (b) procedurally difficult—if not impossible—if the mandates are encoded into state constitutions or federal statutes. Additionally, in some mandated social service areas—like care of the severely mentally ill and many other health care domains—there are nonpublic sources of funding that provide an alternative to government as the only financial source of survival. These funders are themselves the subject of policy action: insurers, for instance, are the subject of regulation that affects the services available to policyholders and thus the streams of funding available to service providers. In the face of a mandate, the clear financial dependence of service providers on legislative decision makers found in the discretionary context gives way to a much more complex picture. Mandates provide much greater certainty with respect to public funding: policy network members are relatively certain there will be a “pie” to divide. The question is, who gets the bigger slices? Seen in this light, the differences in advocacy structures between discretionary and mandatory social service areas may be clearly traced to differences in policy funding context: the funding structure implies the advocacy structure that flows from it provided one also accounts for differences in client populations. Starting with the question of funding structure, there are various ways to seek a bigger slice of a social service pie: legislative action; litigation to enforce or modify a mandate; efforts to revise the regulatory constraints on alternative funders; and administrative processes. Policy funding context determines which of the “advocacy technologies” is most appropriate or even possible. When funding is almost exclusively discretionary, then many advocacy technologies fall away as irrelevant or impractical: organizing to influence legislative actors is the primary method to assure there is a pie at all. When funding is mandated, the range of advocacy technologies available expands significantly: there are legal, legislative, and administrative leverage points which lead interests to organize around their particular client needs and their preferred advocacy technology. However, client factors also matter: some client populations are more able to organize for collective action than others. In ABE, the primary constituency is poorly resourced and sometimes politically excluded (in the case of illegal immigrants) and thus unable to strongly advocate for its interests (though in some other states there are strong immigrant community efforts to mobilize on behalf ESL client subgroup—just not in Newstatia). In MH, clients vary by their degree of function, but most have families that have a deep interest in policy choices. The upshot is that the nature of the funding stream (discretionary versus mandated) interacts with the nature of the client population to determine the array of interest organizations that actively participate in the policy network. These differences in context ramify into the structure of the policy network through the following hypothesized causal chain: Fundamentally, these differences in policy network context are prior to dependence—indeed, dependence is actually brought into being by these features of context. Dependence and client factors elicit advocacy organizations, and collectively these help to determine network structure. Public managers in Newstatia operate in a political/policy environment defined by these factors. As with any research conducted on a highly limited number of cases, our results are not generalizable. However, our findings do suggest that future efforts to understand the role of resource dependence in structuring policy networks—and possibly other interorganizational structures for that matter, including collaborative networks—need to carefully consider the contextual factors that structure dependence. Resource dependence flows from a constitutive policy framework. In this study we have focused on one aspect of that policy framework: the policy funding context. However, other aspects of the policy framework may also matter—for instance, the degree to which funding is local, state, or federal; policy “flow-downs” from state to local or federal to state; or the degree to which policy is institutionalized through law or executive action. Networks reflect the policy funding context and the dependencies called into being by it. Our findings with respect to the role of resource dependence are only as solid as the underlying funding contexts that call dependencies into being. Assumptions and Mechanics of ERGMs In ERGMs, actors are assumed to make relationships based on what they know about the state of the network today (thus not taking into account the past) and what they would prefer in terms of their overall patterns of relationships if they are allowed to make a change to their ties. For instance, if Actor A is tied to Actor B and knows that Actor B is also tied to Actor C, Actor A may choose to make a tie to Actor C because there is an underlying preference for social relations to be “closed”—a friend of a friend is a friend. However, the tie could also be made separate from the preference for closure if A and C are of the same gender and homophily by gender is preferred. The tie becomes even more likely if both are true: there is a general preference for closure and for gender homophily. Thus, ERGMs evaluate across the entire range of proposed factors that may make ties in the dependent network more or less likely. The dependent network (here, the confidential communication network) provides the data on what actual choices were made by network members. The estimated coefficients reflect what factors were most likely to have driven relational choices given the actual relationships reported in the data. The magnitude and statistical significance of the coefficients, like in linear regression, tell us whether the proposed factors—homophily, preference for closure, etc.—appear to have been important to relational choices or not. The relational choices across all actors in network define the totality of the network. Footnotes 1 More specifically, in the ABE policy network, client advocacy organization (i.e., adult learners’ advocacy organization) was not statistically significant in the model and dropped to simplify the model. In the MH policy network, client advocacy organization was not significant at 10% level in the goodness of fit test; thus, it was not included in the model. Appendix Assumptions and Mechanics of ERGMs In ERGMs, actors are assumed to make relationships based on what they know about the state of the network today (thus not taking into account the past) and what they would prefer in terms of their overall patterns of relationships if they are allowed to make a change to their ties. For instance, if Actor A is tied to Actor B and knows that Actor B is also tied to Actor C, Actor A may choose to make a tie to Actor C because there is an underlying preference for social relations to be “closed”—a friend of a friend is a friend. However, the tie could also be made separate from the preference for closure if A and C are of the same gender and homophily by gender is preferred. The tie becomes even more likely if both are true: there is a general preference for closure and for gender homophily. Thus, ERGMs evaluate across the entire range of proposed factors that may make ties in the dependent network more or less likely. The dependent network (here, the confidential communication network) provides the data on what actual choices were made by network members. The estimated coefficients reflect what factors were most likely to have driven relational choices given the actual relationships reported in the data. The magnitude and statistical significance of the coefficients, like in linear regression, tell us whether the proposed factors—homophily, preference for closure, etc.—appear to have been important to relational choices or not. The relational choices across all actors in network define the totality of the network. References Agranoff, Robert, and Michael McGuire. 2001. Big questions in public network management research. Journal of Public Administration Research and Theory  11: 295– 326. Google Scholar CrossRef Search ADS   ———. 2004. Collaborative public management: New strategies for local governments . Washington, DC: Georgetown Univ. Press. Ansell, Chris, and Alison Gash. 2008. Collaborative governance in theory and practice. Journal of Public Administration Research and Theory  18: 543– 71. Google Scholar CrossRef Search ADS   Berry, Jeffrey M. 2005. Nonprofits and civic engagement. Public Administration Review  65: 568– 78. Google Scholar CrossRef Search ADS   Biesta, Gert. 2007. Why ‘what works’ won’t work: Evidence-based practice and the democrative deficit in educational research. Educational Theory  57: 1– 22. Google Scholar CrossRef Search ADS   Bodin, Orjan, and Beatrice Crona. 2009. The role of social networks in natural resource governance: What relational patterns make a difference? Global Environmental Change: Human and Policy Dimensions  19: 366– 74. Google Scholar CrossRef Search ADS   Boer, P., M. Huisman, Tom A. B. Snijders, C. Steglich, L. H. Y. Wichers, and E. P. H. Zeggelink. 2006. StOCNET: An open software system for the advanced statistical analysis of social networks . Groningen, The Netherlands: ICS/SciencePlus Group, Univ. of Groningen. Bonacich, Phillip, and Elisa Jayne Bienenstock. 2009. Structural implications of reciprocal exchange: A power-dependence approach. Social Psychology Quarterly  72: 226– 40. Google Scholar CrossRef Search ADS   Borgatti, S. P. 2002. NetDraw: Graph visualization software . Harvard: Analytic Technologies. Borgatti, S. P., M. G. Everett, and L. C. Freeman. 2002. UCINET 6 for Windows: Software for social network analysis . Harvard: Analytic Technologies. Bowen, T. Renee, Ying Chen, and Hülya Eraslan. 2014. Mandatory versus discretionary spending: The status quo effect. American Economic Review  104: 2941– 74. Google Scholar CrossRef Search ADS   Bryson, John M., Barbara C. Crosby, and Melissa Middleton Stone. 2006. The design and implementation of cross-sector collaborations: Propositions from the literature. Public Administration Review  66( Suppl 1): 44– 55. Google Scholar CrossRef Search ADS   Casciaro, Tiziana, and Mikolaj Jan Piskorski. 2005. Power imbalance, mutual dependence, and constraint absorption : A closer look at resource dependence theory. Administrative Science Quarterly  50: 167– 99. Google Scholar CrossRef Search ADS   Chetkovich, Carol A., and Frances Kunreuther. 2006. From the ground up: Grassroots organizations making social change . Ithaca, NY: Cornell Univ. Press. Child, Curtis D., and Kirsten A. Grønbjerg. 2007. Nonprofit advocacy organizations: Their characteristics and activities. Social Science Quarterly  88: 259– 81. Google Scholar CrossRef Search ADS   Cho, Chung Lae, and Deil S. Wright. 2004. The devolution revolution in intergovernmental relations in the 1990s: Changes in cooperative and coercive state-national relations as perceived by state administrators. Journal of Public Administration Research and Theory  14: 447– 68. Google Scholar CrossRef Search ADS   Eisenhardt, Kathleen M., and Melissa E. Graebner. 2007. Theory building from cases: Opportunities and challenges. Academy of Management Journal  50: 25– 32. Google Scholar CrossRef Search ADS   Emerson, Richard. 1962. Power-dependence relations. American Sociological Review  27: 31– 41. Google Scholar CrossRef Search ADS   Emerson, Kirk, Tina Nabatchi, and Stephen Balogh. 2012. An integrative framework for collaborative governance. Journal of Public Administration Research Theory  22: 1– 29. Google Scholar CrossRef Search ADS   Frank, O., and David Strauss. 1986. Markov graphs. Journal of the American Statistical Association  81: 832– 42. Google Scholar CrossRef Search ADS   Funk, Michelle, Alberto Minoletti, Natalie Drew, Jacob Taylor, and Benedetto Saraceno. 2006. Advocacy for mental health: Roles for consumer and family organizations and governments. Health Promotion International  21: 70– 75. Google Scholar CrossRef Search ADS   Gais, Thomas L., and Jack L. Walker. 1991. Pathways to influence in American politics. In Mobilizing interest groups in America , ed. Jack L. Walker, 103– 21. Ann Arbor: Univ. of Michigan Press. Graddy, E. A., and B. Chen. 2006. Influences on the size and scope of networks for social service delivery. Journal of Public Administration Research and Theory  16: 533– 52. Google Scholar CrossRef Search ADS   Gulati, Ranjay, and Harbir Singh. 1998. The architecture of cooperation: Managing coordination costs and appropriation concerns in strategic alliances. Administrative Science Quarterly  41: 781– 814. Google Scholar CrossRef Search ADS   Hatmaker, Deneen M., and R. Karl Rethemeyer. 2008. Mobile trust, enacted relationships: Social capital in a state-level policy network. International Public Management Journal  11: 426– 62. Google Scholar CrossRef Search ADS   Hillman, A. J., M. C. Withers, and B. J. Collins. 2009. Resource dependence theory: A review. Journal of Management  35: 1404– 27. Google Scholar CrossRef Search ADS   Huang, Kun, and Keith G. Provan. 2007. Resource tangibility and patterns of interaction in a publicly funded health and human services network. Journal of Public Administration Research and Theory  17: 435– 54. Google Scholar CrossRef Search ADS   Ingold, Karin. 2011. Network structures within policy processes: Coalitions, power, and brokerage in Swiss climate policy. Policy Studies Journal  39: 435– 59. Google Scholar CrossRef Search ADS   Ingold, Karin, and Philip Leifeld. 2016. Structural and institutional determinants of influence reputation: A comparison of collaborative and adversarial policy networks in decision making and implementation. Journal of Public Administration Research and Theory  26: 1– 18. Google Scholar CrossRef Search ADS   Isett, Kimberley R., Ines A. Mergel, Kelly Leroux, Pamela A. Mischen, and R. Karl Rethemeyer. 2011. Networks in public administration scholarship: Understanding where we are and where we need to go. Journal of Public Administration Research and Theory  21( Suppl 1): i157– i173. Google Scholar CrossRef Search ADS   Jack L. Walker. 1991. Mobilizing interest groups in America: Patrons, professions, and social movements . Ann Arbor, MI: University of Michigan Press. Johnston, J. M., and B. S. Romzek. 2008. Social welfare contracts as networks: The impact of network stability on management and performance. Administration & Society  40: 115– 46. Google Scholar CrossRef Search ADS   Kenis, Patrick and Raab, Jörg. 2007. Taking stock of policy networks: Do they matter ? In Handbook of public policy analysis: Theory, politics and methods , ed. M. S. Sidney, F. Fisher, and G. J. Miller, 187– 200, Boca Raton, FL: CRC Press. Kettl, D. F. 1996. Governing at the millennium. In Handbook of public administration , ed. J. L. Perry, 5– 18. San Francisco: Jossey-Bass. Klijn, Erik-Hans, and Chris Skelcher. 2007. Democracy and governance networks: Compatible or not. Public Administration  85: 587– 608. Google Scholar CrossRef Search ADS   Knoke, David. 1990. Political networks: The structural perspective . Cambridge, UK: Cambridge Univ. Press. Google Scholar CrossRef Search ADS   ———. 2014. Policy networks. In The sage handbook of social network analysis , ed. John Scott and Peter J. Carrington, 210– 22. Thousand Oaks, CA: Sage Publications. Google Scholar CrossRef Search ADS   Knoke, David, F. U. Pappi, H. Broadbent, and Y. Tsujinaka. 1996. Comparing policy networks: Labor politics in U.S., Germany, and Japan . Cambridge, UK/New York: Cambridge Univ. Press. Google Scholar CrossRef Search ADS   Laumann, Edward, and David Knoke. 1987. The organizational state: Social choice in national policy domains . Madison, WI: Univ. of Wisconsin Press. Laumann, Edward, Peter V. Marsden, and David Prensky. 1989. The boundary specification problem in network analysis. In Research methods in social network analysis , ed. Linton C. Freeman, Douglas R. White, and A. Kimball Romney, 61– 87. Fairfax, VA: George Mason Univ. Press. Milward, H. Brinton, and Keith G. Provan. 2000. Governing the Hollow State. Journal of Public Administration Research and Theory  10: 359– 80. Google Scholar CrossRef Search ADS   Milward, H. Brinton, Keith G. Provan, and Barbara A. Else. 1993. What does the hollow state look like. In Public management: The state of the art , ed. B. Bozeman, 309– 32. San Francisco: Jossey-Bass. Mosley, Jennifer E. 2012. Keeping the lights on: How government funding concerns drive the advocacy agendas of nonprofit homeless service providers. Journal of Public Administration Research and Theory  22: 841– 66. Google Scholar CrossRef Search ADS   OECD. 2012. Glossary of key terms for completing the 2012 OECD Budgeting Practices and Procedures Survey. http://www.oecd.org/gov/budgeting/Budgeting-survey-glossary.pdf. (accessed March 22, 2018) Oliver, Christine. 1990. Determinants of interorganizational relationships: Integration and future directions. Academy of Management Review  15: 241– 65. ———. 1991. Strategic responses to institutional processes. Academy of Management Review  16: 145– 79. CrossRef Search ADS   Park, Hyun Hee, and R. Karl Rethemeyer. 2014. The politics of connections: Assessing the determinants of social structure in policy networks. Journal of Public Administration Research and Theory  24: 349– 79. Google Scholar CrossRef Search ADS   Pattison, Philippa, and Garry Robins. 2002. Neighborhood-based models for social networks. Sociological Methodology  32: 301– 37. Google Scholar CrossRef Search ADS   Pattison, Philippa, and S. Wasserman. 1999. Logit models and logistic regressions for social networks: II. Multivariate relations. British Journal of Mathematical and Statistical Psychology  52: 169– 93. Google Scholar CrossRef Search ADS   Pfeffer, J. 1987. A Resource dependence perspective on intercorporate relations. In Intercorporate relations: The structural analysis of business , ed. M. S. Mizruchi and M. Schwartz, 22– 55. New York, NY: Cambridge Univ. Press. Google Scholar CrossRef Search ADS   Pfeffer, J., and G. Salancik. 1978. The external perspective on organizations: A resource dependence perspective . New York, NY: Harper & Row. Powell, Walter W. 1990. Neither market nor hierarchy: Network forms of organizations. In Research in organizational behavior , ed. Barry M. Staw and Larry L. Cummings, 295– 336. Greenwich, CT: JAI Press. Powell, Walter W., Kenneth W. Koput, and Laurel Smith-Doerr. 1996. Interorganizational collaboration and the locus of innovation: Networks of learning in biotechnology. Administrative Science Quarterly  41: 116– 46. Google Scholar CrossRef Search ADS   Provan, Keith G., Janice M. Beyer, and Carlos Kruytbosch. 1980. Environmental linkages and power in resource-dependence relations between organizations. Administrative Science Quarterly  25: 200– 25. Google Scholar CrossRef Search ADS   Provan, Keith G., and H. Brinton Milward. 1995. A preliminary theory of interorganizational network effectiveness: A comparative study of four community mental health systems. Administrative Science Quarterly  40: 1– 33. Google Scholar CrossRef Search ADS   Raab, Jörg. 2002. Where do policy networks come from? Journal of Public Administration Research and Theory  12: 581– 622. Google Scholar CrossRef Search ADS   Rethemeyer, R. Karl. 2007. Policymaking in the age of internet: Is the internet tending to make policy networks more or less inclusive? Journal of Public Administration Research and Theory  17: 259– 84. Google Scholar CrossRef Search ADS   Rethemeyer, R. Karl, and Deneen M. Hatmaker. 2008. Network management reconsidered: An inquiry into management of network structures in public sector service provision. Journal of Public Administration Research and Theory  18: 617– 46. Google Scholar CrossRef Search ADS   Rhodes, R. A. W. 1997. Understanding governance: Policy networks, governance, reflexivity and accountability . Buckingham: Open University Press. Ripley, Ruth, and Tom A. B. Snijders. 2009. Manual for SIENA Version 4.0 . http://www.stats.ox.ac.uk/~snijders/siena/s_man401.pdf. (accessed March 22, 2018) Robins, Garry, Phillipa Pattison, Yuval Kalish, and Dean Lusher. 2007a. An introduction to exponential random graph (P*) models for social networks. Social Networks  29: 173– 91. Google Scholar CrossRef Search ADS   Robins, Garry, Philippa Pattison, and Stanley Wasserman. 1999. Logit models and logistic regressions for social networks: III. Valued relations. Psychometrika  64: 371– 94. Google Scholar CrossRef Search ADS   Robins, Garry, Tom A. B. Snijders, Peng Wang, Mark Handcock, and Philippa Pattison. 2007b. Recent developments in exponential random graph (P*) models for social networks. Social Networks  29: 192– 215. Google Scholar CrossRef Search ADS   Saidel, Judith R. 1991. Resource interdependence: The relationship between state agencies and nonprofit organizations. Public Administration Review  51: 543– 53. Google Scholar CrossRef Search ADS   Salamon, L. M. 1981. Rethinking public management: Third-party government and the changing forms of government action. Public Policy  29: 255– 75. ———. 1987. Of market failure, voluntary failure, and third-party government: Toward a theory of government-nonprofit relations in the modern welfare state. Nonprofit and Voluntary Sector Quarterly  16: 29– 49. ———. 1995. Partners in public service: Government-nonprofit relations in the modern welfare state . Baltimore, MD: Johns Hopkins Univ. Press. Scheppele, K. L., and Jack L. Walker. 1991. The litigation strategies of interest groups. In Mobilizing interest groups in America , ed. Jack L. Walker, 335– 72. Ann Arbor, MI: Univ. of Michigan Press. Smith, S. R., and M. Lipsky. 1993. Nonprofits for hire: The welfare state in the age of contracting . Cambridge, MA: Harvard Univ. Press. Snijders, Tom A. B. 2001. The statistical evaluation of social network dynamics. Sociological Methodology  31: 361– 395. Google Scholar CrossRef Search ADS   ———. 2002a. Markov chain Monte Carlo estimation of exponential random graph models. Journal of Social Structure  3: 40. ———. 2002b. The SIENA Homepage (Simulation Investigation for Empirical Network Analysis) . http://www.stats.ox.ac.uk/~snijders/siena/. (accessed March 22, 2018) Snijders, Tom A. B., and Marijtje A. J. Duijn. 2002. Conditional maximum likelihood estimation under various specifications of exponential random graph models. In Contributions to social network analysis, information theory, and other topics in statistics: A festschrift in honour of ove frank , ed. J. Hagberg, 117– 34. Stockholm, Sweden: University of Stockholm, Department of Statistics. Snijders, Tom A. B., P. E. Pattison, G. L. Robins, and M. S. Handcock. 2006. New specifications for exponential random graph models. Sociological Methodology  36: 99– 153. Google Scholar CrossRef Search ADS   Sørensen, Eva, and Jacob Torfing. 2005. The democratic anchorage of governance networks. Scandinavian Political Studies  28: 195– 218. Google Scholar CrossRef Search ADS   US Congress Office of Technology Assessment. 1993. Adult literacy and new technologies: Tools for a lifetime . Washington, DC: US Government Printing Office. US Department of Health and Human Services. 1999. Mental health: A report of the surgeon general . Rockville, MD: US Department of Health and Human Services, Substance Abuse and Mental Health Services Administration. Walker, Jack L. 1991. Mobilizing interest groups in America: Patrons, professions, and social movements . Ann Arbor, MI: Univ. of Michigan Press. Google Scholar CrossRef Search ADS   Westmoreland, Timothy M., and Kathryn R. Watson. 2006. Redeeming hollow promises: The case for mandatory spending on health care for American Indians and Alaska natives. American Journal of Public Health  96: 600– 605. Google Scholar CrossRef Search ADS   Woolf, Steven H. 2009. Social policy as health policy. Journal of the American Medical Association  301: 1166– 69. Google Scholar CrossRef Search ADS   © The Author(s) 2018. Published by Oxford University Press on behalf of the Public Management Research Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Public Administration Research and Theory Oxford University Press

How Does Policy Funding Context Matter to Networks? Resource Dependence, Advocacy Mobilization, and Network Structures

Loading next page...
 
/lp/ou_press/how-does-policy-funding-context-matter-to-networks-resource-dependence-fCB9Zixfju
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the Public Management Research Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
1053-1858
eISSN
1477-9803
D.O.I.
10.1093/jopart/muy016
Publisher site
See Article on Publisher Site

Abstract

Abstract This study explores how policy funding context—defined as whether funding for a social service policy domain is discretionary or mandated—affects network structures in social service domains. We present comparative findings from two social service policy networks which differ with respect to funding context: A 47-actor adult basic education policy network that is funded discretionarily and a 40-actor mental health policy network where spending is mandated. Both are located in a US state we pseudonymed “Newstatia.” Using an exponential random graph model, we found that policy funding contexts affect how the locus of resource dependence interacts with the nature of client groups to determine the array of interest organizations engaged in the networks, which leads to differentials in network structure across these domains. We suggest that policy funding contexts are before resource dependence and client factors when explaining network structure. This opens space for reconsideration of the causal claims between policy funding contexts, resource dependence, advocacy mobilization, and network structures. Introduction Network theory and social network analysis have become primary tools used to analyze the multisector, multi-entity processes that lead to policy decisions, both grand and small. These processes go by different monikers, including policy networks (Laumann and Knoke 1987; Kenis and Raab 2007), governance networks (Klijn and Skelcher 2007; Rhodes 1997; Sørensen and Torfing 2005), collaborative governance regimes (Ansell and Gash 2008; Emerson, Nabatchi, and Balogh 2012), cross-sector collaborations (Bryson, Crosby, and Stone 2006), and intergovernmental cooperative arrangements (Agranoff and McGuire 2004). These studies have evolved into three primary streams: policy networks, collaborative networks, and governance networks (Isett et al. 2011). Policy networks focus on policy decision-making processes more than policy implementation processes, whereas collaborative networks emphasize policy implementation more than policy decision making. Work on governance networks considers both decision making and policy implementation jointly (Rethemeyer and Hatmaker 2008). Despite these conceptual differences, all three share a common assumption: policy stakeholders see decision making and implementation processes as a web of strong and weak connections between and among primary organizations with an interest in policy outcomes (Knoke 1990). The iterative exchange of material and social resources among primary organizations generates self-organizing, complex network structures for functional coordination, knowledge sharing, and/or advocacy mobilization. Understanding these interorganizational network structures provides insights into (1) where power resides in networks and how the distribution of power shapes the perceptions, motives, and actions of organizations involved (Knoke 1990) and (2) why and how certain policy actions or agenda items are (or are not) collectively debated, negotiated, designed, and implemented (Bodin and Crona 2009; Ingold 2011; Ingold and Leifeld 2016). However, uncovering latent deep structures (i.e., identifying factors that govern how networks are formed and take shape) is often ignored or obscured by our customary focus on the interactions that facilitate the work of a network but which are themselves governed by broader contextual factors (Knoke 1990). Little is known about how networks become structures for purposes of advocacy while interacting with broader policy funding contexts. Interactions and exchanges are the actions that define a network and bring it into being. But what helps to shape the pattern of interaction—that is, the structure of the network? Much of the literature uses resource dependence theory (RDT) (Pfeffer 1987; Pfeffer and Salancik 1978) to explain the structure of networks (Park and Rethemeyer 2014). For example, organizations participate in networks to access other organizations’ material and social resources, such as funding, regulatory assistance, knowledge, obligations, or legitimacy (Huang and Provan 2007; Oliver 1990, 1991). Constituent organizations in a network build structures through these multilateral and complex interdependencies (Rethemeyer and Hatmaker 2008). In particular, many empirical analyses have focused on the structure of networks in social service policy domains due to the increasing privatization of local, state, and federal provision of public goods and services (Agranoff and McGuire 2001; Graddy and Chen 2006; Huang and Provan 2007; Johnston and Romzek 2008; Knoke 1990; Milward and Provan 2000; Provan and Milward 1995). This work has used RDT as a primary theoretical framework to explain why organizations relate and interact and how “social structural resources” that arise from network configurations are deployed by actors in these networks. However, previous studies have not paid enough attention to variations in network structure that stems, we believe, from differing policy funding contexts. The current conceptual and empirical research on policy networks has not provided a sufficient basis for understanding the interplay between funding environments and network structures. We examine whether the nature of the funding stream itself biases the structure of policy networks. That is, are there systematic differences between service domains that are funded discretionarily by yearly legislative action and those where expenditures are mandated by law and thus subject to legal challenge and executive agency processes as well as legislative allocations. The two funding contexts that we focus on—discretionary versus mandated—are the primary divide in how public resources in social service domains are distributed in the United States. By shifting our attention to the policy funding contexts that define resource flows for entire policy networks and away from the resources available for the survival of individual policy actors in the network, we aim to draw inferences about policy network structures generically using social network analysis. Specifically, we claim that policy funding context affects (1) how the resource allocation mechanism interacts with (2) client factors to determine (3) interest groups’ advocacy involvement and thus (4) the choice of “advocacy technologies” network participants’ use, which then leads to differences in network structures across social service policy domains. This study will improve our understanding of the structure of policy networks by identifying how policy funding contexts alter the locus of resource dependence and shape political mobilization. This study proceeds in four sections. The first discusses why social service funding contexts help to determine resource dependence and mobilization processes within a network. The second reviews our cases and research methods. The third section presents findings and propositions from our comparative studies of (1) a mental health policy network and (2) an adult basic education policy network collected in 2001 and 2005, respectively, in a state we have pseudonymed “Newstatia.” Focusing on differences in the policy funding contexts in which these two policy networks are embedded, we present a mechanism that explains the formation of policy network structure: a causal path between policy funding context, the locus of resource dependence, advocacy mobilization, and network structure. The last section concludes with a discussion of study limitations and implication for future research. Theoretical Discussion Remaining Puzzles Regarding Resource Dependence, Political Mobilization, and Networks in Social Service Policy Domains Previous work on policy networks has discussed the centrality of resources as an environmental factor that shapes the political behavior of organizations involved in policymaking. In these models, policy decisions are a social product that results from the interplay between (1) the level of resources and (2) the political dynamics between resource holders who distribute public funds (e.g., legislators, legislative committees, or state agencies) and resource seekers (e.g., private or nonprofit service providers, advocacy groups) who must acquire resources to provide public goods and services and (often but not always) to survive (c.f., Milward and Provan 2000; Mosley 2012; Saidel 1991; Salamon 1987; Smith and Lipsky 1993; Park and Rethemeyer 2014). Resource interdependence weaves organizations together, and a complicated and persistent set of relational configurations emerge as a policy network. In particular, some policy actors may gain “social structural resources” (SSRs) from their informal roles and structural positions that are mutually agreed upon by members of the network. SSRs differ from “material institutional resources” (MIRs) in that they grant certain actors influential positions from which to mobilize resources from other members—particularly preferential access to actors with state authority. Previous work suggests that SSRs endure over time unless the relational configuration of a policy network substantially changes (Hatmaker and Rethemeyer 2008). However, previous studies have not sufficiently discussed the question of funding mechanisms in explaining what causes variation in the structure of policy networks across social service areas. The missing factor is policy funding context. Although resource exchanges (money, personnel, etc.) are typically voluntary, governmental mandates and their funding flows impose and enforce certain interorganizational connections and even organizational forms (Knoke 2014). The critical issue is that the locus of resource dependence in a network can be shaped by the nature of the policy funding stream. If all funding is discretionary, then the locus of resource dependence resides in the legislative and executive processes that allocate the resources. Here, legislators, legislative committees, and executive officials define the political landscape. These actors seek reliable information to evaluate policy proposals and decide funding allocations (Knoke 2011; Park and Rethemeyer 2014). Alternatively, if funds are allocated by mandate, new avenues for policy change open and thus the locus of resource dependence shifts. One could advocate for new mandates in the legislature; one could advocate for new interpretations of existing mandates in executive agencies; or one could seek legal redress in the courts (administrative and civil). Similarly, policy funding context also helps to shape how advocacy organizations mobilize their resources in networks by determining which organizations and advocacy tactics are considered legitimate. At the most basic level, policy funding context may make traditional interest-based advocacy more or less useful and thus determine whether advocacy organizations form at all. If advocacy organizations do form, policy funding context can also help to determine what tactics such organizations use and which spectrum of policymakers the advocacy tactic is usually directed toward. If funding levels are discretionary, then legislative tactics are clearly useful as legislators and legislative committees have a greater role in the allocation of resources to social services. If funding levels are mandatory, then legal and administrative tactics may come to the fore. Therefore, we focus on the following questions in an effort to explore factors affecting policy network structures: How do differences in policy funding context (discretionary versus mandated) shape the locus of the resource dependence between organizations operating in a given social service domain? How do differences in policy funding contexts influence advocacy mobilization processes in a network? Discretionary Versus Mandated Social Service Policy Funding Contexts Under the banner of “the welfare state,” government makes many promises to socially vulnerable people with respect to their living standards and social well-being. In current federal, state, and local policy decision making and implementation processes, social service policy is generally executed through either discretionary or mandatory budget allocations. Discretionary spending refers to “expenditure that is governed by annual or other periodic appropriations rather than by formulas or criteria set forth in authorizing legislation” (OECD 2012, 3). Examples include health care spending for American Indians and Alaska Natives by the Federal government; early childhood education, health, and nutrition programs for low-income children and families; Head Start; scientific research through the National Institutes of Health and National Science Foundation; and food assistance for Women, Infants, and Children. Specifically, when a social service program is authorized on a multi-year basis with general policy guidelines and maximum spending, funds are annually appropriated by the legislature. Legislators may choose to appropriate amounts that range from zero to the maximum of the authorization. Thus, advocates for a given social service program must, at least annually, seek political support from lawmakers for continuation of the program, or the programs will die. Since there are often no alternative markets for the goods and services that the social service providers produce, providers often fail without annual state contracts or grants (Park and Rethemeyer 2014). In this sense, we expect that resource seekers in a social service area where state resources are distributed discretionarily will actively engage in legislative political action and coalition building (e.g., organizing an industry association) to secure the resources necessary for survival. Legislative advocacy, such as writing letters, sending emails, or making phone calls to members or creating coalitions to collectively lobby for legislation, is often required to maintain state effort when services are discretionarily funded. Additionally, policy research can become an advocacy tool for nonprofits and other policy stakeholders—but a tool that can cut both ways. For organizations that seek change, policy research can be an aggressive advocacy tactic, as it can embarrass legislators and executive agency leaders by demonstrating that current policies and programs are ineffective, inefficient, or both (Berry 2005; Knoke 2011). However, supportive policy research can help to cement political relationships with key legislators and agencies heads by providing compelling evidence that policy is working and that agencies are effective. Indeed, there is some evidence that in unstable funding environments, legislative, and executive decision makers look favorably on supportive advocacy activities such as policy research (Gais and Walker 1991). By contrast, mandatory spending refers to expenditures administered through enacted law. In advance of annual appropriations, the law guarantees necessary money for mandated programs (Westmoreland and Watson 2006). If there are no explicit changes in formulas or criteria written into law, the previous year’s budget bill is applied to the current year (Bowen, Chen, and Eraslan 2014). Mandatory spending is generally characterized as either open-ended or capped. For example, Medicaid and Medicare are open-ended mandatory spending; thus, the federal government must provide guaranteed health services to targeted individuals. The State Children’s Health Insurance Program is an example of capped mandatory spending. In many cases, growth in mandatory spending automatically occurs without legislative intervention (Westmoreland and Watson 2006). Thus, resource seekers in a social service area where state resources are distributed via mandatory allocations may not have strong incentives to engage in legislative politics. Previous work has already demonstrated that once mandatory programs come into being, the organizational landscape around a policy begins to change. Organizations develop that take as their mission defense of the mandatory service. For instance, Gais and Walker (1991) chronicle how a number of occupational associations (for health care providers, social workers, and social service professionals) were actually started with public funds to help support mandatory programs. Once created, the leaders of these groups remained in close contact with administrative agencies that implemented these mandatory social services (Gais and Walker 1991). As interest groups based upon occupational or commercial communities came to support themselves through membership dues or private foundations, they increasingly employed “outside” strategies of political influence, such as mobilizing public opinion through the media, instead of overt “inside” strategies such as legislative advocacy (Gais and Walker 1991). Additionally, a different set of advocacy tools become available in a mandatory environment. Most importantly, judicial processes to assure steady streams of funding (Scheppele and Walker 1991) become viable. Once mandated by law, legal client advocacy organizations may use the courts to (1) provide explicit legal remedies for unfavorable government action with respect to individuals as well as served populations and (2) pursue their policy agendas concerning the passage of favorable legislation and the expansion of public services (Scheppele and Walker 1991). The existing literature on discretionary and mandatory policy contexts provides ample clues that the organizational environment and advocacy strategies that interested parties use differ depending on whether the social service is mandatory or discretionary. Yet the existing work on policy networks does not account for these differences when examining the determinants of network structure. We now turn to filling in that gap. Study Background To examine the relationship between policy funding context and network structure in social service domains, we examine two extreme cases: one where the services provided are funded discretionarily and another where services are largely (though not in all cases) funded through mandates. The discretionary case focuses on policy related to adult basic education. This area of social service policy is characterized by relatively small budgets and low levels of salience beyond provider and user constituencies, making it difficult to receive media attention. Advocacy organizations are more difficult to organize because the user community is generally very disadvantaged. Adult learners tend to be socioeconomically disadvantaged, and English as a Second Language (ESL) communities are further inhibited from political participation through lack of legal or political standing since many service users are immigrants, both legal and illegal. Advocacy organizations tend to focus on government resource flows because it is a discretionary program (Park and Rethemeyer 2014). Policy advocacy is often carried out through research publications for the legitimization of policy (Biesta 2007) by highlighting problems and the return to public investment in literacy. However, educational service providers also engage in more traditional forms of political mobilization by engaging affected communities in letter-writing campaigns, by directly mobilizing their clients, and through efforts to build relationships with the major legislative decision makers. The second case is mandatory in nature and focuses on policymaking for the severely mentally ill. Mental health care services are mostly mandatory at the state level. The policy funding context for mental health care is similar to that of general health policymaking. Health policymaking is “highly politicized, with significant lobbying by pharmaceutical and medical technology industries as well as by health providers and consumer groups” (Child and Grønbjerg 2007, 262). Health policy is controversial and receives extensive media attention (Woolf 2009). Mental health policymaking is characterized by multiple advocacy interests because (1) there are multiple constituencies (e.g., patients, patients’ families, employers, insurance companies, hospitals, pharmaceutical and medical companies, outpatient service providers, and taxpayers), (2) the policy area is heavily regulated, and (3) the funding levels are very substantial (Child and Grønbjerg 2007; Woolf 2009). A previous study suggested that advocacy organizations participate more heavily in health-related policy than in policymaking on education, public benefits, or religion (Child and Grønbjerg 2007). In particular, policy advocacy in health policy is often carried out through litigation. Since diverse constituencies can be mobilized around rights and winning in court secures a more permanent victory, litigation can be an attractive option for advocacy organizations to support public health policy (Scheppele and Walker 1991). Therefore, we expect that social service policy networks in areas of mandated spending are more likely to attract participation by legal client advocacy organizations than when spending is discretionary. Methods Data Collection and Data Sets This study draws on two policy networks collected from (1) a mental health policy network and (2) an adult basic education policy network. The mental health policy network data (40 actors; 37 respondents) was collected in 2001; the adult basic education policy network data (47 actors; 41 respondents) was collected in 2005. Both data were collected using semi-structured interviews in the same state eastern US state (Newstatia) that has more than 10 cities with at least 50,000 inhabitants. Newstatia is a compelling research site for social policy networks as there is both a great deal of demand and large expenditures in both areas of social service. Newstatia has served adult basic learners and the mentally ill for more than a century. At the time of the study, Newstatia ranked in the top quintile in immigrant population per capita, per-learner expenditures, and total expenditures on adult basic education. Newstatia also ranked in the second-highest quintile regarding severely mentally ill patients served and in the top quintile in terms of total and per capita state spending on the severely mentally ill. ABE policy is defined as those decisions that affect the funding or regulation of organizations that provide educational services to individuals 16 years of age or older who are seeking to raise their reading, writing, and/or computational skills to a level closer to that of a high-school graduate (US Congress Office of Technology Assessment 1993). Mental health policy is defined as decisions that affect the quality and quantity of services available from public and private sector sources to children and adults who have severe mental disorders that interfere with some area of social functioning (US Department of Health and Human Services 1999). Policy Funding Contexts and Network Selection While the data for this study was originally collected for other purposes, the networks studied here conform to a theoretical sampling strategy (Eisenhardt and Graebner 2007) to understand how network structures differ by policy funding contexts. The key differentiator is the nature of the funding streams in these social policy domains: services to the severely mentally ill are legally mandated while support for low-literate adults is at the discretion of state decision makers in the legislature. Additionally, the level of resources devoted to the policy areas is substantially different: support for the severely mentally ill is measured in billions of dollars while support for low-literate adults is measured in millions of dollars. Moreover, the two cases differ with respect to resource dependence on state funding: for-profit and nonprofit ABE providers are much more dependent on state contracts and grants than are the for-profits and nonprofits that provide MH services (Rethemeyer 2007). Network Specification This study used a three-step network specification method in the “realist” tradition outlined by Laumann, Marsden, and Prensky (1989). Rather than using researcher-defined criteria (the “nominalist” approach), the realist approach uses informants to identify network members. The first step was to develop a “naïve” universe of potential policy network members through searches of the Internet, newspaper reports, and policy documents. We then asked three informants from the naïve universe who were clearly central to policymaking in these domains to vet and expand the naïve list. Next, we gave seven members of each network the master list and asked the respondents to rate each organization on a scale of 1–3 with respect to their policy influence. As a backstop to our initial specification, we asked each informant we interviewed during the main data collections if any organizations were omitted from the list. The final specifications included 40 actors in the mental health network and 47 actors in the ABE network. Our informant checks found no organization was listed as “missing” by more than 20% of the network members during the data collection, which suggested high convergence on the membership of both networks. Most of the informants were CEOs; a few were government relations specialists. Measurements Network Ties Ties within a policy network can be measured in various ways. We focus on communications that are foundational to the outputs of policy networks: policy decisions (executive, legislative, and/or judicial) that are built on shared understandings of political and policy preferences, knowledge of preference intensity and power distributions, and communal beliefs about policy cause and effect. Common understandings are the product of communication. Thus, network ties were measured by the extent to which a pair of policy actors maintain routine and/or confidential communication relationships (Knoke 1990; Laumann and Knoke 1987). These measures have been widely used to capture structural relationships that affect policy decisions, outputs, failures, and changes that cannot be explained solely by reference to formal, task-based structural design. More importantly, routine communication is used to scan political environments, while confidential communication is used both to establish meaning and to discuss distribution and allocation of resources (Laumann and Knoke 1987; Raab 2002). We presented each informant two rosters that contained the name and contact person for the organizations identified through the network specification procedure. For each roster, we asked informants to estimate the monthly frequency of the two types of communication—routine and confidential—using a zero to seven scale that was anchored to cues for the number of contacts per month. For this study, we used the data on confidential communication ties as the dependent network because confidential communication contains high-value political and policy material (Park and Rethemeyer 2014). The dependent networks were dichotomized for the ERGM analyses, with a “1” assigned to any dyad with communications frequency of at least a few times a year. This cutoff value was chosen because our qualitative data and knowledge of the domain actors suggested that actors that communicated fewer than a few times a year were not regularly engaged in policy discussions sufficiently to influence outcomes (Park and Rethemeyer 2014). Resource Dependence Resource dependence has been measured in a variety of ways (Casciaro and Piskorski 2005; Hillman, Withers, and Collins 2009; Provan, Beyer, and Kruytbosch 1980). We used both objective and subjective measures for resource dependence to capture how MIRs and SSRs may affect tie choice in a policy network. Regarding objective measures of resource dependence, we collected information on the nature of financial flows and regulation—two measures typically used in a resource-scarce environment (Cho and Wright 2004)—through the operationalization of the capacity to control others’ behavior legitimately (Park and Rethemeyer 2014; Provan, Beyer, and Kruytbosch 1980). Using a roster of network members, we asked informants to note (1) which organizations provide them funding and (2) which organizations regulate their activity. Additionally, we considered whether or not each focal actor has formal authority within the public sector conferred by an elected or appointed office or formal position in the civil service. Since formal authority is a scarce resource in highly institutionalized settings like policy networks, other actors without formal authority, such as service providers and advocacy organizations, attempt to gain access to public actors directly or through intermediaries (Knoke et al. 1996). For this reason, informants were asked to indicate whether they were (1) executive actors or (2) legislative actors. As a subjective measure for focal actor’s dependence, we used policy actors’ evaluation of each network member’s influence on decision making in the ABE and MH policy networks. Organizations seek to mold their resource environments through political activity (Emerson 1962; Hillman, Withers, and Collins 2009; Pfeffer and Salancik 1978). To change the resource environment within a policy network, actors evaluate resources—both MIRs and SSRs—that they and other players have and then select their political actions. Since power resides in and is constrained by relationships (Emerson 1962), the pattern of exchange in a network is different from that of dyadic exchange between two actors (Bonacich and Bienenstock 2009). Actors’ perceptions of the influence of others on resource allocations in a network may greatly affect their choice of relational strategies to acquire resources. It is important to include a subjective measure of dependence—perceived influence—to understand how resource dependence (and network members’ perceptions of it) differentially constrains tie selections across social policy domains. To measure influence, we asked each informant to rate the “policy influence” of each member of the roster on a scale from zero to seven. Influence was defined as follows: the ability to get others to believe, think, or act as one prefers with respect to a given policy issue. The more influence an actor has, the more likely it is that policies in line with their preferences will be implemented. To purge the resulting influence scores of raters’ tendency to anchor on high or low values, we transformed each raters’ responses into a Z-score using the mean and standard deviation of each rater’s responses. Organizational Advocacy Information To examine the differential structural participation of advocacy organizations, we first had to identify the nature of all organizations in the network. We used an existing typology from earlier work (Park and Rethemeyer 2014) but refined the advocacy categories to include (1) client advocacy organizations, (2) legal client advocacy organizations, (3) industrial/professional associations and foundations, (4) research and technical support organizations, and (5) service delivery organizations. Network Structural Variables A number of structural variables such as reciprocity, three-cycles, two-paths, and other higher-order structures (e.g., preference for higher outdegree nodes, preference for higher indegree nodes, alternating k-triangles, and alternating two-paths) are included in our model to control for generic social pressures on organizations’ relational choices. For instance, if Actor A is friends with Actor B, and Actor B is friends with Actor C, it is very rare that Actor A is not friends with Actor C because such an instance would violate transitivity (i.e., a tendency for a friend of a friend to be a friend) as a generic social pressure (Park and Rethemeyer 2014). The “microstructures” representing generic social pressures have significant effects on global network structures (Pattison and Robins 2002). Previous studies on policy network have shown that generic social processes pressure policy actors to forge ties that may not be formed otherwise (Ingold and Leifeld 2016; Park and Rethemeyer 2014). Analysis We used an exponential random graph model (ERGM) to analyze our sociometric data. ERGM’s were developed to account for dependence between dyads (Frank and Strauss 1986; Pattison and Wasserman 1999; Robins et al. 2007a; Robins, Pattison, and Wasserman 1999). Standard linear regression models of dyadic data are inherently biased due to these dependencies. Additionally, the dependencies themselves are interesting, as they represent patterns of social interaction that can be analyzed—for instance, reciprocity, the tendency for closed social structures (a friend of a friend is a friend), and so forth. ERGMs can easily accommodate relationships, attributes of an actor, and network structural variables as predictors of the dependent network—here meaning the confidential communications network (Snijders et al. 2006). ERGMs assume that a network may be represented by a random graph that follows a Markov change process. That is, the probability of a tie between any pair of actors is assumed to be random (i.e., independent) given that the rest of the network consists of certain conditionally dependent ties. A tie from actor i to actor j is conditionally dependent only on other possible ties involving i and/or j and other features that systematically make some ties more likely than others—for instance, the well-known preference for homophily in many social interactions (Robins et al. 2007a). For an introduction to stochastic social network analysis, see Robins et al. (2007b); Snijders (2001, 2002a); Snijders and Duijn (2002); for a fuller discussion of modeling procedures, see Park and Rethemeyer (2014). Appendix discusses the assumptions and mechanics of ERGMs in detail. We executed our ERGM using the Simulation Investigation for Empirical Network Analysis (also known as SIENA, see Snijders 2002b) package in StOCNET (Boer et al. 2006) that estimates parameters based on simulation techniques (Markov Chain Monte Carlo maximum likelihood estimation (MCMCMLE) models—see Robins et al. 2007b). We used NetDraw (Borgatti 2002) to create network visualizations and UCINET (Borgatti, Everett, and Freeman 2002) to calculate deterministic network statistics. ERGM models are said to adequately represent the underlying data when the features included in the model (e.g., the number of reciprocal dyads, the prevalence of ties between organizations of a similar age, etc.) may be reproduced through simulations with an average value that is statistically indistinguishable from the measured values in the data. To determine whether the model has “converged,” the average value of the modeled features from a large set of simulations (5,000 in this case) is compared to the measured value using a t-statistic. Convergence is said to occur when the t-value is very near zero—by convention, less than 0.10 for all included features. Both models reported below are fully converged by this standard. Interpretation of results in ERGM is similar to logit parameter estimates—the sign and significance may be directly interpreted, but the magnitudes may not be. However, ERGM creates more complex outputs about attributes that may affect tie formation. One must consider three effects of attributes: ego, alter, and same (or similarity). Ego effects capture whether certain attributes make an actor more “outgoing” (more likely to make connections) if the coefficient is positive and significant. Alter effects capture whether an attribute makes an actor more “popular” (more likely to be sought out by others) if the coefficient is positive and significant. “Same” effects capture homophily when the variable is binary, while “similarity” effects capture homophily when the variable is continuous. Do actors prefer others with the same (or similar) attribute if the coefficient is positive and significant? For an in-depth introduction to ERGM parameters and the interpretation of the estimates, see the SIENA Web site and manual (Ripley and Snijders 2009; Snijders 2002b). Findings The Structure of Discretionary and Mandatory Social Service Policy Networks By taking an inductive stance, our findings help us to develop a series of propositions about the general nature of policy network structure in light of differences in policy funding context. Table 3 reports the final models, including a reduced set of variables found to have significant coefficients through joint goodness-of-fit tests for both ABE and MH policy networks. Our findings indicate that differences in policy funding contexts (discretionary versus mandated) play a major role in determining the locus of the resources dependence, the nature of advocacy mobilization processes, how dependence and advocacy interact with one another, and thus network structures. Table 1. Actor Typology of Mental Health in Newstatia, US Description  Newstatia, US (2001)  State agencies  7  Legislators  8  Legislative committees  4  Industry associations  3  Professional associations/unions  3  Service providers  6  Legal client advocacy organizations  4  Client or family advocacy organizations  5  Newspapers  2  Insurers  3  Researchers/research teams  3  Description  Newstatia, US (2001)  State agencies  7  Legislators  8  Legislative committees  4  Industry associations  3  Professional associations/unions  3  Service providers  6  Legal client advocacy organizations  4  Client or family advocacy organizations  5  Newspapers  2  Insurers  3  Researchers/research teams  3  Note: Total will not add to 40; two advocacy organizations are also government agencies; one research team is also state agencies; five industry/professional associations are also service providers. View Large Table 1. Actor Typology of Mental Health in Newstatia, US Description  Newstatia, US (2001)  State agencies  7  Legislators  8  Legislative committees  4  Industry associations  3  Professional associations/unions  3  Service providers  6  Legal client advocacy organizations  4  Client or family advocacy organizations  5  Newspapers  2  Insurers  3  Researchers/research teams  3  Description  Newstatia, US (2001)  State agencies  7  Legislators  8  Legislative committees  4  Industry associations  3  Professional associations/unions  3  Service providers  6  Legal client advocacy organizations  4  Client or family advocacy organizations  5  Newspapers  2  Insurers  3  Researchers/research teams  3  Note: Total will not add to 40; two advocacy organizations are also government agencies; one research team is also state agencies; five industry/professional associations are also service providers. View Large Table 2. Actor Typology of Adult Basic Education in Newstatia, US Description  Newstatia, US (2005)  State agencies  7  Legislators  7  Legislative committees  5  Industry associations  1  State-funded service providers  12   Community-based organizations  6   Community colleges  0   Municipal agencies  3   School districts  3   Union  0  State-funded technical assistance units  9  Foundations  1  Research/advocacy organizations  3  Learner advocates  1  Consultants  1  Description  Newstatia, US (2005)  State agencies  7  Legislators  7  Legislative committees  5  Industry associations  1  State-funded service providers  12   Community-based organizations  6   Community colleges  0   Municipal agencies  3   School districts  3   Union  0  State-funded technical assistance units  9  Foundations  1  Research/advocacy organizations  3  Learner advocates  1  Consultants  1  View Large Table 2. Actor Typology of Adult Basic Education in Newstatia, US Description  Newstatia, US (2005)  State agencies  7  Legislators  7  Legislative committees  5  Industry associations  1  State-funded service providers  12   Community-based organizations  6   Community colleges  0   Municipal agencies  3   School districts  3   Union  0  State-funded technical assistance units  9  Foundations  1  Research/advocacy organizations  3  Learner advocates  1  Consultants  1  Description  Newstatia, US (2005)  State agencies  7  Legislators  7  Legislative committees  5  Industry associations  1  State-funded service providers  12   Community-based organizations  6   Community colleges  0   Municipal agencies  3   School districts  3   Union  0  State-funded technical assistance units  9  Foundations  1  Research/advocacy organizations  3  Learner advocates  1  Consultants  1  View Large Table 3. MCMCMLE of Policy Networks (ABE and MH)   Parameters  ABE  MH  Structural configuration  Reciprocity  1.3558** (0.1838)  1.1542** (0.2518)    3-cycles  0.0562 (0.0525)  −0.166* (0.0738)  2-paths  0.0126 (0.0132)    Preference for higher outdegree nodes  −0.2684 (0.3907)  −0.2877 (0.3212)  Preference for higher indegree nodes  −8.8049* (4.4680)  −1.0063* (0.4016)  Alternating k-triangles  1.2885** (0.2367)  1.1402** (0.1698)  Alternating 2-paths    −0.0664* (0.031)  Regulators  Regulators ego  −0.1717 (0.4110)    Regulators alter    −0.3289 (0.2404)  Same regulators  −0.2413 (0.2574)    Funders  Funders ego  0.8034* (0.3418)  −0.2713* (0.1331)  Same funders  0.4139* (0.2048)    Regulatory/funding dependence by public actors  Regulated and funded by public actors ego  0.8023** (0.1739)    Regulated and funded by public actors alter  −0.3396† (0.1898)    Same regulated and funded by public actors  0.2550** (0.0984)    Public actors  State actors alter  −0.5488* (0.2753)  0.4142* (0.1758)  Same state actor  0.0266 (0.1757)    Legislator and legislative committee ego  0.6218** (0.2015)    Legislator and legislative committee alter  −0.4122† (0.2255)  0.479** (0.1308)  Same legislator and legislative committee  0.6441** (0.13080  0.4818** (0.1044)  Influence  Influence ego    0.0039 (0.0037)  Influence alter  0.0124** (0.0033)  0.0187** (0.0041)  Influence similarity    1.0373** (0.3084)  NonprofitOrgs.         Legal client advocacy organizations  Legal client advocacy organizations ego    0.3388* (0.1695)   Industrial/professional associations  Industrial & professional associations and foundations ego  −0.0588 (0.2466)  0.5225* (0.2451)   Research and technical support organizations  Same research and technical supports organizations  0.4398** (0.1132)     Service providers  Service providers ego  0.0132 (0.1014)  −0.0448 (0.237)    Parameters  ABE  MH  Structural configuration  Reciprocity  1.3558** (0.1838)  1.1542** (0.2518)    3-cycles  0.0562 (0.0525)  −0.166* (0.0738)  2-paths  0.0126 (0.0132)    Preference for higher outdegree nodes  −0.2684 (0.3907)  −0.2877 (0.3212)  Preference for higher indegree nodes  −8.8049* (4.4680)  −1.0063* (0.4016)  Alternating k-triangles  1.2885** (0.2367)  1.1402** (0.1698)  Alternating 2-paths    −0.0664* (0.031)  Regulators  Regulators ego  −0.1717 (0.4110)    Regulators alter    −0.3289 (0.2404)  Same regulators  −0.2413 (0.2574)    Funders  Funders ego  0.8034* (0.3418)  −0.2713* (0.1331)  Same funders  0.4139* (0.2048)    Regulatory/funding dependence by public actors  Regulated and funded by public actors ego  0.8023** (0.1739)    Regulated and funded by public actors alter  −0.3396† (0.1898)    Same regulated and funded by public actors  0.2550** (0.0984)    Public actors  State actors alter  −0.5488* (0.2753)  0.4142* (0.1758)  Same state actor  0.0266 (0.1757)    Legislator and legislative committee ego  0.6218** (0.2015)    Legislator and legislative committee alter  −0.4122† (0.2255)  0.479** (0.1308)  Same legislator and legislative committee  0.6441** (0.13080  0.4818** (0.1044)  Influence  Influence ego    0.0039 (0.0037)  Influence alter  0.0124** (0.0033)  0.0187** (0.0041)  Influence similarity    1.0373** (0.3084)  NonprofitOrgs.         Legal client advocacy organizations  Legal client advocacy organizations ego    0.3388* (0.1695)   Industrial/professional associations  Industrial & professional associations and foundations ego  −0.0588 (0.2466)  0.5225* (0.2451)   Research and technical support organizations  Same research and technical supports organizations  0.4398** (0.1132)     Service providers  Service providers ego  0.0132 (0.1014)  −0.0448 (0.237)  †p < 0.1; *p < .05; **p < .01. View Large Table 3. MCMCMLE of Policy Networks (ABE and MH)   Parameters  ABE  MH  Structural configuration  Reciprocity  1.3558** (0.1838)  1.1542** (0.2518)    3-cycles  0.0562 (0.0525)  −0.166* (0.0738)  2-paths  0.0126 (0.0132)    Preference for higher outdegree nodes  −0.2684 (0.3907)  −0.2877 (0.3212)  Preference for higher indegree nodes  −8.8049* (4.4680)  −1.0063* (0.4016)  Alternating k-triangles  1.2885** (0.2367)  1.1402** (0.1698)  Alternating 2-paths    −0.0664* (0.031)  Regulators  Regulators ego  −0.1717 (0.4110)    Regulators alter    −0.3289 (0.2404)  Same regulators  −0.2413 (0.2574)    Funders  Funders ego  0.8034* (0.3418)  −0.2713* (0.1331)  Same funders  0.4139* (0.2048)    Regulatory/funding dependence by public actors  Regulated and funded by public actors ego  0.8023** (0.1739)    Regulated and funded by public actors alter  −0.3396† (0.1898)    Same regulated and funded by public actors  0.2550** (0.0984)    Public actors  State actors alter  −0.5488* (0.2753)  0.4142* (0.1758)  Same state actor  0.0266 (0.1757)    Legislator and legislative committee ego  0.6218** (0.2015)    Legislator and legislative committee alter  −0.4122† (0.2255)  0.479** (0.1308)  Same legislator and legislative committee  0.6441** (0.13080  0.4818** (0.1044)  Influence  Influence ego    0.0039 (0.0037)  Influence alter  0.0124** (0.0033)  0.0187** (0.0041)  Influence similarity    1.0373** (0.3084)  NonprofitOrgs.         Legal client advocacy organizations  Legal client advocacy organizations ego    0.3388* (0.1695)   Industrial/professional associations  Industrial & professional associations and foundations ego  −0.0588 (0.2466)  0.5225* (0.2451)   Research and technical support organizations  Same research and technical supports organizations  0.4398** (0.1132)     Service providers  Service providers ego  0.0132 (0.1014)  −0.0448 (0.237)    Parameters  ABE  MH  Structural configuration  Reciprocity  1.3558** (0.1838)  1.1542** (0.2518)    3-cycles  0.0562 (0.0525)  −0.166* (0.0738)  2-paths  0.0126 (0.0132)    Preference for higher outdegree nodes  −0.2684 (0.3907)  −0.2877 (0.3212)  Preference for higher indegree nodes  −8.8049* (4.4680)  −1.0063* (0.4016)  Alternating k-triangles  1.2885** (0.2367)  1.1402** (0.1698)  Alternating 2-paths    −0.0664* (0.031)  Regulators  Regulators ego  −0.1717 (0.4110)    Regulators alter    −0.3289 (0.2404)  Same regulators  −0.2413 (0.2574)    Funders  Funders ego  0.8034* (0.3418)  −0.2713* (0.1331)  Same funders  0.4139* (0.2048)    Regulatory/funding dependence by public actors  Regulated and funded by public actors ego  0.8023** (0.1739)    Regulated and funded by public actors alter  −0.3396† (0.1898)    Same regulated and funded by public actors  0.2550** (0.0984)    Public actors  State actors alter  −0.5488* (0.2753)  0.4142* (0.1758)  Same state actor  0.0266 (0.1757)    Legislator and legislative committee ego  0.6218** (0.2015)    Legislator and legislative committee alter  −0.4122† (0.2255)  0.479** (0.1308)  Same legislator and legislative committee  0.6441** (0.13080  0.4818** (0.1044)  Influence  Influence ego    0.0039 (0.0037)  Influence alter  0.0124** (0.0033)  0.0187** (0.0041)  Influence similarity    1.0373** (0.3084)  NonprofitOrgs.         Legal client advocacy organizations  Legal client advocacy organizations ego    0.3388* (0.1695)   Industrial/professional associations  Industrial & professional associations and foundations ego  −0.0588 (0.2466)  0.5225* (0.2451)   Research and technical support organizations  Same research and technical supports organizations  0.4398** (0.1132)     Service providers  Service providers ego  0.0132 (0.1014)  −0.0448 (0.237)  †p < 0.1; *p < .05; **p < .01. View Large Finding 1: In both ABE and MH policy networks, regulators (who have oversight of policy actors in the network) are not important in structuring policy connections. Finding 2: In both ABE and MH policy networks, funders (who provide funding to policy actors in the network) are important in structuring policy connections. Regarding resource holders’ dynamics, both the ABE and MH networks are affected by funders’ network formation activities—but in different ways—while neither network is structured by relationships built to or from regulators. Turning first to the findings regarding regulation, regulators do not play a role in structuring the network. For both the ABE and MH networks, the coefficients in the Regulators section of the table are statistically insignificant. Regulators do not make connections at a higher rate than other actors in both networks; they are not sought as communications partners in either network; and they do not tend to communicate with each other differentially. While regulators hold a key resource for service providers—authorization to engage in a practice or service—that resource is necessary but not sufficient. Providers can often find ways to satisfy regulatory obligations, but they cease to exist without resources. Policy networks exist primarily to seek material resources, not authorization. Thus, regulators do not play a significant part in creating the structure. However, the structure of both networks is affected by funders, but the ABE network is far more thoroughly affected by relationships driven by the imperative of funding. In most social service policy domains, state funding is an indispensable resource that helps service providers assist socially disadvantaged people (Park and Rethemeyer 2014). Thus state funders are imperative. Previous studies have demonstrated that financial resource dependence is an important shaper of social policy network structure (Park and Rethemeyer 2014). We also find that both networks are affected by funding relationships, but quite differently. Specifically, funding flows (i.e., who funds whom) are a strong predictor of the creation of confidential policy communication ties within the discretionary ABE policy network. Funders ego and Same Funders in ABE are positive and significant at the 5% level. Funders in the ABE network actively shape policy communication, and they have a strong tendency to talk about sensitive political information with each other. Resource holders such as the State Department of Education and the industry association proactively gather reliable tacit, technical, or proprietary information through personal contact in order to make better decisions about resource allocation (Gulati and Singh 1998; Powell 1990; Powell, Koput, and Smith-Doerr 1996). Moreover, because these funding streams are discretionary, the process of collecting data to guide funding must be continuous, intensive, and timed to the legislative calendar. However, in the mandated MH network, our results suggest a different story: MH funders are significant policy actors in structuring the network, but they are not active in the creation of ties with other policy actors (see negative and significant coefficient on Funders ego in MH in table 3) even though there are very substantial funding relationships in the MH network. In fact, funders in the MH network tend to be less likely to create ties to other network members. They do not engage in network formation actively. We will discuss how these findings relate to differences in policy funding context below. Finding 3: In the ABE policy network, financial and regulatory resource seekers (who are given funding and regulated by public sector actors in the network) are important in structuring policy connections. Resource seekers are active and homophilous in forming ties but are less popular as partners in the network. Resource holders and resource seekers make contacts in a policy network differently. Specifically, resource seekers’ relational choices are affected by their evaluation of their dependence on resource holders (Park and Rethemeyer 2014). We examined how relational behaviors of resource seekers differ by policy funding context and found that funding and regulatory dependence is significant in structuring policy connections only within the discretionary ABE policy network. By contrast, even though the resource dependence related variables were included in the model, these variables are not significant predictors of connections within the mandated MH network. The coefficients in table 3 on Regulated & funded by public actors ego and Same funded & regulated by public actors are positive and significant at the 1% level. Also, the coefficient on Funded & regulated by public actors alter is negative and significant at 10% level. In the ABE network, the funding and regulatory dependence relationships are highly correlated; thus, we included only one of these relationships in the model and interpreted the results from a combined funding/regulation dependence perspective. Our funding/regulatory dependence variables are “1” if policy actors are regulated/funded by a public actor within the networks. Thus, policy actors regulated and funded by public actors tend to create ties with others in the ABE policy network. At the same time, they are strongly homophilous: those organizations that are subject to regulation and funding dependence tend to communicate with one another. However, not surprisingly, they are not popular within the ABE network because they do not hold the key financial resources in the network. Our findings and the ABE interview data suggest that those organizations funded and regulated by public actors (mainly ABE service providers and technical assistance units) actively responded to the post-9/11 economic crash by extending their brokerage positions between the provider community and legislative actors (Park and Rethemeyer 2014). They realized that they may not be able to “live off the state” in a hostile budget environment. They were maintaining relationships with resource holders and seeking new relationships with organizations that built successful state funding portfolios (Park and Rethemeyer 2014). Also, as seen in figure 1, resource seekers, such as state-funded ABE service providers (circle in a box) and technical assistance units (box), created a coalition of resource seekers within their own subgroups. In this way, they may be attempting to offset their dependence on resource holders to ensure their survival and cope with uncertainty about state funding. With the data at hand we cannot conclusively demonstrate how much “more” the ABE network structure is the product of resource seekers building relationships as compared to the role of resource seekers in the MH network. However, we know that the ABE policy funding context grants state authorities more discretion regarding financial resource allocations than the MH policy funding context. The power imbalance between resource holders and resource seekers is greater in the discretionary ABE domain than in mandated MH domain. We can say that there is no evidence in the ERGM results that the structure of the MH policy network is driven by resource seekers: every variable in the “Regulatory/Funding Dependence By Public Actors” section of table 3 is insignificant. This difference in structural determinants highlights the relationship between network structure and policy funding context in a social service policy domain. From these findings about resource holders and seekers, we state the following research propositions: Figure 1. View largeDownload slide ABE Policy Network, MDS Layout Figure 1. View largeDownload slide ABE Policy Network, MDS Layout Proposition 1: Being a financial resource holder, not a regulatory resource holder, is a stronger driver of relational structures within social service policy networks. Proposition 2: Within social service networks, the extent to which resource dependence affects relational structures differs by policy funding context. Proposition 3: Resource interdependent relationships are more important in structuring policy network in a policy context where expenditures are discretionary than in a policy context where expenditures are mandatory. Proposition 3a: When fewer resources are to be discretionarily distributed, financial and regulatory resource seekers actively create ties and develop internal communication with resource seekers. Finding 4: In the mandatory MH network, executive actors (who hold public authority) are popular as policy partners; however, they are not popular in the discretionary ABE network. Finding 5: In the discretionary ABE network, legislators and legislative committee members tend to actively form ties. Finding 6: In both the discretionary ABE and mandated MH policy networks, legislators and legislative committee members tend to form ties with each other. Based on Knoke et al.’s (1996) study, we examined how public actors (i.e., state agencies, legislators, and legislative committees) affect network structures in both the ABE and MH networks because these actors’ primary resource is public authority. We expected that public actors shape network structures differently depending on the degree of discretion over the level and continuation of funding (discretionary versus mandated). As shown in table 3, we found that state actors are not sought by other actors in ABE network (see the negative and significant coefficient on State actors alter in the ABE network); in fact, the negative coefficient suggests that state alters are unpopular as relationship partners. By contrast, MH policy actors seek state actors as their networking partners within the MH policy network (see the positive and significant coefficient on State actors alter in the MH network). Turning to the legislative actors within the networks, legislators and legislative committees in the ABE network have a strong preference for making more connections (see the positive and significant coefficient on Legislator & legislative committee ego at 1% level in table 3). However, ABE legislators are not sought by other policy actors (see the negative and significant coefficient on Legislator & legislative committee alter at 10% level) and may be avoided (as the negative coefficient implies). By contrast, legislative actors in the MH network are highly sought by other policy actors within the network (see the positive and significant coefficient on Legislator & legislative committee alter at 1% level in table 3). There is also a tendency for the legislative actors to form ties with each other in both the ABE and MH networks. This is evidenced by the coefficients on Same legislator & legislative committee in the ABE and MH networks, which are both positive and significant at the 1% level. Taken together, in the face of differences in funding context, public resource holders are active in the ABE network while they are passive in MH. In the ABE network, legislative actors who control financial flows tend to create direct ties with others, while both executive and legislative actors are not structurally “popular.” This finding suggests that when social policy is discretionary and the financial pie is not particularly large, network structure is differentiated by legislative members’ activities. Legislators and resource holders can increase political and policy monitoring capacity by intensively communicating with legislative members (see the right-side clusters created by up-triangles and “overlapping triangles” in figure 1). Our ABE interview data also suggests that the budget debacle of 2001–2002 drove network structure: A large set of legislative actors became actively involved in the ABE policy network as resource pressures mounted (see Park and Rethemeyer 2014). Legislative actors reach out within the network and especially to peers in order to better coordinate yearly appropriations. This coordination became critical as the 2001–2002 budget crisis drove deep cuts to discretionary funding. However, in the MH network, both executive and legislative actors shape the network by receiving ties from others, not by sending ties. Public sector organizations are endowed with public authority; they depend on other organizations’ information to inform their decision making (Knoke et al. 1996). Walker (1991) demonstrated that public interest groups are mobilized to serve as reliable sources of information by executive agencies as part of the policy process. Simply put, in a social service policy domain, executive actors tend to receive incoming ties from others without developing outgoing ties. Information flows toward executive agencies in this scheme. Moreover, in the MH context, (1) the social service in question—care of the severely mentally ill—is an entitlement rather than a discretionary expenditure, (2) substantially more public resources are at stake (more than $1 billion versus about $100 million), and (3) alternative financial resources exist inside and outside the network in the form of private insurers who must cover some insured individuals with severe mental illnesses. These aspects of the policy funding context make legislators in the MH network less driven to assert control through information gathering via extensive ties structures. A large portion of the mental health budget is on “autopilot” through mandates. Instead, legislative actors may selectively choose among those actors who are actively seeking ties with them to address more narrowly defined issues—for instance, the creation and content of a formulary, which was a major issue at the time this data was collected. While the formulary was of deep interest to some organizations, it was not an existential threat to the overall service structure. The structure is different in ABE, where yearly budget decisions can have sweeping and even existential implications. Legislators are more outgoing but must contend with constraints from powerful nonlegislative actors who wish to assert control over interactions with lawmakers. As seen in figure 2, the MH network is less differentiated but more structurally complex than the ABE network. From these findings, we have derived the following propositions: Figure 2. View largeDownload slide MH Policy Network, MDS Layout Figure 2. View largeDownload slide MH Policy Network, MDS Layout Proposition 4: Legislators and legislative committee members monitor policy environments by creating ties and intensively communicating among themselves when (a) social services are discretionary and (b) financial resources are relatively scarce. Proposition 5: Public actors monitor policy environments by receiving information from others when (a) a social service is mandatory, (b) large amounts of public resources are at stake, and (c) alternative financial sources exist outside of the network. Proposition 6: Legislators and legislative committee members are strongly homophilous in shaping social service policy networks. Finding 7: In both the ABE and MH policy networks, policy actors with high influence scores are popular. As we expected, influential actors in both networks are sought out by other policy actors since both coefficients on Influence alter are positive and significant at the 1% level. However, perceived influence constrains actors’ behavior in choosing instrumental ties in the ABE and MH networks differently. In ABE, influential actors, such as the State Department of Education, legislative committees (especially the Senate and House Ways and Means Committees), legislators, research and advocacy organizations, service providers (literacy volunteer organizations), the industry association (which we pseudonymed NAABE—Newstatia Alliance for Adult Basic Education), and research and technical assistance units are extensively sought by other policy actors. In the case of the MH network, influential actors, such as the state agency (Department of Mental Health), committees (Senate and House Ways and Means Committees), legislators, newspapers, industry association (which we pseudonymed NAMH—Newstatia Alliances for Mental Health), insurers, and client and family advocacy organizations are sought for ties within the MH network. Also, these influential actors tend to restrict their interactions to a set of powerful peers within the MH network (see the positive and significant coefficients on influence alter and influence similarity at 1% significance level in MH in table 3). Thus, information is more closely held in the MH network. Proposition 7: Influential policy actors are strongly sought by other policy actors across social service policy networks despite differences in policy funding context. Finding 8: In the MH network, legal client advocacy organizations and industrial and professional associations are more outgoing. Finding 9: In the ABE network, research and technical support organizations prefer homophilous policy ties. As noted, we focused on five types of nonprofits: (1) client advocacy organizations, (2) legal client advocacy organizations, (3) industrial/professional associations and foundations, (4) research and technical support organizations, and (5) service delivery organizations. We dropped one type—client advocacy organizations—from our ABE and MH models due to the lack of statistical significance.1 Thus, only four types of nonprofit organizations are used for the analysis. In our categorization, advocacy means promoting a policy agenda on behalf of socially marginalized groups through the courts, the legislature, administrative agencies, and/or the public at large and service delivery implies the provision of services directly to individuals and families (Chetkovich and Kunreuther 2006). However, our categorization is not mutually exclusive: service providers advocate, and some advocacy groups provide services. We found that MH and ABE policy networks are shaped differently by these advocacy and service delivery organizations. The coefficients on both Legal client advocacy organizations ego and Industrial/professional associations and foundations ego in the MH network are positive and significant at the 5% significance level (table 3). This finding suggests that legal client advocacy organizations and professional associations and foundations tend to actively engage in creating ties with other policy actors in the MH network. In MH, there are a number of specialized advocacy organizations, including interest groups for family members of people with mental disorders, professional associations that represent those who work with the severely mentally ill, and organizations that use legal means to (1) advocate for policies and laws against discrimination, (2) improve services and secure the just treatment of mentally ill people, and (3) assist clients seeking access to services for those who are severely mentally ill (Funk et al. 2006). These organizations exist because statutes mandate MH services for those with qualifying conditions. Newstatia also financially supports some legal advocacy activities in our case. Figure 3. View largeDownload slide Conceptual Model Linking Policy Funding Context to Network Structure Figure 3. View largeDownload slide Conceptual Model Linking Policy Funding Context to Network Structure Thus, the MH network includes both advocacy organizations that can facilitate mass mobilization for legislative action and organizations that can seek policy change through enforcement of legal rights of citizens with severe mental illnesses. Here, legal client advocacy organizations, representing the formally intended beneficiaries by MH law and policy, play a pivotal role in structuring the MH network. Also, since the health field is highly politicized and institutionalized (Child and Grønbjerg 2007), professional and industrial associations in MH (those that represent inpatient mental health facilities, mental health clubhouses, outpatient clinics, and a union of MH workers) invested in advocating with state actors who have the power to modify regulations and MH programs. For example, the industry association (NAMH) advocates in legislative and executive processes for both institutional members as well as individuals receiving services and their families. The professional lobbyists in these industry associations give testimony in legislative processes and try to educate policymakers about the impact of MH issues. Since government programs are a key source of revenue for the members of the MH industrial and professional associations (Child and Grønbjerg 2007), these associations tend to actively engage in the MH network by seeking new ties with other policy actors. MH’s mandatory policy funding context combined with the existence of secondary interest groups that represent client and industry interests mitigate toward systematic advocacy mobilization processes (e.g., institutionalized lobbying and litigation). In the discretionary ABE context, client advocacy organizations are not a significant factor shaping the ABE network. However, activities by research and technical support organizations were significant. In our first ERGM of the ABE network, we included the adult learners’ advocacy organizations as a variable, but the coefficient was statistically insignificant. This result was not surprising: the learner’s advocacy organization was small, poorly funded, and not well known in Newstatia. Most ABE learners are socioeconomically disadvantaged and thus tend to lack the resources needed to support a robust advocacy organization. Additionally, ABE learners do not have secondary interested parties (like family members in the case of the mentally ill) who are financially and politically positioned to advocate for them. Finally, in the absence of mandates, there is no way to use the courts to promote ABE learner interests and thus no legal client advocacy organizations engaged in this social service policy domain. These policy funding context differences have restricted the opportunities for client advocacy to systematic mobilization of clients by providers. Taken together, our data suggests that client and industry advocacy organizations are more active when legal rights are granted to protect groups of citizens, whereas these activities are lessened when there are no rights to enforce. However, in the ABE network, research and technical assistance organizations emerged as major advocates. The coefficient on Same research and technical support organizations is positive and significant at 1% significance level. This result implies that research and technical support units in the ABE network tend to communicate intensely among themselves. Our field research confirmed that these organizations provided critical infrastructure and coordination for political and policymaking activities by private sector ABE providers. While these organizations certainly consider the interests of their clients—indeed, they exist to serve clients—it is also indisputable that their advocacy was deeply tied to the perspectives of service providers and other funded units. ABE learners’ advocacy organizations do not play a central leadership role in ABE policymaking; only the provider organizations and industry association do. From this analysis, we state the following research propositions: Proposition 8: In policy contexts where services are legally mandated and secondary interested parties exist, legal client advocacy organizations and industrial & professional associations and foundations play an essential role in structuring social service policy networks. Proposition 9: In policy contexts where services are provided using discretionary funds, providers, research & technical support organizations, and other funded entities are important in shaping social service policy networks. Discussion and Conclusion Networks have become an increasingly common feature of both policymaking and service delivery in a wide range of social safety net programs (Kettl 1996; Milward, Provan, and Else 1993; Salamon 1981, 1995). While much of the research has focused on how network structure affects access, participation, and eventually outcomes, few studies examine the relationship between policy funding contexts and structures in social service policy networks. We seek to close this gap by using comparative quantitative analysis of two policy networks—one focused on a discretionary context (ABE) and another on a mandatory context (MH)—to induce a set of propositions about the relationship between funding context and network structure. Here, we have explicitly examined the interaction between policy funding contexts and dominant theoretical traditions that have informed previous work on social policy networks—resource dependence and advocacy mobilization. Our analyses suggest that the nature of the policy funding stream—discretionary or mandatory—affects the nature of dependence between network actors and the type of advocacy mobilization that occurs. In a discretionary social service policy context, funding depends on annual allocations from the legislature. In this context, resource seekers create advocacy coalitions within the policy networks that focus on legislative advocacy. The network structure flows from the need to keep the legislative spigot open. Highly influential intermediary policy actors—actors that do not necessarily fund providers—also play a role in structuring policy networks in a discretionary context. In the ABE example, these actors include the State Board of Education, a legislative committee focused on social services, a state-based educational research center, an industry association, and several state-funded technical assistance units. These organizations intermediate and control interactions between resources holders—legislators—and resource seekers—funded programs and support organizations. It is through interactions between resource seekers and intermediary actors that resource holders coordinate discretionary funding flows. The structure of the discretionary social service policy network is tied to intermediary actors’ efforts and developed to assure there was a financial “pie” of any size to divide. When much of the funding for a social service policy is programmed and mandated, legislative success is not necessarily about survival. Mandates put “guide rails” around the degree to which legislators may choose winners and losers in the absence of legislation to revoke or greatly modify an existing mandate. Revising a mandate is usually (a) politically difficult—if not impossible—given established coalitions and (b) procedurally difficult—if not impossible—if the mandates are encoded into state constitutions or federal statutes. Additionally, in some mandated social service areas—like care of the severely mentally ill and many other health care domains—there are nonpublic sources of funding that provide an alternative to government as the only financial source of survival. These funders are themselves the subject of policy action: insurers, for instance, are the subject of regulation that affects the services available to policyholders and thus the streams of funding available to service providers. In the face of a mandate, the clear financial dependence of service providers on legislative decision makers found in the discretionary context gives way to a much more complex picture. Mandates provide much greater certainty with respect to public funding: policy network members are relatively certain there will be a “pie” to divide. The question is, who gets the bigger slices? Seen in this light, the differences in advocacy structures between discretionary and mandatory social service areas may be clearly traced to differences in policy funding context: the funding structure implies the advocacy structure that flows from it provided one also accounts for differences in client populations. Starting with the question of funding structure, there are various ways to seek a bigger slice of a social service pie: legislative action; litigation to enforce or modify a mandate; efforts to revise the regulatory constraints on alternative funders; and administrative processes. Policy funding context determines which of the “advocacy technologies” is most appropriate or even possible. When funding is almost exclusively discretionary, then many advocacy technologies fall away as irrelevant or impractical: organizing to influence legislative actors is the primary method to assure there is a pie at all. When funding is mandated, the range of advocacy technologies available expands significantly: there are legal, legislative, and administrative leverage points which lead interests to organize around their particular client needs and their preferred advocacy technology. However, client factors also matter: some client populations are more able to organize for collective action than others. In ABE, the primary constituency is poorly resourced and sometimes politically excluded (in the case of illegal immigrants) and thus unable to strongly advocate for its interests (though in some other states there are strong immigrant community efforts to mobilize on behalf ESL client subgroup—just not in Newstatia). In MH, clients vary by their degree of function, but most have families that have a deep interest in policy choices. The upshot is that the nature of the funding stream (discretionary versus mandated) interacts with the nature of the client population to determine the array of interest organizations that actively participate in the policy network. These differences in context ramify into the structure of the policy network through the following hypothesized causal chain: Fundamentally, these differences in policy network context are prior to dependence—indeed, dependence is actually brought into being by these features of context. Dependence and client factors elicit advocacy organizations, and collectively these help to determine network structure. Public managers in Newstatia operate in a political/policy environment defined by these factors. As with any research conducted on a highly limited number of cases, our results are not generalizable. However, our findings do suggest that future efforts to understand the role of resource dependence in structuring policy networks—and possibly other interorganizational structures for that matter, including collaborative networks—need to carefully consider the contextual factors that structure dependence. Resource dependence flows from a constitutive policy framework. In this study we have focused on one aspect of that policy framework: the policy funding context. However, other aspects of the policy framework may also matter—for instance, the degree to which funding is local, state, or federal; policy “flow-downs” from state to local or federal to state; or the degree to which policy is institutionalized through law or executive action. Networks reflect the policy funding context and the dependencies called into being by it. Our findings with respect to the role of resource dependence are only as solid as the underlying funding contexts that call dependencies into being. Assumptions and Mechanics of ERGMs In ERGMs, actors are assumed to make relationships based on what they know about the state of the network today (thus not taking into account the past) and what they would prefer in terms of their overall patterns of relationships if they are allowed to make a change to their ties. For instance, if Actor A is tied to Actor B and knows that Actor B is also tied to Actor C, Actor A may choose to make a tie to Actor C because there is an underlying preference for social relations to be “closed”—a friend of a friend is a friend. However, the tie could also be made separate from the preference for closure if A and C are of the same gender and homophily by gender is preferred. The tie becomes even more likely if both are true: there is a general preference for closure and for gender homophily. Thus, ERGMs evaluate across the entire range of proposed factors that may make ties in the dependent network more or less likely. The dependent network (here, the confidential communication network) provides the data on what actual choices were made by network members. The estimated coefficients reflect what factors were most likely to have driven relational choices given the actual relationships reported in the data. The magnitude and statistical significance of the coefficients, like in linear regression, tell us whether the proposed factors—homophily, preference for closure, etc.—appear to have been important to relational choices or not. The relational choices across all actors in network define the totality of the network. Footnotes 1 More specifically, in the ABE policy network, client advocacy organization (i.e., adult learners’ advocacy organization) was not statistically significant in the model and dropped to simplify the model. In the MH policy network, client advocacy organization was not significant at 10% level in the goodness of fit test; thus, it was not included in the model. Appendix Assumptions and Mechanics of ERGMs In ERGMs, actors are assumed to make relationships based on what they know about the state of the network today (thus not taking into account the past) and what they would prefer in terms of their overall patterns of relationships if they are allowed to make a change to their ties. For instance, if Actor A is tied to Actor B and knows that Actor B is also tied to Actor C, Actor A may choose to make a tie to Actor C because there is an underlying preference for social relations to be “closed”—a friend of a friend is a friend. However, the tie could also be made separate from the preference for closure if A and C are of the same gender and homophily by gender is preferred. The tie becomes even more likely if both are true: there is a general preference for closure and for gender homophily. Thus, ERGMs evaluate across the entire range of proposed factors that may make ties in the dependent network more or less likely. The dependent network (here, the confidential communication network) provides the data on what actual choices were made by network members. The estimated coefficients reflect what factors were most likely to have driven relational choices given the actual relationships reported in the data. The magnitude and statistical significance of the coefficients, like in linear regression, tell us whether the proposed factors—homophily, preference for closure, etc.—appear to have been important to relational choices or not. The relational choices across all actors in network define the totality of the network. References Agranoff, Robert, and Michael McGuire. 2001. Big questions in public network management research. Journal of Public Administration Research and Theory  11: 295– 326. Google Scholar CrossRef Search ADS   ———. 2004. Collaborative public management: New strategies for local governments . Washington, DC: Georgetown Univ. Press. Ansell, Chris, and Alison Gash. 2008. Collaborative governance in theory and practice. Journal of Public Administration Research and Theory  18: 543– 71. Google Scholar CrossRef Search ADS   Berry, Jeffrey M. 2005. Nonprofits and civic engagement. Public Administration Review  65: 568– 78. Google Scholar CrossRef Search ADS   Biesta, Gert. 2007. Why ‘what works’ won’t work: Evidence-based practice and the democrative deficit in educational research. Educational Theory  57: 1– 22. Google Scholar CrossRef Search ADS   Bodin, Orjan, and Beatrice Crona. 2009. The role of social networks in natural resource governance: What relational patterns make a difference? Global Environmental Change: Human and Policy Dimensions  19: 366– 74. Google Scholar CrossRef Search ADS   Boer, P., M. Huisman, Tom A. B. Snijders, C. Steglich, L. H. Y. Wichers, and E. P. H. Zeggelink. 2006. StOCNET: An open software system for the advanced statistical analysis of social networks . Groningen, The Netherlands: ICS/SciencePlus Group, Univ. of Groningen. Bonacich, Phillip, and Elisa Jayne Bienenstock. 2009. Structural implications of reciprocal exchange: A power-dependence approach. Social Psychology Quarterly  72: 226– 40. Google Scholar CrossRef Search ADS   Borgatti, S. P. 2002. NetDraw: Graph visualization software . Harvard: Analytic Technologies. Borgatti, S. P., M. G. Everett, and L. C. Freeman. 2002. UCINET 6 for Windows: Software for social network analysis . Harvard: Analytic Technologies. Bowen, T. Renee, Ying Chen, and Hülya Eraslan. 2014. Mandatory versus discretionary spending: The status quo effect. American Economic Review  104: 2941– 74. Google Scholar CrossRef Search ADS   Bryson, John M., Barbara C. Crosby, and Melissa Middleton Stone. 2006. The design and implementation of cross-sector collaborations: Propositions from the literature. Public Administration Review  66( Suppl 1): 44– 55. Google Scholar CrossRef Search ADS   Casciaro, Tiziana, and Mikolaj Jan Piskorski. 2005. Power imbalance, mutual dependence, and constraint absorption : A closer look at resource dependence theory. Administrative Science Quarterly  50: 167– 99. Google Scholar CrossRef Search ADS   Chetkovich, Carol A., and Frances Kunreuther. 2006. From the ground up: Grassroots organizations making social change . Ithaca, NY: Cornell Univ. Press. Child, Curtis D., and Kirsten A. Grønbjerg. 2007. Nonprofit advocacy organizations: Their characteristics and activities. Social Science Quarterly  88: 259– 81. Google Scholar CrossRef Search ADS   Cho, Chung Lae, and Deil S. Wright. 2004. The devolution revolution in intergovernmental relations in the 1990s: Changes in cooperative and coercive state-national relations as perceived by state administrators. Journal of Public Administration Research and Theory  14: 447– 68. Google Scholar CrossRef Search ADS   Eisenhardt, Kathleen M., and Melissa E. Graebner. 2007. Theory building from cases: Opportunities and challenges. Academy of Management Journal  50: 25– 32. Google Scholar CrossRef Search ADS   Emerson, Richard. 1962. Power-dependence relations. American Sociological Review  27: 31– 41. Google Scholar CrossRef Search ADS   Emerson, Kirk, Tina Nabatchi, and Stephen Balogh. 2012. An integrative framework for collaborative governance. Journal of Public Administration Research Theory  22: 1– 29. Google Scholar CrossRef Search ADS   Frank, O., and David Strauss. 1986. Markov graphs. Journal of the American Statistical Association  81: 832– 42. Google Scholar CrossRef Search ADS   Funk, Michelle, Alberto Minoletti, Natalie Drew, Jacob Taylor, and Benedetto Saraceno. 2006. Advocacy for mental health: Roles for consumer and family organizations and governments. Health Promotion International  21: 70– 75. Google Scholar CrossRef Search ADS   Gais, Thomas L., and Jack L. Walker. 1991. Pathways to influence in American politics. In Mobilizing interest groups in America , ed. Jack L. Walker, 103– 21. Ann Arbor: Univ. of Michigan Press. Graddy, E. A., and B. Chen. 2006. Influences on the size and scope of networks for social service delivery. Journal of Public Administration Research and Theory  16: 533– 52. Google Scholar CrossRef Search ADS   Gulati, Ranjay, and Harbir Singh. 1998. The architecture of cooperation: Managing coordination costs and appropriation concerns in strategic alliances. Administrative Science Quarterly  41: 781– 814. Google Scholar CrossRef Search ADS   Hatmaker, Deneen M., and R. Karl Rethemeyer. 2008. Mobile trust, enacted relationships: Social capital in a state-level policy network. International Public Management Journal  11: 426– 62. Google Scholar CrossRef Search ADS   Hillman, A. J., M. C. Withers, and B. J. Collins. 2009. Resource dependence theory: A review. Journal of Management  35: 1404– 27. Google Scholar CrossRef Search ADS   Huang, Kun, and Keith G. Provan. 2007. Resource tangibility and patterns of interaction in a publicly funded health and human services network. Journal of Public Administration Research and Theory  17: 435– 54. Google Scholar CrossRef Search ADS   Ingold, Karin. 2011. Network structures within policy processes: Coalitions, power, and brokerage in Swiss climate policy. Policy Studies Journal  39: 435– 59. Google Scholar CrossRef Search ADS   Ingold, Karin, and Philip Leifeld. 2016. Structural and institutional determinants of influence reputation: A comparison of collaborative and adversarial policy networks in decision making and implementation. Journal of Public Administration Research and Theory  26: 1– 18. Google Scholar CrossRef Search ADS   Isett, Kimberley R., Ines A. Mergel, Kelly Leroux, Pamela A. Mischen, and R. Karl Rethemeyer. 2011. Networks in public administration scholarship: Understanding where we are and where we need to go. Journal of Public Administration Research and Theory  21( Suppl 1): i157– i173. Google Scholar CrossRef Search ADS   Jack L. Walker. 1991. Mobilizing interest groups in America: Patrons, professions, and social movements . Ann Arbor, MI: University of Michigan Press. Johnston, J. M., and B. S. Romzek. 2008. Social welfare contracts as networks: The impact of network stability on management and performance. Administration & Society  40: 115– 46. Google Scholar CrossRef Search ADS   Kenis, Patrick and Raab, Jörg. 2007. Taking stock of policy networks: Do they matter ? In Handbook of public policy analysis: Theory, politics and methods , ed. M. S. Sidney, F. Fisher, and G. J. Miller, 187– 200, Boca Raton, FL: CRC Press. Kettl, D. F. 1996. Governing at the millennium. In Handbook of public administration , ed. J. L. Perry, 5– 18. San Francisco: Jossey-Bass. Klijn, Erik-Hans, and Chris Skelcher. 2007. Democracy and governance networks: Compatible or not. Public Administration  85: 587– 608. Google Scholar CrossRef Search ADS   Knoke, David. 1990. Political networks: The structural perspective . Cambridge, UK: Cambridge Univ. Press. Google Scholar CrossRef Search ADS   ———. 2014. Policy networks. In The sage handbook of social network analysis , ed. John Scott and Peter J. Carrington, 210– 22. Thousand Oaks, CA: Sage Publications. Google Scholar CrossRef Search ADS   Knoke, David, F. U. Pappi, H. Broadbent, and Y. Tsujinaka. 1996. Comparing policy networks: Labor politics in U.S., Germany, and Japan . Cambridge, UK/New York: Cambridge Univ. Press. Google Scholar CrossRef Search ADS   Laumann, Edward, and David Knoke. 1987. The organizational state: Social choice in national policy domains . Madison, WI: Univ. of Wisconsin Press. Laumann, Edward, Peter V. Marsden, and David Prensky. 1989. The boundary specification problem in network analysis. In Research methods in social network analysis , ed. Linton C. Freeman, Douglas R. White, and A. Kimball Romney, 61– 87. Fairfax, VA: George Mason Univ. Press. Milward, H. Brinton, and Keith G. Provan. 2000. Governing the Hollow State. Journal of Public Administration Research and Theory  10: 359– 80. Google Scholar CrossRef Search ADS   Milward, H. Brinton, Keith G. Provan, and Barbara A. Else. 1993. What does the hollow state look like. In Public management: The state of the art , ed. B. Bozeman, 309– 32. San Francisco: Jossey-Bass. Mosley, Jennifer E. 2012. Keeping the lights on: How government funding concerns drive the advocacy agendas of nonprofit homeless service providers. Journal of Public Administration Research and Theory  22: 841– 66. Google Scholar CrossRef Search ADS   OECD. 2012. Glossary of key terms for completing the 2012 OECD Budgeting Practices and Procedures Survey. http://www.oecd.org/gov/budgeting/Budgeting-survey-glossary.pdf. (accessed March 22, 2018) Oliver, Christine. 1990. Determinants of interorganizational relationships: Integration and future directions. Academy of Management Review  15: 241– 65. ———. 1991. Strategic responses to institutional processes. Academy of Management Review  16: 145– 79. CrossRef Search ADS   Park, Hyun Hee, and R. Karl Rethemeyer. 2014. The politics of connections: Assessing the determinants of social structure in policy networks. Journal of Public Administration Research and Theory  24: 349– 79. Google Scholar CrossRef Search ADS   Pattison, Philippa, and Garry Robins. 2002. Neighborhood-based models for social networks. Sociological Methodology  32: 301– 37. Google Scholar CrossRef Search ADS   Pattison, Philippa, and S. Wasserman. 1999. Logit models and logistic regressions for social networks: II. Multivariate relations. British Journal of Mathematical and Statistical Psychology  52: 169– 93. Google Scholar CrossRef Search ADS   Pfeffer, J. 1987. A Resource dependence perspective on intercorporate relations. In Intercorporate relations: The structural analysis of business , ed. M. S. Mizruchi and M. Schwartz, 22– 55. New York, NY: Cambridge Univ. Press. Google Scholar CrossRef Search ADS   Pfeffer, J., and G. Salancik. 1978. The external perspective on organizations: A resource dependence perspective . New York, NY: Harper & Row. Powell, Walter W. 1990. Neither market nor hierarchy: Network forms of organizations. In Research in organizational behavior , ed. Barry M. Staw and Larry L. Cummings, 295– 336. Greenwich, CT: JAI Press. Powell, Walter W., Kenneth W. Koput, and Laurel Smith-Doerr. 1996. Interorganizational collaboration and the locus of innovation: Networks of learning in biotechnology. Administrative Science Quarterly  41: 116– 46. Google Scholar CrossRef Search ADS   Provan, Keith G., Janice M. Beyer, and Carlos Kruytbosch. 1980. Environmental linkages and power in resource-dependence relations between organizations. Administrative Science Quarterly  25: 200– 25. Google Scholar CrossRef Search ADS   Provan, Keith G., and H. Brinton Milward. 1995. A preliminary theory of interorganizational network effectiveness: A comparative study of four community mental health systems. Administrative Science Quarterly  40: 1– 33. Google Scholar CrossRef Search ADS   Raab, Jörg. 2002. Where do policy networks come from? Journal of Public Administration Research and Theory  12: 581– 622. Google Scholar CrossRef Search ADS   Rethemeyer, R. Karl. 2007. Policymaking in the age of internet: Is the internet tending to make policy networks more or less inclusive? Journal of Public Administration Research and Theory  17: 259– 84. Google Scholar CrossRef Search ADS   Rethemeyer, R. Karl, and Deneen M. Hatmaker. 2008. Network management reconsidered: An inquiry into management of network structures in public sector service provision. Journal of Public Administration Research and Theory  18: 617– 46. Google Scholar CrossRef Search ADS   Rhodes, R. A. W. 1997. Understanding governance: Policy networks, governance, reflexivity and accountability . Buckingham: Open University Press. Ripley, Ruth, and Tom A. B. Snijders. 2009. Manual for SIENA Version 4.0 . http://www.stats.ox.ac.uk/~snijders/siena/s_man401.pdf. (accessed March 22, 2018) Robins, Garry, Phillipa Pattison, Yuval Kalish, and Dean Lusher. 2007a. An introduction to exponential random graph (P*) models for social networks. Social Networks  29: 173– 91. Google Scholar CrossRef Search ADS   Robins, Garry, Philippa Pattison, and Stanley Wasserman. 1999. Logit models and logistic regressions for social networks: III. Valued relations. Psychometrika  64: 371– 94. Google Scholar CrossRef Search ADS   Robins, Garry, Tom A. B. Snijders, Peng Wang, Mark Handcock, and Philippa Pattison. 2007b. Recent developments in exponential random graph (P*) models for social networks. Social Networks  29: 192– 215. Google Scholar CrossRef Search ADS   Saidel, Judith R. 1991. Resource interdependence: The relationship between state agencies and nonprofit organizations. Public Administration Review  51: 543– 53. Google Scholar CrossRef Search ADS   Salamon, L. M. 1981. Rethinking public management: Third-party government and the changing forms of government action. Public Policy  29: 255– 75. ———. 1987. Of market failure, voluntary failure, and third-party government: Toward a theory of government-nonprofit relations in the modern welfare state. Nonprofit and Voluntary Sector Quarterly  16: 29– 49. ———. 1995. Partners in public service: Government-nonprofit relations in the modern welfare state . Baltimore, MD: Johns Hopkins Univ. Press. Scheppele, K. L., and Jack L. Walker. 1991. The litigation strategies of interest groups. In Mobilizing interest groups in America , ed. Jack L. Walker, 335– 72. Ann Arbor, MI: Univ. of Michigan Press. Smith, S. R., and M. Lipsky. 1993. Nonprofits for hire: The welfare state in the age of contracting . Cambridge, MA: Harvard Univ. Press. Snijders, Tom A. B. 2001. The statistical evaluation of social network dynamics. Sociological Methodology  31: 361– 395. Google Scholar CrossRef Search ADS   ———. 2002a. Markov chain Monte Carlo estimation of exponential random graph models. Journal of Social Structure  3: 40. ———. 2002b. The SIENA Homepage (Simulation Investigation for Empirical Network Analysis) . http://www.stats.ox.ac.uk/~snijders/siena/. (accessed March 22, 2018) Snijders, Tom A. B., and Marijtje A. J. Duijn. 2002. Conditional maximum likelihood estimation under various specifications of exponential random graph models. In Contributions to social network analysis, information theory, and other topics in statistics: A festschrift in honour of ove frank , ed. J. Hagberg, 117– 34. Stockholm, Sweden: University of Stockholm, Department of Statistics. Snijders, Tom A. B., P. E. Pattison, G. L. Robins, and M. S. Handcock. 2006. New specifications for exponential random graph models. Sociological Methodology  36: 99– 153. Google Scholar CrossRef Search ADS   Sørensen, Eva, and Jacob Torfing. 2005. The democratic anchorage of governance networks. Scandinavian Political Studies  28: 195– 218. Google Scholar CrossRef Search ADS   US Congress Office of Technology Assessment. 1993. Adult literacy and new technologies: Tools for a lifetime . Washington, DC: US Government Printing Office. US Department of Health and Human Services. 1999. Mental health: A report of the surgeon general . Rockville, MD: US Department of Health and Human Services, Substance Abuse and Mental Health Services Administration. Walker, Jack L. 1991. Mobilizing interest groups in America: Patrons, professions, and social movements . Ann Arbor, MI: Univ. of Michigan Press. Google Scholar CrossRef Search ADS   Westmoreland, Timothy M., and Kathryn R. Watson. 2006. Redeeming hollow promises: The case for mandatory spending on health care for American Indians and Alaska natives. American Journal of Public Health  96: 600– 605. Google Scholar CrossRef Search ADS   Woolf, Steven H. 2009. Social policy as health policy. Journal of the American Medical Association  301: 1166– 69. Google Scholar CrossRef Search ADS   © The Author(s) 2018. Published by Oxford University Press on behalf of the Public Management Research Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.

Journal

Journal of Public Administration Research and TheoryOxford University Press

Published: Mar 28, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off