From Policy Networks to Policy Preferences: Organizational Networks in the Opt-Out Movement

Organizational networks shape education policy by influencing power holders and elites, but do they have similar effects on grassroots activists? We use data from the National Survey on Opting Out (2016 and 2018; n = 2,909) to examine the role of organizational networks in mobilizing activists in the opt-out movement (a movement in which parents and caregivers refuse to have their children sit for standardized tests). Despite characterizations of the opt-out movement as a bunch of “soccer moms” disappointed with their children’s test scores, our findings show that opt-out is in fact a structured movement reliant on social movement organizations (SMOs) with agendas that go beyond standardized tests. Further, we demonstrate a small but significant correlation between contact with SMOs and individual policy From policy networks to policy preferences 2 preferences. These patterns suggest that organizational networks may inform education policy by creating a social space for activists to learn about different policy ideas in education. We discuss implications for research and practice.


Introduction
During the past decade, scholars have increasingly focused on the role of networks in education politics and policy. As Stephen Ball has noted, the shift from a bureaucratic, statecentered policy making apparatus to one that includes a range of actors, from business to philanthropy to advocacy organizations, has impacted policy ideas, governance and public discourse and represents a new way of "getting policy work done" (Ball, 2008, p. 762). The idea of utilizing organizational networks to understand policy making and policy implementation has been gaining momentum, even as education policy has typically been largely oriented toward hierarchically organized state education authorities (Russell, Meredith, Childs, Stein & Prine, 2014). Indeed, one of the attractions of studying policy networks is to shed light on the increasingly complex power dynamics of education policy making, as networks mobilize their resources to influence education policy at the local, national and international levels.
In education, the research on policy networks has typically focused on outcomes at the state or national levels, exploring the impact that networked financing, discourse, and agenda setting have on contentious education policy. For instance, networks have been used to analyze the politics of charter school policy and advocacy (Au & Ferrare, 2014;Mintrom & Vergari, 1998;Scott, 2009) and in studies on the growing role of philanthropic dollars in education reform (e.g., Reckhow & Snyder, 2014). However, the role of grassroots efforts to change education policy remains underexplored. Thus, we know little from the policy networks literature about the role of networks that are operating from the 'bottom up,' such as educational social movements that respond to and challenge state education policy.
Luckily, scholars of social movements and sociologists of education have developed useful theoretical and analytical tools to examine the role of organizations and networks in contentious politics. Social movements theory suggests an important role for networks, both individual and organizational, in mobilizing and connecting activists and resources in their efforts to challenge authority (Diani, 1992;Heaney & Rojas, 2007, 2008Zald & Ash, 1966). Diani defines social movements as "networks of informal interactions between a plurality of individuals, groups and/or organizations, engaged in political or cultural conflicts, on the basis of shared collective identities" (Diani, 1992, p.8). While social movements are comprised of individual activists, the literature notes a particular role for social movements organizations (SMOs) in moving from individual to collective action. SMOs can be understood as the 'carriers' of social movements, giving structure, resources and the force of collective action to a movement's aims, though here too most of the literature on the impact of SMOs measures impact as a function of legislative change at the state or national policy levels (Burstein & Linton, 2002).
In this paper, we use the case of the opt-out movement as a bridge to connect the literature on policy networks with that of social movements in order to better elucidate the impact of networks at the grassroots level. Specifically, we explore the links between individual activists and SMOs, and whether or not activists' connections to SMOs are associated with individual policy preferences. While activists may have certain individual dispositions that draw them to a movement, we focus on whether those activists who are connected to a network have different preferences than those who are not. The opt-out movement is an appropriate case for this work as it has been one of the more highly visible protest movements in education politics during the past five years.1 The movement, in which participants have protested against high-stakes testing as a form of educational accountability, is largely thought to be focused on test refusal. However research has shown that the movement is about more than just opposition to high-stakes-testing, and activists in the movement have expressed concern about a variety of current educational policies and reform efforts forthcoming).
In this paper, we describe a study of activists in the opt-out movement and their organizational network. Data come from an original survey project, the National Survey on Opting Out, that the authors conducted in 2016 and 2018. The survey instrument includes a question on affiliation/contact with SMOs, questions on policy attitudes, and a host of socio-demographic background variables. We use social network analysis (SNA) and multivariate regression models to answer three related research questions: RQ #1: What is the organizational network of the opt-out movement? RQ #2: Which activists in the opt-out movement are connected to social movements organizations (SMOs)? RQ #3: How do activists with different levels of connection to the opt-out movement vary in their attitudes towards educational reform?
After presenting the SMO network, and analyzing which activists were more likely to be connected with the organizational network, we find that organizational networks do in fact make a difference in individual policy attitudes. Controlling for sociodemographic characteristics, we find that activists that have contact with the organizational network have different views on a variety of educational reform topics than those activists who do not have contact with the organizational network. We also find that the connected activists have different views than the public on certain education policy issues. This study contributes to the literature on policy networks in education by considering social movements and their organizational networks as important types of networks in contentious education politics. In addition, the study offers a way to bridge two lines of scholarship that are often disconnected (i.e., policy networks and social movements theory). Furthermore, the study contributes to the growing literature on the opt-out movement in the U.S., and to research on social movements and policy preferences in education.
The rest of the paper proceeds as follows. We first discuss the social science literature on policy networks, SMOs and education policy, followed by more specific information about what we know to date about the opt-out movement in the United States. After explaining our data and methods, we present findings, including a visualization the opt-out SMOs network, an analysis of activist connectedness, and evidence of network effects on activists' educational policy views. We then discuss these findings and suggest avenues for future research.

Literature Review Policy Networks
Social science literature has explored the importance of policy making and policy change, but political science and sociology have emphasized different aspects of how networks may impact policy. The policy literature understands policy networks as "sets of formal and informal linkages between governmental and other actors structured around […] public policy making and implementation" (Rhodes, 2006, p. 426). These networks can act as mediators of ideas, linkages between political institutions or governance structures in and of themselves. The rise of networks, some have argued, has led to new forms of governance -"government without governing" (Rhodes, 2006, p. 428) -as governments have had to incorporate new actors, new policy interests and ideas, as well as additional expertise into the policy process (Ball, 2008). The ways in which network actors influence both policy and each other has been a key focus of the policy network theory research (Howlett, Mukherjee & Koppenjan, 2017).
The policy literature has focused generally on how organizational networks influence policy at the state or federal level. The most prominent conceptualizations of policy networks have tended to focus on the role of organized interest groups, especially "advocacy coalitions" (Sabatier & Jenkins-Smith, 1993;Sabatier & Weible, 2007). The advocacy coalitions framework (ACF) identifies beliefs as the causal driver of political change, meaning that shifts in the beliefs of individual policy actors, coupled with available resources and situated in a shifting political context (e.g. changes in socioeconomic conditions, governing coalitions, public opinion), leads to the advocacy of new policy positions. A review of 80 ACF case studies published from 1987 to 2006 demonstrated how the ACF has been applied to host of policy issues, especially those in environmental and health policies (Weible, Sabatier & McQueen, 2009). However, the outcomes of interest in these cases are most commonly policy impacts at the state and federal levels. Though research has shown that several types of individuals, e.g., street-level bureaucrats, policy entrepreneurs, policy brokers and others, can play important roles in the policy process, the role of individuals as policy agents is under explored in the policy network context. Scholars note that the move to more networked governance opens up a space to revisit the role of individuals "not as atomistic actors operating on their own, but rather as actors situated in various structures" (Olsson & Hysing, 2012, p. 257).

Social Movements and Social movements Organizations
Social movements theory also recognizes the role of networks and their impact on policy. Sociologists have focused on social movements and social movements organizations as key constructs in the process of contentious politics and policy (Armstrong & Bartley, 2007;McCarthy, 2013;McCarthy & Zald, 1977). Indeed, the assumption that political organizations like SMOs have an impact on policy is so pervasive "that we treat it as a core hypothesis of democratic politics" (Burstein & Linton, 2002, p. 384). In the social movements literature, scholars focus on networks of social movements organizations (SMOs) as key to collective action activities, as SMOs can mobilize material resources, create alliances, spread information and coordinate efforts (Crossley & Ibrahim, 2012;Diani, 1992). Additionally, they can shape the broadly held policy preferences of the movement (McCarthy & Zald, 1977). As Lohman argues (in Burstein & Linton, 2002, p. 387), SMOs can also play a key role in generating new information that policy makers need to respond to if they do not want their chances of re-election to suffer.
Burstein and Linton's seminal 2002 study examined the impact of SMOs, interest groups, political parties and public opinion on public policy and found that the impact that SMOs and interest groups have on policy is mediated by public opinion. They found that when public opinion is taken into account, political organizations' direct impact declines or disappears. SMOs affect policy directly to the extent that their activities provide elected officials with information and resources relevant to those policy makers' prospects for reelection.

Policy Networks, SMOs and Education Policy Research
In education policy research, the role of networks has been taken up primarily through the use of SNA as the methodological tool of choice. As mentioned previously, most of these studies assess the effects of social networks at the local or national level. For instance, the study of education policy networks has been increasingly used to unpack the dynamics of such policy arenas as foundation giving (e.g., Ferrare & Reynolds, 2015;Reckhow & Snyder, 2014), charter schools (Au & Ferrare, 2014;Kretchmar, Sondel & Ferrare, 2014;Scott, 2009), educational change at the local level (e.g., Daly & Finnigan, 2010), and national education reform (e.g., Baeket al., 2018). There are some studies that look at the impact of individual social linkages (i.e., social networks) as supporting or constraining the implementation of education reform (e.g., Daly & Finnegan, 2010;Daly, Moolenaar, Bolivar & Burke, 2010), but these studies analyze networks as social networks, not policy networks per se. Additionally there is limited research on the impact of SMOs on education policy; most of the social movements literature focuses on welfare and economic policy. For instance, in Uba's (2009) meta-analysis of 74 journal articles that reported results on the impact of SMOs or interest groups on policy change, only one had an education focus. The relevance of movement participation at the individual level remains under explored (Van Stekelenburg & Klandermans, 2013).
In this paper, we contribute to the growing literature on policy networks in education by connecting the literature on SMOs to the question of how SMOs may impact individual policy preferences. We highlight the role that networks play in education politics not only by mobilizing individuals to take action on contested policy issues, but also by informing individual views on policy. Understanding individual attitudes is important because research suggests that voter preferences are a cause, and not just an outcome, of government choice on policy. We believe that better understanding this process is important in order to fully leverage the relationship between policy networks, SMOs and policy change.

The Opt-out Movement
The opt-out movement, in which parents and caregivers refuse to have their children take federally-mandated annual standardized tests, came quickly to the fore in the mid-2010s as a prominent public challenge to test-based accountability measures in the United States. In 2009, forty-five states and the District of Columbia adopted the Common Core State Standards and their accompanying standardized assessments. These new assessments met with growing resistance and starting around 2014 individuals and groups mobilized to protest the new assessments. This mobilization became known as the opt-out movement (Hursh, Deutermann, Rudley, Chen & McGinnis, 2020;McDermott, Robertson & Jensen, 2014).
Research on the opt-out movement has been fairly limited, though some scholars have tried to understand the demographics and motivations of its participants as well as the movement's organization and structure. In a study of New York districts, Chingos (2015) found that after controlling for the number of students receiving free and reduced lunches, districts with lower test scores had higher opt-out rates. Bennett (2016) found that demographic differences influenced attitudes towards testing and that response to testing requirements varied widely by state. A national survey of opt-out activists from 47 states found that respondents were mostly parents or caregivers of school-aged children and almost half of the activists were teachers or educators (Pizmony-Levy & Green Saraisky, 2016). In a study of the Colorado opt-out movement, researchers used school-level data to demonstrate that widespread participation in opt-out was most prevalent in suburban and rural areas and in higher SES communities with high performing schools (Clayton, Bingham & Ecks, 2019). Additionally, scholars have looked at the role of women in the movement (Schroeder, Currin & McCardle, 2018).
In other research focusing on movement structure and organization, Wang (2017) mapped how the New York opt-out network forged coalition ties, framed messages and mobilized policy solutions. Wang describes the network structure in New York and suggests how disparate grassroots actors were mobilized through network ties and how their message gained coherence and political clout through the strength of the movement's organizational networks. Another study focusing on New Jersey highlighted that state's organizational network and found that three politically liberal parent-led groups collaborated to mobilize opposition to the PARCC test, and also worked across ideological lines with conservative groups to support anti-testing policy and legislation (Supovitz, Stephens, Kubelka, McGuinn, & Ingersoll, 2016). Paquin Morel (forthcoming) analyzed the discursive frames used by opt-out activists in New York. Using data from dedicated Facebook groups, Paquin Morel shows how activists use socially oriented frames in the initial stages of the movement and then moved to politically neutral frames in order to mobilize a broad base. Additionally, scholars have looked at public opinion about the opt-out movement (Pizmony-Levy & Cosman, 2017).
As with most social movements, the opt-out movement relied on SMOs to mobilize individuals and other resources to support its anti-testing efforts Hursh et al., 2020;Wilson, forthcoming). To our knowledge, there is little scholarship that examines the role of SMOs in the opt-out movement (Supovitz et al., 2016 andWang, 2017 being exceptions). Furthermore, scholars have yet to use the opt-out movement to explore networks and policy preferences. This study aims to address these gaps.

Data and Methods
Our analysis is based on data from the National Survey on Opting Out, an original survey program developed by the authors. The primary goal of the survey program was to explore the politics and mobilization of the opt-out movement. We sought to better understand why and how people became involved in the opt-out movement, as well as activists' perspectives on education policy and reform. We present data from two waves of surveys. The first survey was conducted from January 20 to March 31, 2016 (n=1,611). The second survey was conducted from March 7 to May 18, 2018 (n=1,298) to assess trends and changes in the movement over time.
Similar to previous research on social movement activism, we drew a non-probability sample of individuals affiliated with the opt-out movement. To minimize sampling bias, a list of national, state and local groups that maintained social media channels was constructed from web-based queries. After receiving permission from the various group administrators, surveys were shared electronically through the groups and their social media platforms. This included posting links to the survey on Facebook and Twitter. In addition, surveys were shared with colleagues and other individuals active in the movement. To expand the reach of the survey, messages about the survey included hashtags such as #optout, #optout2016, and #optout2018.
The survey instrument was designed as a web-based, self-administered questionnaire. Because there is limited research on the opt-out movement, we drew on a range of sources to inform the instrument, including scholarly research, media coverage, materials and artifacts produced by individuals and organizations affiliated with the movement, and consultations with key informants. Some items were taken directly from existing public opinion surveys (e.g., Phi Delta Kappa [PDK] Poll) that had previously asked respondents about standardized assessment and/or opting out. The final survey probed three broad areas. The first section gathered data on participant engagement with the opt-out movement (e.g., reasons for participation and modes of activism). The second section focused on attitudes toward education policy reform, and the use of testing and assessment in education. The third section gathered data on sociodemographic information.
Despite the richness of the data, two caveats should be mentioned. First, the study does not involve a random selection of respondents and this design could have implications for the representativeness of the sample. Second, and relatedly, although the sample includes participants from 47 states, close to half of respondents (46.5 percent) come from three states: New York, Florida, and New Jersey. All three states witnessed high rates of opting out, and their representation in the sample could be a simple reflection of the high mobilization. This pattern could have implications not only for the structure of the organizational network, which we discuss later in the paper, but also for policy preferences of individual activists (see sensitivity analysis). However even given these limitations, we believe the data can still provide important insights into the role of organizational networks in the opt-out movement.

Measures
Contact with social movements organizations (SMOs). The main variable in our analysis is whether or not respondents were contacted for activism by an organization. Following questions about engagement with opt-out activism (e.g., opting out and other forms of protest), the survey asks respondents "Have you ever been contacted by an organization to participate in any optout activities?" This is a binary variable coded 1 for yes and 0 for no. Respondents who answered positively were then asked to type in the names of all the organizations that contacted them ("If yes, which organization(s) have contacted you to participate in any opt-out activities?"). We discuss the analysis of this open-ended question later. This question reveals information about which SMOs participated in the mobilization of activists by contacting them regarding opting out. Contact could have come in a variety of forms, such as an e-mail, a post on social media, a phone call, or an inperson contact. We do not assume that these SMOs were entirely responsible for the individual's participation in the movement.
Attitudes towards educational reform. To assess the extent to which affiliation with SMOs could have implications for policy, the survey includes two sets of questions on policy preferences. Following Burstein (1998Burstein ( , 2003, we posit that public policy preferences serve as signals for policy makers about which ideas resonate with their electorate; these signals can translate to policy action. Both sets of questions are adapted from the PDK Poll of the Public's Attitudes Toward the Public Schools (2014,2015).
The first set of questions evaluates the perceived importance of five ideas for improving public schools in the community: (a) quality of teachers, (b) expectations for what students should learn, (c) effectiveness of principals, (d) how much money the schools have to spend, and (e) using tests to measure what students have learned. These variables are measured with a 4-point Likert-type scale ranging from 1 (not at all important) to 4 (very important).
The second set of questions evaluates the locus of governmental responsibility for five aspects of public schools: (a) paying for the K-12 public education system, (b) deciding what is taught in the school, (c) holding schools accountable for what students learn, (d) determining the right amount of testing, and (e) setting educational standards for what students should know. The responses for these variables are trichotomous: federal government, state government, and local school board.

Controls
In all multivariate models we include a host of socio-demographic controls including gender, race/ethnicity, age, education, income, employment status, teaching profession, and region of residence. Given the political nature of the opt-out movement, we also include political party identification. Descriptions of the variables, metrics, and descriptive statistics are outlined in Table 1.

Analytical Strategy
There are at least three ways to investigate the organizational structure -or network -of a given social movement. First, one can explore official websites of SMOs to examine hyperlinks to other SMOs. In this approach, two organizations are tied in the network if one organization refers to another organization. Another approach is to examine texts (e.g., news stories and press releases) to identify SMOs that are active in the movement. In this strategy, two organizations are tied in the network if they are mentioned as co-sponsors of an event or if they appear at the same event. A third alternative is to analyze the co-mobilization or co-recruitment of activists. In this approach, two organizations are tied in the network if they contact the same activists. All three modes of operationalizing organizational networks are commonly used in the literature. Given our research questions and our research design (a survey of activists) we use the third approach in this study.
We analyzed the structure of the opt-out movement using social network analysis (SNA) and prepared visual representation of the network using UCINET 6.289 and NetDraw 2.097 (Borgatti, Everett, & Freeman, 2007). We manually coded the open-ended responses to the question "which organization(s) have contacted you […]?" When possible, we distinguished between national organizations and state or local chapters, for example "United Opt Out" and "New Jersey Opt Out." Generic names, such as "local groups," were removed from the dataset. This process resulted in a two-mode matrix with respondents (in rows) and organizations (in columns) (for discussion and review see Wasserman & Faust, 1994). Following Heaney and Rojas' work (2007, 2008 on the U.S. Anti-War movement, we derived the organizational network from the co-contacts by organizations in our data. In other words, we converted the two-mode dataset into a one-mode dataset. Two organizations are tied in this network if they contacted the same individual to participate in opt-out activities. The meaning of a tie in this network is that the organizations have overlapping memberships (and perhaps overlap in ideology, strategy, and/or tactics).
The analytical approach we are taking here does have limitations. Our analysis is based on a sample of opt-out activists, thus the social structures we reveal are dependent on the individuals that participated in our surveys. One implication is that SMOs engaging many opt-out activists have a relatively stable position in the network across years, whereas SMOs engaging small numbers of optout activists will vary relatively more with respect to their position in the network.
In addition to SNA, we used logistic regression to examine the predictors of having contact with SMOs. We estimate models for each year separately (i.e., 2016 and 2018). In additional analysis (not reported, available upon request) we also estimated models with interaction terms to evaluate differences between years. We used ordinal logistic regression models to examine the perceived importance of ideas for improving public schools, and we used multinomial logistic regression models to examine the locus of governmental responsibility for aspects of public schools. In total we estimated three models. Model 1 includes only the measure of contact with SMOs. Model 2 adjusts for socio-demographic characteristics. Model 3 introduces an interaction term between contact with SMOs and year.

Results
As noted earlier, the role of SMOs in mobilizing activists varies across space and time. Indeed, descriptive statistics show that in 2016 one-third of respondents (33.2%) said SMOs contacted them to participate in opt-out activities, whereas in 2018 the figure dropped to one-fourth (25.4%). We first explore the organizational network that emerges from the interaction between individual activists and SMOs, and the extent to which this network changed between 2016 and 2018. We then examine how the likelihood of opt-out activists being mobilized by SMOs varies by socio-demographic characteristics and by year. We devote the last part of the analysis to the association between SMOs mobilization and the policy attitudes of opt-out activists.

The Organizational Network of Opt-out Activists
We examine patterns of connections between SMOs in the opt-out movement by following previous social movements scholars' suggestions that the "overlapping relationships between activists constitute the basis of relationships among organizations" (Heany & Rojas, 2008, p. 59; see also Diani, 2004). For example, if a respondent indicated that s/he was contacted by Organization A and Organization B, then there is a tie between these organizations because they have common target audience.
We begin by presenting the organizational network of the opt-out movement, as derived from the 2016 survey. Overall, activists mentioned a total of 167 organizations. Figure 1 presents the main component in the network, which includes 40 organizations (see Appendix A with names and additional information). Two organizations are tied in this network if they contacted at least the same two individuals to participate in any opt-out activity; the thickness of the lines reflects the strength of the tie between the organizations. At the heart of the diagram we find four central SMOs: United Opt Out (UOO), Bad Ass Teachers (BATs), New York State Allies for Public Education (NYSAPE), New York State United Teachers (NYSUT). UOO and NYSEP are focused on standardized tests, as reflected in their websites: UOO serves as a focused point of unyielding resistance to corporate education reform, test-centric educational practices, and the privatization and destruction of public education […] UOO's mission is to strengthen public education; fight corporate based reforms that are threatening the concept and existence of an educational system that is publicly funded, quality in nature and available to all; and, in particular, to end the practice of punitive, high-stakes and related activities that are fraudulently being used as "proof" of the incompetence of public education/teachers. NYS Allies for Public Education are parents, educators and community members who firmly believe in the power of public education and its fundamental link to the success of a thriving community and a transparent, democratic government. We believe excessive testing and inappropriate sharing of private student data without parent consent threaten the future of our students, our schools, and our state. While meaningful assessments are an essential component of a world-class education, the NYS Common Core (Next Generation Learning Standards) standardized assessments are aligned with unproven reforms neither supported by vigorous research nor vetted by educators and parents.
But other key organizations bring with them different agendas and ideas that go beyond standardized testing. For example, the mission of BATs is to "give a voice to every teacher who refuses to be blamed for the failure of our society to erase poverty and inequality" (badassteacher.org). NYSUT, which is a federation of more than 1,200 local unions, is dedicated to improving not only the working conditions of its members, but also their professions: "We're united in a common commitment to improve the quality of education and healthcare for the people of New York" (NYSUT, 2019). In addition to the four aforementioned SMOs, the network includes state and local chapters of opt-out and BATs groups, organizations dedicated to public education (e.g., Network for Public Education, and Save Our Schools [SOS]), teachers unions, and organizations dedicated to specific educational issues (Fair Test, Stop Common Core NYS). The Network for Public Education, for example, is an advocacy group whose mission is "to preserve, promote, improve and strengthen public schools for both current and future generations of students" (NPE, 2019). The composition of the network suggests that SMOs could engage opt-out activists on a myriad of educational issues and not only in issues related to standardized tests.
The organizational network of the opt-out movement in 2018 is somewhat different than the 2016 network, as illustrated in Figure 2. The main difference between the two years is the size of the network. In 2018, activists mentioned a total of 88 organizations (about half of the number in 2016). Only 46 organizations were mentioned in both 2016 and 2018 surveys (see Appendix A). Another important difference is the disappearance of many state and local chapters of opt-out and BATs groups. However, Figure 2 suggests that the heart of the organizational network remained stable across the years with the following SMOs remaining central: United Opt Out (UOO), Bad Ass Teachers (BATs), NYS Allies for Public Education (NYSAPE), and the Network for Public Education. The composition of the network suggests that in 2018 SMOs continued to engage optout activists on many educational issues, and not only on issues related to standardized tests.

Who Is More Likely To Be Contacted by SMOs?
Table 2 presents logistic regression models to examine the likelihood that activists say they were contacted by SMOs to participate in any opt-out activities. We begin with the results for the 2016 survey (Model 1). SMOs are more likely to engage men, older respondents, and residents of the Northeast. Moreover, SMOs are more likely to contact liberals and conservatives than politically 'middle of the road' respondents. For a more concrete interpretation of these results, Table 3 contains predicted probabilities based on the models in Table 2. For example, when other independent variables are at their mean levels, men have a .42 probability of having a contact with SMOs; the probability decreases to .32 for women. Respondents in the Northeast have a .39 probability of having a contact with SMOs, while the probability decreases in other regions (West = .30, South = .29, Midwest = .22). As for political ideology, liberals have a .39 probability of having a contact with SMOs, followed by conservatives with a .33 probability and others (middle of the road) with a .24 probability.
Next, we examine the results for the 2018 survey (Table 2, Model 2). Overall, the patterns are similar, except for gender and political ideology. SMOs engage men and women similarly. Importantly, in 2018, SMOs were more likely to contact liberals than conservatives and others (middle of the road). As illustrated in Table 3, when other independent variables are at their mean levels, liberals have a .29 probability of having a contact with SMOs, while the probability decreases for conservatives (.15) and for others (.22).
In supplementary analysis (not reported, available upon request), we estimated additional models with the pooled sample (i.e., a combination of 2016 and 2018 datasets). Specifically, we conducted a series of analyses in which we included interaction terms between year and selected socio-demographic variables. We began by exploring the interactions between gender and year, and region and year. These interaction terms were found to be statistically nonsignificant. We then explored the interaction between political ideology and year. The model confirmed that SMOs equally engaged liberals and conservatives in 2016, but not in 2018.

Organizational Network and Policy Preferences
The final step in our analysis focuses on the policy implications of the organizational network of the opt-out movement. We assess the extent to which activists who were contacted by SMOs hold different attitudes towards educational reform and policy than those who did not have contact with the SMO network. We argue that finding an association between contact with SMOs and attitudes towards educational reform is an important indication of the possible link between organizational networks in the opt-out movement and educational policy preferences. Table 4 presents a summary of ordinal logistic regression models predicting the perceived importance of five ideas for improving public schools in the community. For each outcome variable we estimated two models. Model 1 is a bivariate correlation between contact with SMOs and the outcome; Model 2 adjusts for socio-demographic controls. To better contextualize the results, Figure 3 contains aggregate results for the general public views (PDK, 2014) and predicted probabilities based on Model 2 in Table 4. Figure 3 facilitates two comparisons: (a) between the general public and opt-out activists, and (b) within the opt-out movement (based on a connection with SMOs). Overall, opt-out activists have different views than the general public on the importance of different ideas for improving public schools in the community. Compared to the general public, optout activists are less enthusiastic about setting standards (expectations for what students should learn) or using standardized tests; at the same time, opt-out activists are more enthusiastic about the role of school leadership (effectiveness of the principals) and resources (how much money the schools have to spend). Significance test for differences between opt-out activists with and without contact with SMOs: * p < 0.05, ** p < 0.01, *** p < 0.001 In three out of the five ideas, we find a statistically significant difference between opt-out activists who were mobilized by SMOs and other activists: (a) quality of teachers, (b) expectations for what students should learn, and (c) using tests to measure what students have learned. In the case of "expectations for what students should learn," the difference is large enough to be of value in a practical sense: opt-out activists with no connection to the organizational network have a probability of .57 to say this idea is very important; for opt-out activists with any connection to the organizational network the probability decreases to .49.
The locus of governmental responsibility is also a key point in the debate over educational policy in the U.S. Stakeholders hold very different views on the roles of the federal government, state government, and local school boards. Table 5 presents a summary of multinomial logistic regression models for where opt-out activists assign the main responsibility for five aspects of public schools. For each outcome variable we estimated the same models as before. To better contextualize the results, Table 6 contains aggregate results for the general public views (PDK, 2015) and predicted probabilities based on Model 2 in Table 5   Overall, opt-out activists are more likely than the general public to place responsibility for educational policy at the local school board or state government levels (versus at the federal level). The only exception is paying for the K-12 public education system. Here, we find little if any difference between opt-out activists and the general public. About half of the respondents place responsibility for funding on the state, and the rest place responsibility for funding on the federal government or local school board.
Opt-out activists who were mobilized by SMOs are more likely than their counterparts who were not mobilized by SMOs to attribute responsibility for educational policy to local school boards and state government (versus federal government). This pattern is evident in four out of five domains: (a) deciding what is taught in the school, (b) holding schools accountable for what students learn, (c) determining the right amount of testing, and (d) setting educational standards for what students should know. Moreover, the differences between activists persist when we adjust for sociodemographic characteristics. For example, opt-out activists with no connection to SMOs have a probability of .21 to place responsibility for "setting educational standards for what students should know" at the federal level, but this figure drops to .12 among opt-out activists with connections to the organizational network.
We conducted several methodological checks on our results to assess whether the patterns are sensitive to the sample composition, with 30.5 percent of the respondents residing in New York (not reported, available upon request). First, we estimated all models with the sub-sample of New York and with the sub-sample of other states. Second, we conducted a series of analyses in which we included an interaction term between state (New York vs. other states) and mobilization by SMOs. Patterns were consistent across all models, suggesting that the patterns are not driven by the large representation of activists from New York in the sample.

Discussion
Using data from two waves of the National Survey on Opting Out (2016 and 2018), we use social network analysis (SNA) in two distinct ways in this study. First, as a methodological tool for mapping the structure of an educational social movement, we use SNA to depict the structure of the opt-out movement. This allows us to discern which SMOs are central to bringing resources to the movement, especially in terms of potential mobilization to participation. Second, as a conceptual approach that focuses on relationships between actors, we use SNA to examine whether activists with a link to opt-out organizations hold different policy preferences than other activists (those without a link to opt-out organizations). This allows us to explore the potential role of SMOs in education policy.
The evidence described in this study shows that both national level (e.g., United Opt Out; BATs) and state level (e.g., SOS NJ; Washington State Against Common Core) organizations conducted major opt-out outreach efforts. The 2016 network identified organizations active in several states (New Jersey, Georgia, Michigan, Washington, Florida, Oregon) with New York State most heavily represented. In the 2018 network, several key national organizations remain while many of the state organizations are not named. This suggests that organizations in the Northeast, particularly in New York and New Jersey, continued their mobilization efforts while SMOs in other parts of the country were less active. Overall, our findings paint a picture of an organizational policy network that is relatively stable over time, shrinking slightly between the first and second waves of the survey, in 2016 and 2018 respectively.
These patterns have implications for the opt-out movement. The organizational network we identified includes SMOs that are dedicated to the particular contested policy issue of standardized testing, as well as SMOs that address standardized testing policy as part of a larger organizational agenda. Indeed, many of the SMOs in the network are concerned with education policy issues that go beyond standardized testing to issues of public education in general. For instance, while the Badass Teachers Association (BATs), which is central in the opt-out network in both 2016 and 2018, clearly states in its mission statement that it is against high-stakes testing to evaluate teachers, students and schools, that goal is just one of thirteen goals described in the organization's mission statement (BATs, 2019). BATs sees itself as dedicated to educational social justice broadly defined. We suspect that activists who are exposed to organizations in the opt-out network like BATs are being exposed to policy information and message framing not only about opt-out but about other education policy issues as well. In other words, the organizational network crates a space where activists can learn about multiple ideas and agendas regarding education.
The prominence of New York in the network is a reflection of our sample composition (one-third of all survey respondents are from New York) but may also reflect the importance of organizational density on activism. New York has consistently had the highest opt-out rates in the country, with state rates hovering around 20% during the last several years. Our data cannot directly address the question of whether NY opt-out rates are consistently high because of the organizational density of its network, or whether high opt-out rates or network structure came first. Nevertheless, the pattern suggests there could be a relationship between a stable SMOs network and high opt-out rates.
Not all opt-out activists have the same likelihood to be mobilized by SMOs. Our analysis shows that political ideology plays an important role. In 2016, self-identified liberals and conservatives reported on connections to opt-out SMOs at about equal rates, with very few 'middle of the road' participants reporting on such connections. This reflects the fact that the opt-out movement brought together, at least initially, concerns from across the ideological spectrum, with liberals concerned with the use of high-stakes testing in teacher evaluations and the growing role of corporate interests in education, and conservatives protesting against the growing role of the federal government in education policy (Pizmony-Levy & Green Saraisky, 2016). By 2018, the SMOs network is mobilizing fewer people overall and shows a significant drop among conservative respondents. This pattern might be related to broader shifts in the political context (e.g., the election of President Donald Trump and the appointment of Betsy DeVos as Secretary of Education). With a new federal administration, concerns that have motivated conservative activists in the past -such as the Common Core State Standards and the role of the federal government in education policywere perceived as "solved" (see Pizmony-Levy & Green Saraisky, forthcoming). In addition to political ideology, we also find that SMOs engage differently with men and women. Compared to women, men are more likely to report having contact with SMOs. Past research offers little insight into the relationship between gender and SMOs. Recent surveys show men and women are equally likely to be internet users (Perrin & Duggan, 2015) and that women are more likely than men to participate in education activism (Pizmony-Levy, Pallas & Green Saraisky, 2018). In other words, internet use and civic engagement on educational issues cannot explain the aforementioned gender gap. It is possible that women rely on other social networks or channels for participation (e.g., parent networks) while men are more likely to rely on organizational networks (Currin, Schroeder & McCardle, forthcoming). Future research could address this issue.
In this study we examined not only the organizational networks of the opt-out movement, but also a possible mechanism through which organizational networks may impact policy change: the policy preferences of activists. The evidence from our multivariate analysis suggests that activists who have network contact hold different policy preferences on a range of issues beyond those connected to opting out. The patterns are especially clear when we analyze views towards the locus of governmental responsibility for different aspects of public schools. Importantly, these patterns hold even after we adjust for individual characteristics. If activist policy preferences are associated with exposure to the broader educational agendas of the SMOs in the opt-out network, this link is important to understanding possible SMOs impact on education policy.
This further clarifies the conditions under which networks can matter for policy change. Past research has demonstrated how organizational networks are important for resource mobilization, collective identity, issue framing and capacity building, but there is less agreement about the political consequences of SMOs (Amenta, Caren, Chiarello & Su, 2010). While it is not the only determinant of political outcomes for SMOs, clearly one important channel is the SMOs ability to define problems, create frames and offer policy solutions that resonate with activists (Cress & Snow, 2000), thereby affecting activist policy preferences. The literature also differentiates between activists' preferences and those of general public opinion (Amenta, et al, 2010;Dencik & Cable, 2017). It is theorized that SMOs directly affect policy only insofar as they are providing policymakers with information that affects their re-election prospects, and that this SMO influence is reduced or disappears when public (majority) opinion is taken into account (Burstein & Linton, 2002). However, if SMOs contact is affecting activist policy preferences more broadly than on a contentious issue, SMOs can affect policy to the extent that they are shifting not only activist preferences but broader policy debates. At the individual level, the shift in activist preferences might also impact policy if activists vote for those policy makers who promise to support voters' preferred policy preferences. This is an initial examination of this possible link but it broadens our understanding of how a policy network can work at the grassroots rather than the elite level (e.g. lobbying and established interest groups) that much of the policy networks literature explores.
One limitation of the study is that we do not know how the initial tie between activists and SMOs was formed. It is possible that activists themselves, wanting to participate in opt-out related activities or for other reasons, might have reached out to one of the organizations directly. Another related limitation is that we do not know whether opt-out activists, once contacted by a SMO, participated in any of the opt-out activities promoted by the SMO. Indeed, future research should examine the links between contact with SMOs and modes of participation in the opt-out movement (e.g., contacting elected officials and policy makers).
Another direction for future research would be to explore the network of SMOs using information from sources other than surveys. For example, researchers could construct the organizational network using information on board members and staff. This would allow for exploring the interlocking relationships between SMOs and their role in facilitating change (or stability) within the opt-out movement. Yet another possible direction is to analyze news reports on the opt-out movement (similar to the work of Wang, 2017). Social-movements scholars have used this approach to study changes in SMOs collaboration and the role of networks in the diffusion of protest tactics (e.g., Wang & Soule, 2012;. Scholars and policy makers have recognized the importance of networks for education policy and implementation. The opt-out movement and the findings of our study suggest that networks operate not only at the macro-level (federal and state government), but also at the microlevel where activists and the public are the main actors. Scholars of policy studies should continue to pay attention to the grassroots networks that connect people, organizations, and agendas.

About the Authors Nancy Green Saraisky
Teachers College, Columbia University nlg2004@columbia.edu Nancy Green Saraisky is a research associate and adjunct assistant professor in the Department of International and Transcultural Studies. She received her PhD in political science and comparative education from Columbia University, where her studies were funded by the National Science Foundation. Recent projects focus on the relationship between data, politics and policy as it relates to educational assessment, both domestically and internationally. She also researches education activism and the rhetoric of education policy making.

Oren Pizmony-Levy
Teachers College, Columbia University op2183@tc.columbia.edu Oren Pizmony-Levy is an associate professor in the Department of International and Transcultural Studies. He holds a PhD in sociology and comparative and international education from Indiana University-Bloomington. His research and teaching focus on the intersection between education and social movements. One line of research examines the accountability movement and its role in the emergence of international assessments of student achievement (e.g., TIMSS, PIRLS, and PISA). Another line of research examines the Opt Out movement in the United States.

Wayne Au
University of Washington, Bothell wayneau@uw.edu Wayne Au is an educator, activist, and scholar who focuses on issues of race, class, and power in schooling. He is a professor in the School of Educational Studies at the University of Washington-Bothell, where he currently serves as dean of diversity and equity. Au  Readers are free to copy, display, distribute, and adapt this article, as long as the work is attributed to the author(s) and Education Policy Analysis Archives, the changes are identified, and the same license applies to the derivative work. More details of this Creative Commons license are available at https://creativecommons. Please send errata notes to Audrey Amrein-Beardsley at audrey.beardsley@asu.edu Join EPAA's Facebook community at https://www.facebook.com/EPAAAAPE and Twitter feed @epaa_aape.