Opting Out in the Empire State: A Geographic Analysis of Opting Out in New York, Spring 2015 & 2016


by Kathryn P. Chapman, Lydia Ross & Sherman Dorn - 2020

Background: Recently, states have experienced widely varying participation in annual assessments, with the opt-out movement concentrated in New York State and Colorado. Geographic variation between and within states suggests that the diffusion of opting out is multilayered and an appropriate phenomenon to explore geographic dimensions of social movements in education.

Purpose: The study analyzes the geographic patterns of opting out from state assessments in school districts in New York State.

Research design: We conducted linear regression and geographically weighted regression on district-level proportions of third- through eighth-grade students in local public school districts for 2015 and 2016 (n = 623), excluding New York City and charter schools. Independent variables included the district-level proportion of students with disabilities, identified as English Language Learners, and identified as White; census-based small-area child poverty estimates for the districts; and the geographic population density of the district. Linear regressions excluded racial and ethnic dummy variables to reduce collinearity problems, and geographically weighted regression limited geographically varying coefficients to child poverty and population density based on preliminary analyses.

Findings: The unweighted ordinary least squares (OLS) of district-level opting out in both spring 2015 and spring 2016 are weakly predictive as a whole (adjusted R2 < .20). In both years, population density was a statistically significant but low-magnitude predictor of change in opt-out behavior using OLS. The proportion of students with Individualized Education Plans was positively associated with opt-out behavior, and district-level child poverty was negatively associated with opt-out behavior. The proportion of White students was a statistically significant positive predictor of opt-out behavior in spring 2015 but not statistically significant for 2016, though with a coefficient in the same direction (positive). Analyzing the same data with geographically weighted regression more than doubled the adjusted R2 for each year and demonstrated that there were areas of New York State where the coefficients associated with child poverty and population density reversed direction, with suburban Long Island and the western upstate region as areas with a magnified negative association between district-level child poverty and opting-out percentages.

Conclusions: In the past five years, social networks have enabled the long-distance organizing of social and political movements in education, including opting-out and teacher walkouts. However, the long-distance transmission of ideas does not explain intrastate variations. In this study, geographically weighted regression revealed the local variations in relationships between opting-out and two key variables. Local networks still matter critically to social organizing around education.




INTRODUCTION


At first impression, American elementary and secondary students in 2014–2015 were more likely to skip state standardized assessments than in the prior decade, with an explicit argument that the standardized assessments were actively harmful to students and that teachers were being coerced into teaching specifically to the assessments. This mass departure from state standardized assessments quickly became known as the opt-out movement and had many teachers and parents as passionate and vocal supporters. Although whispers of the movement began in 2012, with reports that a number of school administrators and teachers were refusing to give new Common Core-aligned assessments to their students as part of a field-testing operation (Pizmony-Levy & Saraisky, 2016), the opt-out effort gained national prominence in 2014–2015. The state of New York became ground zero for the movement as approximately 200,000 students in total, and sometimes up to 89% of students in specific school districts, opted out of state-mandated standardized testing during the 2014–2015 academic year, propelling the state into center stage of the standardized test opt-out movement.


Much like other states throughout the United States, the opt-out movement started small in New York and grew gradually from parent to neighborhood, and from school to school district. The process that led to the opting-out pattern in New York was complex and multilayered, involving parents, teachers, administrators, and the state teachers’ union affiliate. A significant factor was the dramatic increase in the role of student test scores in teacher and other educator evaluations, in response to the state’s involvement in the federal Race to the Top grant competition and as mandated by state law. This change in personnel evaluations prompted protests by teachers, administrators, and parents (e.g., Winerip, 2011). Networking was involved at both the long distance and local levels, and tracing the involvement of parents demonstrates one dimension of this layered process. Jeannette Deutermann, a mother who observed how difficult homework and assessments had become for her son during the rise of the Common Core State Standards (CCSS), became an unlikely spokesperson for Long Island (Solnik, 2015). After speaking with education professionals at her son’s school and other parents in the school district who were experiencing similar challenges with their children, Deutermann turned to the Internet and Facebook. Through this research, she learned of parents from upstate New York who had chosen to opt their children out of the state standardized assessments. She dug a bit deeper and decided that opting her son out would be the best option to alleviate his anxiety (Solnik, 2015).


As she developed connections with parents and guardians throughout New York, and with United Opt Out, a then-small organization in Colorado, Deutermann gained a better understanding of the opt-out movement, the existence of a code in New York State for students who do not sit for state tests,1 and local arguments over tensions between testing and the developing needs of the students of her community. Through these connections and additional social networking, Deutermann founded Long Island Opt Out, an independent organization and Facebook group of over 24,000 concerned individuals. This story is similar throughout the United States as parents use their social networking abilities to research and discover information about both local and more widespread movements. In turn, they share their findings with other parents and guardians within their neighborhoods, schools, and school districts, and in approximately 90 independent Facebook groups2 (see also Pizmony-Levy & Saraisky, 2016).


Deutermann’s story is one of geographic proximity and transfer of power. She relied upon her individual resources or social capital to make connections with and influence members of the concerned public that were located geographically close to her or within her school district and state. Along with other concerned parents, guardians, and children, Deutermann challenged the New York State education department and fought against an assessment system that they perceived as unjust. Castells (2007) might describe the leadership position that Deutermann assumed and the role that Long Island Opt Out played within the greater opt-out movement as the expansion of a network society; Facebook and other Internet tools allow local and distant networks to feed into each other. We believe that by studying the development of the opt-out movement, specifically in New York, we will begin to understand the important role of geographic proximity within the greater opt-out movement, or how the local is critical in the spread of broader movements in education politics.


LITERATURE


The contemporary opt-out phenomenon is both geographically dispersed and uneven, as some states such as New York, Colorado, and Rhode Island have had a high percentage of students not take state assessments, while students in most other states continue to take state assessments (e.g., Bennett, 2016; Ujifusa, 2015a, 2015b). The pattern among states is puzzling. For example, the variation between states does not appear to correspond with union membership or political power; and though the New York State teachers’ union affiliate supported the opt-out movement (e.g., Gee, 2015; Karlin, 2015), that statewide affiliate support did not translate into uniform responses by parents across the state. While 15 states have some provision for opting out of state testing beyond medical exemptions (Croft & Lee, 2016), those states cluster in the West, and those policies do not appear strongly related to states that recently have had high opt-out rates.3 To understand this unevenness, we look to both the limited literature on the contemporary pattern of opting out and also to adjacent opt-out-like behavior of parents in withdrawing their children from public education activities.


CONTEMPORARY OPT-OUT MOVEMENT


The opt-out phenomenon in New York State began in the later years of the No Child Left Behind Act (NCLB) and associated regulations, as well as the implementation of state plans associated with the Race to the Top of 2010. Significant numbers of students opting out interfere with two key mechanisms tied to NCLB and New York State’s teacher evaluation policy, implemented as part of its Race to the Top project: (1) the reporting of disaggregated student achievement data and (2) the use of student achievement data as part of teacher evaluation systems. As explained below, the motivations on behalf of opt-out movement organizers sometimes included an explicit desire to interfere with these mechanisms, but the reasoning behind the network was complex and mixed different issues.


Attitudes Toward Testing and a Social Movement


The limited amount of literature on opting out includes both general information about attitudes toward testing and analyses of opting out as a social movement. In recent polling, a significant minority of both parents and the general public think that parents should have the option to remove public school students from state testing (Henderson, Peterson, & West, 2016; Phi Delta Kappan, 2016). While the percentages vary by question wording, sample, and subgroup, support for the option to avoid testing ranges in percentile from low 20s to mid 40s (also see Bennett, 2016, on race, ethnicity, and support for opting out). A survey of self-identified members of the opt-out movement showed that respondents that skewed toward opting out were more educated, White, liberal, and had a higher income than the country as a whole. Self-identified supporters of opting out were also deeply concerned about “teaching to the test” and the Common Core curriculum standards (Pizmony-Levy & Saraisky, 2016). To some extent, the findings of Pizmony-Levy and Saraisky are reflected in smaller-scale studies, such as in New Jersey (Supovitz, Stephens, Kubelka, McGuinn, & Ingersoll, 2016).


Several scholars have started to explore opting out as a social phenomenon. Szolowicz (2017) found that political spectacle theory was a useful lens of analysis. Mitra, Mann, and Hlavicik (2016) asserted that the opt-out movement was a reflection of the contested nature of schooling. Analyzing a single advocacy website, they argued that public schooling is a contested space within which opting out is an opportunity to contest certain values about schooling. Wang (2017) formally analyzed the social networks of major players in New York State as represented in dozens of press reports, including supporters from both the state testing system and the opt-out movement. In Wang’s analysis, the primary opt-out players in the state included six advocacy groups, six teachers’ unions (five locals plus the state affiliate), some parent–teacher organizations, as well as individual parents, guardians, students, and educators. She concluded that in both the social network analysis and in public debates, advocates of opting out formed a dialectical relationship with the networks of testing supporters: Each reacted to the other network, and as a consequence, advocates of opting out gained experience in forming persuasive arguments.


New York as Epicenter


In Harris and Fessenden’s (2015) analysis of opting out in The New York Times, they found that in the spring of 2015, approximately 200,000 third- through eighth-grade students, or 20%, refused to take state-mandated standardized tests. According to their research and that of Olenick (2015), that amount was approximately quadruple the number of students who refused in the previous academic year. In comparison, another journalist reported that approximately 1,000 students opted-out on Long Island in spring 2013, and in the 2014–2015 academic year, more than 65,000 students refused to take the state assessments (Solnik, 2015).


The spring 2015 statistic of 200,000 students opting out statewide also appeared in additional newspaper articles, in which the authors discussed how currently no federal sanctions would be placed on a state that had high opt-out rates (Clukey, 2015; Taylor, 2015). These authors also mentioned that it would be possible in the future for states to level sanctions on school districts, if the state-level education authorities discovered that the school districts had officially encouraged students to opt-out (Clukey, 2015; Taylor, 2015). This loss-of-funding threat was considered pointless by some education officials, as they struggled with realizing an effective way of pressuring superintendents to dissuade parents and guardians from participating in the opt-out movement (Taylor, 2015).


In August 2015, Clukey reported that the education commissioner of New York was working with her staff to create a public-relations toolkit for superintendents. This toolkit was to include legal arguments that superintendents could make to discourage parents and guardians from opting their children out of the state assessments (Clukey, 2015; Solnik, 2015). After these toolkits were distributed, the education commissioner and superintendents faced additional resistance as many school board members across the state had won seats basing their campaigns on support for the opt-out movement (Tyrrell, 2015).


New York State Unified Teachers (NYSUT), the state teachers’ union of New York, released a recorded phone call message during the 2014–2015 academic year that was credited for encouraging more than 160,000 of the 200,000 students to opt-out (Lankes, 2015). With the message’s success, and the financial support of grassroots education activists, the call went out again encouraging parents and families to opt their children out of the spring 2016 state assessments (Clukey, 2016). When the refusal data was released for the 2015–2016 academic year, the opt-out rate rose despite the best efforts of the New York State Education Department (NYSED).


Why was New York the center of opting out in 2015 and again in 2016? Wang’s (2017) analysis in New York focused on the network of social movements. That is not the only possible explanation for why New York became the most prominent state for opting out in spring 2015 and spring 2016. After the sudden increase in testing nonparticipation in spring 2014, the bulk of attention surrounding the opt-out movement focused on states with large non-test-taking behavior, such as New York, and informal data from advocacy organizations (e.g., Chingos, 2015; Harris & Fessenden, 2015). Newspaper articles written about the opt-out movement featured key players in the advocacy organizations and traced how the movement started in upstate New York and then made its way across the state to Long Island (Solnik, 2015). The end result was a dramatic increase in opting out for the 2014–2015 academic year, what Klein (2015) called civil disobedience. Others described the upsurge as the result primarily of efforts by the teachers’ unions in the state (e.g., Bennett, 2016; Brody, 2015).


While initial analyses of the opt-out movement (e.g., Chingos, 2015; Harris & Fessenden, 2015) used public school district-level data and took a brief look at the rate of opting out for students qualifying for free or reduced-price lunch, the authors did not dive deeper to identify the characteristics of the children that were opting out. They provided a global perspective of how the opt-out movement was affecting the state of New York. This global look or model used across the state gave one perspective, while looking specifically at the school districts and characteristics of students gives another.


A BROADER CONTEXT FOR OPTING OUT


The literature described above focuses on the last few years. But it would be a mistake to see the last few years of opting out as a singular event, especially when the recent phenomenon of opting out fits into at least three different contexts. One context is the question whether opting out in the United States in 2015 and 2016 is unique as a protest against standardized testing. A second context is where opting out fits into the broader issue of nonparticipation in public education. A third context is where recent events, especially in New York State, fit into the longer history of teachers’ unions and their relationships with parents and communities.


Test Nonparticipation Before 2015


In the United States there have been occasional protests or boycotts of test administrations before 2015, in the past decade most notably in Seattle in 2013 (e.g., Brenneman, 2013; Zubrzycki, 2013). Then, teachers at Garfield High School boycotted the administration of locally chosen tests that were separate from the state assessments. Seattle’s administrators chose the Measures of Academic Progress (MAP) as an additional layer of assessment beyond the Washington state annual tests. Commonly marketed as a so-called benchmark test for use several times a year, MAP became a target of teachers who perceived the time devoted to the additional testing as direct competition with their own instruction and their autonomy in the classroom. Other, smaller boycotts occurred in New York in 2001 (Shin, 2001; also see Rothstein, 2001). Internationally, more widespread attempts to boycott tests occurred in England and Wales in 1993 and again in 2004 (e.g., The Sats story, 2004), and in British Columbia in 2009 (Poole, 2015).


Opting Out as a Partial Withdrawal From Participation in Public Education


The opt-out phenomenon of 2015 and 2016 fits in the narrow context of test boycotts. But one can also look at opting out as part of a broader repertoire of student and family withdrawal from participation in public education. Parents upset with local public schools have resorted over the years to a range of tactics and strategies in response. For example, there is a long history of boycotts of schools as part of civil rights protests either around access to educational opportunities or broader civil rights issues (e.g., Barrera, 2004; Bernal, 1998; Danns, 2002, 2003; Mabee, 1968; Petrzela, 2016). The boycott of schools is a tactic of explicit withdrawal of attendance to pressure local public schools. There is also the permanent withdrawal of students to place them in private schools, from the creation of parallel networks of Catholic schools in the 19th century to the creation of segregation academies by White racist parents who withdrew their children from local public schools during the civil rights movement (e.g., Fuquay, 2002; Lazerson, 1977; Ravitch, 2000; Taeuber & James, 1982; Walder & Cleveland, 1971). More recently, homeschooling is a withdrawal from formal schooling entirely, not just public education, based on a variety of motives that parents have for homeschooling their children (e.g., Isenberg, 2007; Kunzman & Gaither, 2013). Taken in this context, opting out of state testing is a much milder form of withdrawal from participation in the common processes of public education. When focused on the state test, one could see this as a symbolic withdrawal of legitimacy from state accountability systems as well as a mechanism to gum up the gears of accountability.  


Teachers Unions and Opting Out


The previous two paragraphs contrast the role of teachers unions (in attempted boycotts of tests in the United Kingdom and British Columbia), on the one hand, with the role of parents and sometimes students (in withdrawal from schools), on the other. In New York State, the opt-out movement was advanced through the actions of both parents and teachers, in collaboration. As the history of teachers unions shows, that collaboration is neither rare nor guaranteed (e.g., Goldstein, 2014; Murphy, 1990; Perlstein, 2004). Teachers have had their own form of nonparticipation in schooling: strikes. When a teachers union strikes, an essential component of the political dynamic is the union’s relationship with communities, especially parents. From that historical experience, it is a short step for teachers unions to encourage nonparticipation in tests by parents, as long as that nonparticipation is temporary and does not translate into broader withdrawal of participation in public schools. For a teachers union, a strike is a temporary, contingent withdrawal of participation to pressure management for specific ends; striking teachers do not want to leave their employment permanently. This is different from the broad range of nonparticipation for parents and is close only to attendance boycotts and the opt-out movement.


GEOGRAPHIC ANALYSIS TO COMPLEMENT SOCIAL NETWORK ANALYSIS


The work of Wang (2017) and others to identify the social networks that shaped opting out in New York State is important to identify themes and issues that are salient to those active in local and state opt-out movements. This work suggests that these networks allowed the spread of ideas about parent concerns with testing and the uses of testing, especially around teacher evaluation and the wave of accountability policies associated with Race to the Top and the Obama administration’s NCLB waivers. These concerns could appear in any school district and any state. What is not explained by the literature on the opt-out movement’s social networking is variation. Why did New York become a hotspot for opting out, and within New York, why did some school districts have far higher proportions of students opting out than others? The networking nature of the opt-out movement explains how ideas can have long reaches, but generating a critical mass of parents who withdraw their children from school in test weeks requires local organizing. To understand the patterns of local variation, we must turn to geographic analysis.


DATA SOURCES


We used data from the NYSED to analyze opt-out behavior in New York. The NYSED releases district-level information on test-taking rates for third through eighth grade each year (New York State Education Department, 2017). For this study, we focused our analysis on data from the spring 2015 and spring 2016 testing cycles. For each district, the reports include total number of students eligible for testing and the number of students who refused testing, which is further broken down by English Language Learners (ELLs), students with disabilities, and economically disadvantaged students. Charter schools were omitted from the analysis. After merging all of the data source files, we had 623 school districts left for analysis. It is important to note that the NYSED did not report test-taking data for New York City. Due to the lack of data, and because New York City had remarkably high test-taking rates in comparison to the rest of the state, we excluded New York City from analysis in this study.


To aid in our analysis, we also collected demographic information about the school districts in New York. The NYSED reported this demographic data for the 2011–2012 academic year. We first collected race and ethnicity data for each district. We also collected data on categorical program participation for the 2014–2015 academic year, specifically the number of students with disabilities and those students identified as ELLs in each school district. Finally, we utilized small-area poverty estimates for school districts from U.S. Census data (2014). The data from this file includes the percentage of children in poverty in each school district.


For the geographical analyses of this study, we used TIGER/Line shapefiles (2014). The shapefiles provided geographical boundaries of school districts in the state of New York, which we used to create the maps and conduct analyses.


METHODS


Our analysis followed two approaches: a standard regression approach that assumed the relationships between district characteristics and opt-out percentages were constant across the state, and a geographically weighted approach that assumed those relationships could vary by geographic region. In both approaches, this study models district-level spring 2015 and spring 2016 opt-out behavior, with the proportion opting out selected as the dependent variable. For each year, we began with a standard linear regression as the first analysis, with student proportion White, student proportion with Individualized Education Plans (IEPs), Census-based district child poverty estimates, and district population density as predictor variables. In both years, district-level ethnic–racial proportions were highly collinear (VIF values: percent White = 27.7, percent Hispanic = 10.6, and percent Black = 8.3), and in choosing among proportions White, Latino, and Black, we chose White as the indicator most consistent with the hypothesis that White privilege was associated with higher opting out at the district level. In the results for ordinary least squares (OLS) regression, the coefficients indicated a predicted relationship at the district level.


Because opting out spread as a social movement among parents and teachers, we investigated geographic variations in those relationships as evidence that the social-movement nature of opting out played out in local dynamics, analyzing that potential variation through geographically weighted regression (GWR) (e.g., Fotheringham, Brunsdon, & Charlton, 2002). GWR is a form of locally weighted regression, a general term that refers to predicted relationships that vary by the value of a predictor variable. In general, locally weighted regression estimates outcomes (and thus relationships) along a predictor variable’s range by considering only the cases that are near each (predictor) value, weighting cases closest to the (predictor) value most heavily. This kernel (or nearby) weighting is usually a formula that quickly falls to zero outside the neighborhood of each (predictor) value calculated in locally weighted regression. While most locally weighted regression is along one dimension at a time, GWR is a two-dimensional version, where local kernel weights are set by distance on the plane (local geography).


In practice, GWR requires the pre-selection of variables that are likely to vary across space, leaving others with fixed regression coefficients. Thus, GWR results in a combination of spatially static regression coefficients (as in a standard linear regression) and also geographically varying coefficients for a subset of predictor variables. For the geographically varying predictors, there is both a point estimate and standard error for each data location, which in this analysis was local school districts. However, because of the variable pre-selection and the simultaneous calculation of several hundred coefficient point estimates and standard errors, it is common to view the results of GWR outside of a null-hypothesis-testing framework.


There are other methods of analyzing geographically varying relationships, especially if there is a reason to believe that the geographic relationships are linear in one direction, or where there is a theoretical justification to lump together geographic units that can be treated as dummy variables for analysis (Fotheringham et al., 2002). GWR is thus most appropriate for cases where there is not an a priori justification for other methods, and where one suspects geographic variation but the nature of that variation is not clear in advance. The mechanics of GWR assume that relationships at any point in space are similar to that point’s neighbors—in this case, that nearby school districts would have similar relationships in opt-out behavior—but that there is no clear reason to lump points together in advance. In the case of opting out, the social movement in 2015 and 2016 suggests that local dynamics were important in organizing opting out, but there is not enough information to group New York districts in particular ways.


Based on preliminary analyses, we selected two predictors as the most likely to vary geographically: district-level child poverty and population density.4 The final GWR results thus include two predictors where coefficients did not vary geographically (student proportion White and student proportion with IEPs) and two where GWR was used to predict specific coefficients for each school district outside New York City (child poverty estimates and district population density). For weighting in local coefficient estimates, we used a flexible natural neighbor approach, minimizing the Akaike information criterion with a finite-sample correction (AICc).5


FINDINGS


GENERAL OPT-OUT BEHAVIOR


The opt-out rate for New York school districts varied considerably in spring 2015 and spring 2016. The average opt-out rate for districts for spring 2015 was 27.5% and increased in spring 2016 to 29.1%.6 The minimum opt-out rate changed from 0% in 2015 to 1% in 2016, while the maximum opt-out rate by district was 89% for both years.


To better understand the geography of opt-out behavior, we created thematic maps of opt-out proportions by district for 2014–2015 and 2015–2016 (see figure 1) using ESRI (2014) (also known as ArcGIS). At a first glance, in both spring 2015 and 2016, opting out appeared to be most prevalent in suburban Long Island, though there are swaths of upstate New York where opt-out percentages were relatively high.


Figure 1. Opt-out proportions. Spring 2015 on left, spring 2016 on right. Legend for both images.

[39_23062.htm_g/00002.jpg][39_23062.htm_g/00004.jpg]

[39_23062.htm_g/00006.jpg]


STANDARD LINEAR REGRESSIONS


We conducted two OLS regressions to specify the relationship between opt-out behavior and selected predictors (student proportion White, student proportion with IEPs, district population density, and proportion child poverty) for opt-out percentages at the district level in both spring 2015 and spring 2016.7 An OLS approach to opting out within a state assumes that relationships do not vary geographically.


For the 2014–2015 academic year, the model was only weakly predictive (adjusted R2 = .108) of opt-out behavior. The selected independent variables accounted for 11% of the variance in opt-out behavior. The coefficients are reported in Table 1.  


Table 1. Linear Regression, New York School District Opt-Out Proportions, Spring 2015


Variable

Coefficient

SE

t

p

VIF

Constant

0.15

 0.05

3.24

.001

 

Population density

< 0.01

 < 0.01

2.42

.016

1.74

Proportion White students

0.17

 0.04

3.84

< .001

1.67

Proportion students with IEPs

0.49

 0.20

2.49

.013

1.14

Proportion child poverty

–0.57

 0.08

–6.90

< .001

1.24


All of the independent variables were statistically significant predictors of opt-out behavior in New York school districts for spring 2015 (at p < .05). Though statistically significant, population density did not result in a large change in opting out, with an estimated coefficient of less than 0.001. Proportion students White and proportion students with IEPs were both positively associated with opt-out behavior, with each percentage point resulting in a 0.17 and 0.49 percentage-point increase in opt-out rate, respectively.8 Proportion child poverty was negatively associated with opt-out behavior, as for each percentage-point increase in child poverty, opt-out behavior was predicted to decrease 0.57 percentage points.


The regression model for the 2015–2016 academic year was also only weakly predictive (adjusted R2 = .163) of opt-out behavior. The predictors accounted for 16% of the variance in opt-out behavior. Table 2 shows the coefficients from the model.


Table 2. Linear Regression, New York School District Opt-Out Proportions, Spring 2016


Variable

      Coefficient

 SE

          t

             P

VIF

Constant

0.29

0.05

5.95

.000

 

Population density

0.00

0.00

3.31

.001

1.74

Proportion White students

0.07

0.04

1.590

.112

1.67

Proportion students with IEPs

0.41

0.20

2.06

.040

1.13

Proportion child poverty

–0.77

0.08

–9.19

.000

1.24


The results from the spring 2016 regression model were very similar to the spring 2015 regression model. Across all the variables, the direction of the relationships remained the same. Population density remained a statistically significant but low-magnitude predictor of change in opt-out behavior. The proportion of students with IEPs also was again a statistically significant predictor that was positively associated with opt-out behavior, and the coefficient was slightly higher in spring 2016 than in spring 2015 (0.49 and 0.41, respectively). Proportion child poverty was still negatively associated with opt-out behavior, and the magnitude of the coefficient was higher in spring 2016 than in spring 2015 (–0.77 and –0.57, respectively). Proportion of White students was not a statistically significant predictor of opt-out behavior in spring 2016 (p > .05), but it was still positively associated with opt-out behavior.


In the regression models, the constant was higher for spring 2016 than spring 2015 (0.29 and 0.15, respectively). The increase in constant indicates the increase in opt-out rates across New York school districts from spring 2015 to spring 2016, with all independent variables holding constant.


GEOGRAPHICALLY WEIGHTED REGRESSIONS


To better understand opt-out behavior in New York school districts, we then conducted GWR. As discussed earlier in the methods section, GWR extends the traditional regression framework by allowing for local parameters of the variables to vary (Fotheringham et al., 2002).


For both years, the GWR resulted in a substantial increase in accounted variance in the models compared with the OLS models with spatially static coefficients. For the 2014–2015 academic year, the adjusted R2 = .355 with the GWR, which approximately tripled the variance accounted for in the model. The accounted variance also increased considerably for the 2015–2016 year, with an adjusted R2 = .475. Both of these models included all of the variables from the standard regressions presented in the previous section.


Also as discussed in the methods section, we included only child poverty and population density as geographically varying coefficients in our analysis. In figures 2, 3, and 4, we present natural neighbor maps that show opt-out behavior in relation to child poverty and population density by year.9


Figure 2. Geographic relationships between district child poverty and opting out. Spring 2015 on left, spring 2016 on right. Legend for both images

[39_23062.htm_g/00008.jpg][39_23062.htm_g/00010.jpg]

[39_23062.htm_g/00012.jpg]


Figure 3. Geographic relationships between district child poverty and opting out. Spring 2015, highlighting differences between Westchester County and Long Island. Legend for image.


[39_23062.htm_g/00014.jpg]


[39_23062.htm_g/00016.jpg]


Figure 4. Geographic relationships between district population density and opting out. Spring 2015 on left, spring 2016 on right. Legend for both images.

[39_23062.htm_g/00018.jpg][39_23062.htm_g/00020.jpg]

[39_23062.htm_g/00022.jpg]


Dark gray areas of the maps indicate estimates of local coefficients with a more positive relationship between child poverty and opting out or population density and opting out, and light gray areas indicate estimates of local coefficients with a less positive or more negative relationship between child poverty and opting out or population density and opting out. Only the dark gray and light gray areas of the maps should be considered discernible from zero. This analysis already includes the standard linear regression with its child poverty coefficient. Long Island and the western tip of the state (southwest of Buffalo) are areas where the existing negative correlation is magnified. Westchester County, Rockland County, and the area near Plattsburgh show a reverse tendency, with a much less negative (and in some areas positive) relationship between local child poverty and opting out. We highlight this difference between Long Island and Westchester County in Figure 3.


DISCUSSION


Viewed as analysis supplemental to OLS, GWR suggests that child poverty and population density had a relationship with opting out that varied by where public school families lived in New York. Although news analyses have discussed the potential relationship of opting out and geography (e.g., Harris & Fessenden, 2015), academic opt-out research has overlooked geography, a proven important variable in our study. When we accounted for geographical variance in our models, we nearly tripled the variance accounted for, a result that speaks to the potential use of GWR as an analytical tool. While the data was still noisy, we found a moment of clarity by using GWR: geography mattered in the patterns of opting out.


Social networks, social media, and the use of social capital were essential tools for parents and guardians in the opt-out movement in New York. These social networks allowed the idea of opting out to cross long distances. This long-distance transmission of ideas is important in education politics. Recently, social media platforms such as Facebook have aided not just in the states of New York and Colorado with the opt-out movement, but they have also facilitated organizing in the states of West Virginia, Oklahoma, Kentucky, and Arizona in their teacher walkout events and organizations (e.g., Bell, 2018; Daniels, 2018; Levenson, Mezzofiore, & Williams, 2018; Virtanen & Raby, 2018).


However, the long-distance transmission of ideas does not explain the specifics of how students and their families behave. In New York, neither long-distance social networks nor the observed demographic characteristics of districts explain much of the local variation in opting out in 2015 and 2016. There is something inherently local about opting out, a phenomenon we think is demonstrated by the varying relationships between key social indicators within a district and the behavior of students and their families. When child poverty is related to opting out differently in suburban Long Island than in Westchester County, with different signs, the local nature of organizing behavior is as important as either general demographic characteristics or the long-distance social networking that is evident at the state or national level.


As illustrated through this study, GWR can be a critical tool in understanding how the opt-out movement varied across the state of New York. In the future, GWR along with qualitative methods could be used to study how other education policy movements, such as school choice programs, admissions policies, and curriculum adoption, expand throughout and between states over time. This research could aid agencies as well as education researchers and practitioners in determining how education policies are or are not adopted by different counties and communities of people throughout a state.


We see the case of New York as one part of a larger, nationwide opt-out movement. Although we found geographic areas and specific school districts in New York that could be categorized as supportive or not supportive of the movement, we know that may not be the case for all states. Through future research, we may find that some states are more uniformly in support of or opposition to the opt-out movement. By expanding our research into other states we will begin to further understand the potential impact of geography on the opt-out movement more broadly.


Notes


1. The “999 code,” a score of 999, or “not tested” is used in place of a final assessment score for any student who refuses to take a state standardized assessment (The University of the State of New York, 2013).


2. Facebook search term was “opt-out.” Authors carefully analyzed the group titles to determine relevance.


3. New York and Rhode Island are not among the opt-out provision states, while Utah has had high state participation while requiring notification to parents of opt-out provisions.


4. For the analyses, we used GWR 4.0 (Nakaya, 2015), which provides both exploratory and final diagnostic measures. Based on advice from expert GWR practitioners, we limited the geographically varying predictor set to those variables where geographic variability was most likely to make a difference in a range of model fit indicators (including AICc) (S. Fotheringham, personal correspondence). We selected the adaptive bisquare (aka Tukey biweight) for the geographic kernel function.


5. In general, the Akaike information criteria (AIC) compares models by the information lost when assuming that the model generates the data in question; this includes both the number of model parameters and the model likelihood function. AICc provides a small-sample correction for this approach. Neither AIC nor AICc are goodness-of-fit measures and are more appropriate in model selection—or, in this case, to pick the kernel size in a model to minimize AICc.


6. The unit of analysis here is the district (not student levels). Each district is weighted equally in the averaging of opt-out rates for the state. These figures overstate the actual opt-out rate of the entire state, as the figures were calculated without weighting the opt-out rate by enrollment in the district.


7. The OLS regression results presented in this paper utilize an unweighted regression. Rerunning the regression with district enrollments as weights does not substantially change the results.


8. “Percentage-point” explanations here are algebraically identical to using proportions for both dependent and independent variables, and are used for readability.


9. For both of the figures, it is important to note that New York City was excluded from this analysis, and any mapping of data in New York City is an artifact of the statewide map.


Acknowledgement


We would like to thank Dr. A. Stewart Fotheringham for his advice on this project; he is not responsible for findings or conclusions.


References


ArcMap 10.3 (2014) [Computer software. Redlands, CA: ESRI (Environmental Systems Resource Institute. 


Barrera, B. (2004). The 1968 Edcouch-Elsa High School walkout: Chicano student activism in a South Texas community. Aztlán: A Journal of Chicano Studies29(2), 93–122.


Bell, M. (2018, March 2). Oklahoma teachers coordinate walkout on social media. ABC Tulsa. Retrieved from http://ktul.com/news/local/oklahoma-teachers-coordinate-walkout-on-social-media


Bennett, R. E. (2016, April). Opt out: An examination of issues (Research Report No. RR-16-13). Princeton, NJ: Educational Testing Service. Retrieved from http://onlinelibrary.wiley.com/doi/10.1002/ets2.12101/full  


Bernal, D. D. (1998). Grassroots leadership reconceptualized: Chicana oral histories and the 1968 East Los Angeles school blowouts. Frontiers: A Journal of Women Studies19(2), 113–142.


Brenneman, R. (2013, February 5). Test boycott puts Seattle teachers in national spotlight. Education Week. Retrieved from http://www.edweek.org/ew/articles/2013/02/06/20testing.h32.html


Brody, L. (2015, April 10). N.Y. teachers union robocalls members about state tests. The Wall Street Journal. Retrieved from http://www.wsj.com/articles/n-y-teachers-union-robocallsmembers-about-statetests-1428681388


Castells, M. (2007). Communication, power and counter-power in the network society. International Journal of Communication1(1), 238-266.


Chingos, M. (2015). Who opts out of state tests? Washington, DC: Brookings Institution. Retrieved from http://www.brookings.edu/research/papers/2015/06/18-chalkboard-who-opts-out-chingos


Clukey, K. (2015, August 25). Education commissioner begins plan to combat opt outs. Politico. Retrieved from http://www.politico.com/states/new-york/albany/story/2015/08/education-commissioner-begins-plan-to-combat-opt-outs-024920


Clukey, K. (2016, April 4). Opt-out fight continues as state tests begin. Politico. Retrieved from https://www.politico.com/states/new-york/albany/story/2016/04/opt-out-fight-continues-as-state-tests-begin-033073


Croft, M., & Lee, R. (2016, June). Survey and analysis of state opt-out and required test participation legislation (ACT Working Paper 2016-03). Retrieved fromhttp://www.act.org/content/dam/act/unsecured/documents/Working-Pap er-2016-03-Survey-and-Analysis-of-State-Opt-Out-and-Required-Participation.pdf


Daniels, M. (2018, April 19). How social media helped teachers organize in Arizona, across nation. AZ Central. Retrieved from https://www.azcentral.com/story/news/local/arizona-education/2018/04/19/social-media-helped-teachers-organize-arizona-across-nation/532318002/


Danns, D. (2002). Black student empowerment and Chicago: School reform efforts in 1968. Urban Education37(5), 631–655.


Danns, D. (2003). Chicago high school students’ movement for quality public education, 1966–1971. The Journal of African American History88(2), 138–150.


Fotheringham, A. S., Brunsdon, C., & Charlton, M. (2003). Geographically weighted regression: The analysis of spatially varying relationships. West Sussex, UK: John Wiley & Sons.


Fuquay, M. W. (2002). Civil rights and the private school movement in Mississippi, 1964–1971. History of Education Quarterly42(2), 159–180.


Gee, D. J. (2015, March 30). NYSUT president urges parents to opt out of state tests. Buffalo News. Retrieved from https://buffalonews.com/2015/03/30/nysut-president-urges-parents-to-opt-out-of-state-tests/


Goldstein, D. (2014). The teacher wars: A history of America’s most embattled profession. New York, NY: Doubleday.


Harris, E. A., & Fessenden, F. (2015, May 20). “Opt-out” becomes anti-test rallying cry in New York State. The New York Times. Retrieved from http://www.newyorktimes.com/2015/05/21/nyregion/opt-out-movement-against-common-core-testing-grows-in-new-york-state.html


Henderson, M., Peterson, P. E., & West, M. R. (2016, Winter). The 2015 EdNext poll on school reform. Education Next, 16.


Isenberg, E. J. (2007). What have we learned about homeschooling? Peabody Journal of Education82(2–3), 387–409.


Karlin, R. (2015, March 3). Going on the offense, NYUT’s Magee calls for test boycott [Web log post]. Retrieved from https://blog.timesunion.com/capitol/archives/231494/going-on-the-offense-nysuts-magee-calls-for-test-boycott/


Klein, G. (2015, April 30). CCS BOE approves budget. The Daily Star. Retrieved from http://www.coopercrier.com/news/ccs-boe-approves-budget/article_61813884-6398-5b83-8d62c99e0ca217b8.html


Kunzman, R., & Gaither, M. (2013). Homeschooling: A comprehensive survey of the research. Other Education, 2(1), 4–59.


Lankes, T. (2015, April 30). NYSUT chief Magee gives union more-aggressive public face. The Buffalo News. Retrieved from http://buffalonews.com/2015/04/30/nysut-chief-magee-gives-union-more-aggressive-public-face/


Lazerson, M. (1977). Understanding American Catholic educational history. History of Education Quarterly17(3), 297–317.


Levenson, E., Mezzofiore, G., & Williams, D. (2018, April 3). These crumbling textbooks show why Oklahoma teachers are walking out. CNN. Retrieved from https://www.cnn.com/2018/04/03/us/oklahoma-teachers-textbooks-trnd/index.html


Mabee, C. (1968). A Negro boycott to integrate Boston schools. The New England Quarterly, 41, 341-361.


Mitra, D., Mann, B., & Hlavicik, M. (2016). Opting out: Parents creating contested spaces to challenge standardized tests. Education Policy Analysis Archives, 24(31).


Murphy, J. (1990). Blackboard unions: The AFT and the NEA, 1900–1980. Ithaca, NY: Cornell University Press.


Nakaya, T. (2015). Geographically weighted generalized linear modeling. In C. Brunsdon & A. Singleton (Eds.), Geocomputation: A Practical Primer. Thousand Oaks, CA: Sage


New York State Education Department. (2017). English language arts (ELA) and mathematic assessment results. Retrieved from http://www.p12.nysed.gov/irs/ela-math/


Olenick, J. (2015, August 22). Despite opt-outs, state tests show there’s room for improvement. Lockport Union-Sun & Journal. Retrieved from http://www.lockportjournal.com/news/local_news/despite-opt-outs-state-tests-show-there-s-room-for/article_bcb3c519-2b61-50f1-ac87-eb7a5ac91931.html


Perlstein, D. H. (2004). Justice, justice: School politics and the eclipse of liberalism. New York, NY: Peter Lang.


Petrzela, N. M. (2016). Classroom wars: Language, sex, and the making of modern political culture. New York, NY: Oxford.


Phi Delta Kappan. (2016, September). Why school: The 48th annual PDK poll of the public’s attitudes toward the public schools. Retrieved from http://pdkpoll2015.pdkintl.org/wpcontent/uploads/2016/08/pdkpoll48_2016.pdf


Pizmony-Levy, O., & Saraisky, N. G. (2016, August). Who opts out and why? Results from a national survey on opting out of standardized tests (Research Report). New York, NY: Teachers College, Columbia University. Retrieved from http://www.tc.columbia.edu/media/news/docs/Opt_Out_National-Survey----FINAL-FULL-REPORT.pdf


Poole, W. (2015). Defending teachers’ rights and promoting public education: Evolving and emerging union strategies within a globalized neoliberal context. In N. Bascia (Ed.), Teachers unions in public education: Politics, history, and the future (pp. 33–52). New York, NY: Palgrave Macmillan.


Ravitch, D. (2000). The great school wars: A history of the New York City public schools. Baltimore, MD: Johns Hopkins.


Rothstein, R. (2001, May 30). A rebellion is growing against required tests. The New York Times, p. B9. Retrieved from LexisNexis.


Shin, P. H. B. (2001, May 5). More kids to boycott state tests. New York Daily News, p. 6. Retrieved from LexisNexis.


Solnik, C. (2015, August 3). Education’s Erin Brockovich. Long Island Business News. Retrieved from http://libn.com/2015/08/03/educations-erin-brockovich/


Supovitz, J.A., Stephens, F., Kubelka, J., McGuinn, P., & Ingersoll, H. (2016). The bubble bursts: The

2015 opt-out movement in New Jersey. CPRE Working Papers. Retrieved from http://repository.upenn.edu/cpre_workingpapers/13


Szolowicz, M. A. (2017). Opt out! Understanding resistance to the Common Core’s testing regime through political spectacle. (Doctoral dissertation, University of Arizona). Retrieved from http://hdl.handle.net/10150/622994


Taeuber, K. E., & James, D. R. (1982). Racial segregation among public and private schools. Sociology of Education, 55(2–3), 133–143.


Taylor, K. (2015, August 20). New York schools with many students who skipped tests won’t lose money. The New York Times. Retrieved from https://www.nytimes.com/2015/08/21/nyregion/new-york-schools-with-many-students-who-skipped-tests-wont-lose-money.html


The Sats story. (2004, August 24). The Guardian. Retrieved from https://www.theguardian.com/education/2004/aug/24/schools.sats


The University of the State of New York. (2013). New York State student information repository system (SIRS) manual: Reporting data for the 2013–14 school year. Albany, NY: The State Education Department Information and Reporting Services. Retrieved from http://www.p12.nysed.gov/irs/sirs/2013-14/2013-14SIRSManual9-4_20131206.pdf


TIGER/Line shapefiles [Machine readable data files].(2014). Suitland-Silver Hill, MD: U.S. Census Bureau. Retrieved from https://lib.asu.edu/GIS/repository.


Tyrrell, J. (2015, May 20). Many candidates endorsed by LI Opt-Out group win seats. Newsday. Retrieved from http://www.newsday.com/long-island/many-candidates-endorsed-by-li-opt-out-group-win-seats1.10452800


Ujifusa, A. (2015a, December 23). Education Department asks 13 states to address low test-participation rates. Education Week. Retrieved from http://blogs.edweek.org/edweek/campaign-k12/2015/12 twelve_states_asked_to_address.html?cmp=eml-enl-eu-news


Ujifusa, A. (2015b, July 9). Smarter balanced opt-out rates top 25 percent for Washington State 11thgraders. Education Week. Retrieved from https://blogs.edweek.org/edweek/state_edwatch/2015/07/smarter_balanced_opt-out_rates_top_25_percent_for_wash_st_high_school_juniors.html


U.S. Census Bureau, Small Area Income and Poverty Estimates. (2014, December). 2013 poverty estimates for school districts [Data file]. Retrieved from http://www.census.gov/did/www/saipe/data/schools/data/2013.html


Virtanen, M., & Raby, J. (2018, March 9). Social media, networking, key to W. Va. teacher walkout. U.S. News & World Report. Retrieved from https://www.usnews.com/news/best-states/west-virginia/articles/2018-03-09/social-media-networking-key-to-w-va-teacher-walkout


Walder, J. C., & Cleveland, A. D. (1971). The South’s new segregation academies. The Phi Delta Kappan53(4), 234–239.


Wang, Y. (2017). The social networks and paradoxes of the opt-out movement amid the Common Core State Standards implementation: The case of New York. Education Policy Analysis Archives, 25(34). Retrieved from http://dx.doi.org/10.14507/epaa.25.2757


Winerip, M. (2011, November 27). Principals protest increased use of test scores to evaluate educators. The New York Times. Retrieved from http://www.nytimes.com/2011/11/28/education/principals-protest-increased-use-of-test-scores-to-evaluate-educators.html


Zubrzycki, J. (2013, January 23). NEA supports Seattle teachers protesting standardized test. Education

Week. Retrieved from

http://blogs.edweek.org/edweek/District_Dossier/2013/01/nea_supports_seattle_teachers_.html




Cite This Article as: Teachers College Record Volume 122 Number 2, 2020, p. 1-24
https://www.tcrecord.org ID Number: 23062, Date Accessed: 11/29/2021 3:16:54 PM

Purchase Reprint Rights for this article or review