Influence over School Discipline Policy: Variation across Levels of Governance, School Contexts, and Time

Little research explores the relative influence of various stakeholders on school discipline policy. Using data from the SASS and ordered logistic regression, this study explores such influence while assessing variation across schools types and changes over time. Principals consistently rate themselves and teachers as the most influential stakeholders over setting school discipline policy. The proportion of racial minorities in a school predicts greater influence from higher levels of governance while charter schools report less. Increases in influence of principals and teachers over time are documented. The results may inform both policymakers and practitioners as they work to improve equitable disciplinary outcomes for students.


Introduction
School discipline has increasingly become an issue for policymakers and practitioners alike.The evidence suggests that exclusionary practices such as suspension and expulsion are predictive of a host of negative outcomes including decreased academic performance, failure to attain a high school degree, and a greater likelihood of being involved in the criminal justice system (Arcia 2006;Curran, 2016a;Costenbader & Markson 1998;Marchbanks et al. 2015;Mowen & Brent, 2016;Nicholson-Crotty, Birchmeier, & Valentine, 2009;Noltemeyer, Ward, & Mcloughlin, 2015;Peguero, & Bracy, 2015).Such outcomes, in turn, potentially have large economic costs to society (Marchbanks et al., 2015;Rumberger & Losen, 2016).Furthermore, the use of such practices disproportionately impacts minority students, with the rate of suspensions and expulsions being approximately three times larger for Black students than White students (Balfanz, Byrnes, & Fox, 2015;Curran, 2016a;Gregory, Skiba, & Noguera, 2010;Rafaelle-Mendez, 2003;Rocque, 2010;Rocque & Paternoster, 2011;Shollenberger, 2015;Skiba et al., 2002).Responding to this evidence, the U.S. Department of Education and Department of Justice released a "Dear Colleague" letter calling for improved equity in school discipline and a move away from exclusionary practices, and many states and school districts have implemented reforms to school discipline policy (Steinberg & Lacoe, 2017; U.S. Department of Education & Department of Justice, 2014).
Improvements in this area are complicated, however, by the varied sources of laws, regulations, and policies governing school discipline.School discipline is determined by federal law, state law, district policy, school policy, and the classroom rules and procedures established by individual teachers along with the political pressures of outside stakeholders such as parents (Curran, 2017;Kennedy-Lewis, 2014).Despite the large number of policy actors, we know little about the

Background Theoretical Framework
Education policy, including that related to school discipline, exists in a tiered system within which influence over policy is exerted by actors at various levels of governance from federal and state governments to school level personnel like principals and teachers (Barr, Dreeben, & Wiratchai, 1983;Berends, 2009;Datnow & Park, 2009;McLaughlin & Talbert, 1993).Additionally, school policy is influenced by societal trends and by external stakeholders such as parents and members of the broader community.According to Kingdon (2010), the broader public contributes to the definition of salient problems while also placing political constraints on possible solutions.Salient examples include the adoption of zero tolerance disciplinary policies in schools following expansion of punitive methods in the criminal justice system or in response to high-profile media events such as the shooting at Columbine High School (Casella, 2003;Richards, 2004;Skiba & Peterson, 2000).In many cases, these policy shifts may result from the need for "symbolic" action to appease public pressure after high profile events (Noguera, 1995).The formal levels of governance combined with such external pressures each contribute to the policy milieu that ultimately governs the development and application of school discipline.As a result, this study situates the examination of influence over discipline across a number of stakeholders.While all schools experience the influence of various entities over policy, the relative influence of each may be expected to vary across settings.For instance, many urban school districts experienced a push towards site-based management during the 1980s and 1990s, a phenomenon that potentially shifted influence from the district level to that of the school (Chubb & Moe, 1990;Cross & Reitzug, 1996).Indeed, research examining hiring practices has shown significantly larger gains in principal influence over hiring in urban settings as compared to non-urban settings over the last several decades (Engel, Cannata, & Curran, 2015).
Similarly, one might expect differences in levels of influence across sectors of schools.For instance, though not always the case in practice, charter schools are theoretically granted more autonomy across a number of domains and may have weaker ties to formal school districts (Finnigan, 2007;Gawlik, 2007;Mavrogordato, Goldring, & Smrekar, 2017;Wohlstetter & Chau, 2004).Though evidence points to increases in autonomy for non-charter school principals as well (Steinberg, 2014), influence over discipline policy in charter schools might be expected to be more localized than that of traditional public schools.Prior work is mixed on this question with some studies finding a small but significantly higher level of principal influence over discipline in charters as compared to traditional schools while other work finds no significant difference (Gawlik, 2008;Ni, Yan, & Pounder, 2017).
Finally, schools serving larger proportions of certain subgroups (minority students or lower income students) and schools with higher levels of student misconduct might be expected to be subject to differential pressures to develop policy within their walls or increased pressure from external policies targeted to their settings.Indeed, empirical evidence has shown that schools serving larger proportions of minority students are more likely to employ exclusionary and zero tolerance type discipline policy (Curran, 2017;Welch & Payne, 2010).

Sources of Influence over Discipline Policy
Like many aspects of formal schooling, school discipline policy is affected by laws, policies, and practices at various levels of governance.At the federal level, laws such as the Gun-Free Schools Act and the recently passed Every Student Succeeds Act contain provisions either requiring certain disciplinary responses for given offenses or opening the door to the use of school discipline measures as metrics for school accountability plans (U.S.Department of Education, n.d.).At the same time, the federal Department of Education issues guidance around a number of discipline issues such as exclusionary discipline and corporal punishment (U.S.Department of Justice & U.S. Department of Education, 2014).
At the state level, the extent of laws related to school discipline is wide, with laws pertaining to different disciplinary approaches, the use of school resources officers, professional development around discipline, and issues of monitoring and reporting among others (U.S.Department of Education, 2016).The U.S. Department of Education has recently compiled a compendium of state laws related to school discipline and safety, the entire document of which consists of over 4,000 pages (U.S.Department of Education, 2016), and many states have recently or are currently revising aspects of school discipline (Oosting, 2016, June 2;White, 2014).
School districts leverage influence over school discipline through the formal codification of district policy as well as through less formalized directives to school leaders and district personnel.These policies typically codify requirements from higher levels of governance, such as that required by the federal laws, such as the Gun-Free Schools Act, or by state statutes, such as bans on corporal punishment (Curran, 2016a;Curran & Kitchin, 2017).Additionally, district policy typically contains a number of provisions that are specific to the given district.For instance, many districts ban corporal punishment through district policy despite their state allowing for its use (Chason, 2014).As of 2015, nearly a quarter of the nation's largest school districts had recently implemented disciplinary reforms, many aimed at reducing the use of exclusionary discipline (Steinberg & Lacoe, 2017).
At the school level, the principal has the authority and responsibility of setting school-wide behavioral expectations and administering discipline for major infractions.With regard to disparities in discipline, prior work has documented that exclusionary or zero tolerance policies are more likely to be used in schools serving larger proportions of minority students while restorative practices are less likely to be used in these environments (Curran, 2017;Payne & Welch, 2010;2015).Coupled with other work that shows that much of the Black-White discipline gap is due to variation across schools rather than within schools (Anderson & Ritter, 2016), the role of the principal in setting such school-wide disciplinary policies and administering discipline becomes a highly relevant driver of both rates of discipline and racial disparities in discipline.
At the classroom level, teachers' exercise choice regarding when, how, and whom to discipline.In the language of policy implementation, teachers may be rightly considered "street level bureaucrats" and serve to implicitly create policy from the bottom up as a result of the choices they make regarding implementation of school disciplinary practices (Lipsky, 1980).These decisions are reflected in the decision making of teachers around classroom management plans and classroom rules, of which there may be considerable heterogeneity within a school (Thornberg, 2007).With regard to racial disparities in discipline, prior work has shown that teacher characteristics, such as a teacher's own race, can impact the likelihood that minority students experience exclusionary discipline (Lindsay & Hart, 2017).
Finally, parents represent a potentially influential set of stakeholders on school discipline policy.Though not formally in a position of policymaking within the structure of schools, parents may exert influence through engagement with formal organizations such as parent-teacher associations, through voice in school board elections and meetings, and through informal pressure via direct communication with district personnel.

Research on Influence Over School Discipline Policy
While the potential influence of each of these groups of stakeholders is clear, to date, very little research has systematically examined such influence over school discipline policy.Those that have done so tend to focus on a limited setting, a single year, or a single stakeholder.For example, in a recent study, researchers found principals' reports of influence over discipline policy were greater than that of other stakeholders such as teachers, local school boards, state boards of education, or parent associations; however, the study was limited to data from a single year (Ni, Yan, & Pounder, 2017).In another study, Bloom & Owens (2011) found that in 2004 principals reported a higher level of influence over disciplinary policy as compared to influence over staffing, curriculum, or funding issues but that the level of influence did not vary systematically across urban and non-urban schools (Bloom & Owens, 2011).Other work confirms the finding that principals perceive school discipline to be one of the areas of their highest level of influence (Ni, Yan, & Pounder, 2017;Ware & Kitsantas, 2011) and suggests that throughout the late 1980s and early 1990s, there were increases in the perception of principal influence over discipline policy (Shen, 1998).
In addition to studying principal perceptions of discipline, prior research has examined teachers' perceived autonomy over disciplining students.As in Bloom and Owens' (2011) study of principals, research has found that discipline is an area in which teachers perceive a high level of autonomy, more so than curricular decisions such as selecting textbooks or choosing content to be taught (Sparks, Malkus, & Ralph, 2015).Interestingly, there have been slight declines in the level of teacher autonomy over discipline between the years of 2003 to 2011, a trend that also holds for domains other than discipline (Sparks et al., 2015).
Other work suggests that, in the view of school leaders, teachers' influence over discipline, while high, is less than that of principals (Ni, Yan, & Pounder, 2017;Ware & Kitsantas, 2011) and can vary depending on whether one considers school-wide discipline policy or discipline policy within the classroom (Shen, 1998).For instance, examining nationally representative data from the early 1990s, Shen (1998) found that only about 35% of teachers reported a great deal of influence on school-wide discipline policy while nearly 70% reported a great deal of influence on disciplining students within their classrooms.
While the prior literature provides some insight into the relative level of influence of principals and teachers on discipline policy, little is known about the relative influence of other stakeholders such as state boards of education, local school boards, or parent groups.An exception to this is a recent paper using data from the 2007-08 Schools and Staffing Survey which estimated principals' perceptions of influence over discipline policy by district staff and local school boards to be less than that of principals and teachers but higher than that of state educational agencies or parent associations (Ni, Yan, & Pounder, 2017).This study, however, was limited to data from a single year and did not speak to how such trends have changed over time or varied across different contexts.
The current study builds on the prior work by addressing limitations around the stakeholders considered, the way influence varies across contexts, and the way influence has changed over time.In particular, this study addresses the relative influence of numerous stakeholders, over an extended period of time, and across a number of school contexts.In the next section, I turn to a presentation of the data and analytic approach.

Data
I draw on data from the Schools and Staffing Survey (SASS), a nationally representative survey conducted seven times between 1987 and 2011.Each wave of the SASS provides data on approximately 7,000 schools and includes surveys administered to both school leaders and teachers.For the purposes of this study, I limited the primary analyses to data from the 2007-2008 administration of the SASS (Tourkin et al., 2010).This is the most recent year in which data were collected on principal perceptions of a full range of stakeholders' influence on discipline policy.The analytic sample consisted of observations with available data on stakeholder influence over discipline policy as well as key covariates such as urbanicity, charter status, minority student enrollment, students eligible for free or reduced price lunch, and frequency of problem behaviors in the school.After list-wise deletion of observations missing data on any of these key covariates (approx.15%), the analytic sample consisted of 6,430 observations (rounded to the nearest 10 in compliance with IES restricted data regulations).Each observation represents a single principal in a single school.
In models exploring changes in influence over discipline policy over time, I used an expanded analytic dataset that contains data from multiple iterations of the SASS.In particular, I focused on data from years 1990 to 2007 (the 1987-88 and 2011-12 iterations included the influence question for only a subset of stakeholders).With the exception of models assessing the influence of district staff (which was not inquired about in the 1990 iteration of the survey), these models use an analytic sample of 36,670 observations.

Dependent Variables
As a part of the survey, principals responded to a question in which they rated the influence of a variety of stakeholders on setting discipline policy in their school.In particular, they rated the degree of influence that the state board of education, the local school board, school district staff, principals, teachers, and parent associations have over setting discipline policy.Though the question has been consistently administered across waves of the SASS, the number of response categories has varied across years.The earliest iterations included a Likert scale ranging from 1 to 6, with 1 representing no influence and 6 representing a great deal of influence.The scale was modified to 5 categories for the 1999 administration and to 4 categories (ranging from no influence to major influence) for years 2003 forward.For consistency, in models using multiple years of data, I recoded the earlier years' scales to a 0 to 3 scale ranging from no influence to major influence which aligns with the four item scale used in later iterations of the SASS.More details on this variable and the other study covariates can be found in Appendix Table A1.

Independent Variables
I focus on five key independent variables in this study.These variables include measures of urbanicity, charter school status, minority student enrollment, free and reduced price lunch enrollment, and frequency of problem behaviors.I describe each in detail here.
Urbanicity.I use a binary indicator (0/1) for whether a school is located in an urban setting.Urbanicity was measured using different numbers of categories across iterations of the SASS, reflecting changes in the Census definition of urbanicity across this period.The binary indicator used in this study, however, reflects the distinction between urban environments defined as "large or midsize central cities" and non-urban environments defined as "urban fringe of large or mid-size city" or "small town/rural".As shown in Table 1, approximately a quarter of schools are categorized as urban across each year of the data.
Charter status.Charter status is operationalized as a binary indicator of whether the public school is a charter school or not.According to the SASS, a charter school "is a public school that, in accordance with an enabling state statute, has been granted a charter exempting it from selected state or local rules and regulations" (U.S.Department of Education, 2007).They may include newly created charters as well as schools converted to charters.Given their emergence and expansion over the time period of data examined, charter schools appear only in the 2003 and 2007 waves of the SASS and, even in 2007-08, represent less than 3% of schools (see Table 1).Consequently, the charter school indicator is not included in the longitudinal analyses.

Minority student enrollment. I use a school level report of minority student enrollment.
As part of the schools portion of the SASS survey, principals (or their designee) completed survey questions that assessed the total enrollment of the school as well as enrollment of students by racial/ethnic category (Hispanic/Latino, White, Black, Asian or Pacific Islander, or American Indian or Alaska Native).I use a measure created by SASS that represents the proportion of students in the school that are non-White.As expected given demographic trends nationally, the percentage of minority students increased significantly from around a quarter of students in 1990 to approximately 41% of students in the 2007-08 school year (see Table 1).Free and reduced price lunch enrollment.I use a school level report of the proportion of students approved for free or reduced price lunch in the school.This measure comes from the school survey and is calculated from the number of students approved and the school's overall enrollment.As shown in Table 1, the percentage of students approved for FRPL increased from around 34% of students in 1990 to approximately 46% of students in 2007.
Frequency of problem behaviors.I operationalize frequency of problem behaviors as a composite of a series of items in which principals report the frequency of thirteen different problem behaviors.In particular, principals report the frequency with which physical conflict, robbery/theft, vandalism, alcohol use, drug abuse, weapons, physical abuse of teachers, student racial tensions, bullying, verbal abuse of teachers, disorder in the classroom, disrespect for teachers, and gangs are problems in the school.Principal responses range from 1 (never happens) to 5 (happens daily).I treat the average of these 13 items as a continuous measure of the frequency of problem behaviors in general in the school.As a consistent set of indicators of frequency of problem behaviors were not available before the 2003-04 iteration of the SASS, this variable is not included in the longitudinal analyses.

Methods
A series of descriptive analyses are used to document relative influence of various stakeholders over discipline policy, the degree to which such influence varies by characteristics of schools, and how such influence has changed over time.To explore the degree to which principals' perceptions of various stakeholders' influence over discipline policies varies, I first show descriptive statistics (means and standard errors) for principals' reported influence of each stakeholder on school discipline policy.The differences in the reported means for each stakeholder group reflect differences in perceived relative influence.
Next, to address the degree to which such influence varies by characteristics of the school or student population served, I reran the descriptive statistics conditional on a series of school characteristics.In particular, I ran conditional means based on urbanicity, charter status, and quartiles of proportion of students approved for free or reduced price lunch, quartiles of proportion of students that are racial minorities, and the quartiles of the frequency of problem behaviors.Comparing means of principal reported influence across each of these groups provides descriptive evidence on differences in perceived influence of each stakeholder across the different characteristics of schools.T-tests were used to test the statistical significance of the differences.
While conditional means provide insight into the raw gaps in influence over discipline that exist across subsets of schools, they do not provide insight into the degree to which such differences persist when controlling for other aspects of the school.For instance, differences in influence over discipline policy across schools serving higher or lower proportions of minority students may be partially explainable by differences in the poverty level of the school, differences in urbanicity of the school, or in the frequency of problem behaviors in the school.To explore the degree to which differences in influence over discipline policy across subgroups of schools persist when accounting for other characteristics of the school, I estimated ordered logistic regression models of the following form: (1) Influencesds = β0 + β1Urbands + β2Charterds + β3Prop_Minorityds + β4Prop_FRPLds + β5Freq_Behds +eds Where Influencesds represents the principal rated influence of a given stakeholder for school s in district d, Urban represents a binary indicator for being in an urban environment, Charter represents a binary indicator for being a charter school, Prop_Minority represents the proportion of students in the school that are racial/ethnic minorities, Prop_FRPL represents the proportion of students in the school approved for free or reduced price lunch, and Freq_Beh represents the average rating of the frequency of problem behaviors, and e is an error term.The regressions were clustered at the district level.The coefficients of interest are β1 through β5 which can be interpreted as the relationship between each of these variables of interest and principals' rating of influence of the stakeholder over setting discipline policy controlling for the other characteristics of the school.
Following these models, I sought to understand the degree to which influence over discipline policy has changed over the last several decades.To do so, I used an expanded set of data that included four additional waves of the SASS (years 1990 through 2007).I then ran a series of ordered logistic regression models predicting influence of each stakeholder over setting discipline policy from a continuous year variable as follows.
(2) Influencesds = β0 + β1Yeards +eds For school characteristics that were consistently available across waves of the SASS, I also ran versions of this model interacting the year measure with school characteristics (urbanicity, proportion of racial minority students, and proportion of students approved for free or reduced price lunch).These models take the following form: (3) Influencesds = β0 + β1Yeards + β2SchoolCharacteristicds + β3Year*SchoolCharacteristicds + eds Where the key coefficient of interest is β3 which reflects the differential time trend for different subgroups of schools.Models 2 and 3 were clustered at the district-year level.Such models allow for an examination of the degree to which changes over time vary for certain subgroups of schools.

Results
The findings of the analysis suggest significant differences across stakeholders in perceived influence over discipline policy.While differences across school characteristics were less pronounced, some significant differences by school type, racial composition, socio-economic composition, and frequency of problem behaviors were apparent.Finally, sizable changes in the level of influence over time for several stakeholder groups were apparent.In this section, I highlight each of these findings in detail.

Relative Influence over Setting Discipline Policy
First, I find that, in the view of principals, influence over setting discipline policy varies greatly across stakeholders.Figure 1 displays the average principal reported influence over discipline policy for state boards of education, local school boards, district staff, principals, teachers, and parent associations for the 2007-2008 school year.As shown, principals view themselves as holding the greatest influence over setting discipline policy, with teachers a near second.In terms of the descriptors available in the survey item, both principals and teachers near having a "major influence" (value of 3 on 0-3 scale) over setting discipline policy.This level of influence contrasts greatly with that of state boards and parent associations, each of which sits closer to "minor influence" (value of 1 on 0-3 scale).Local school boards and district staff straddle the middle with a perceived influence near "moderate influence" (value of 2 on 0-3 scale).As shown in Table 1, principals have consistently rated their and teachers' influence over school discipline policy above other stakeholders across all years of the data.

Variation by School Characteristics
Next, I find that the level of perceived influence over setting discipline policy varies across some school types.Table 2 presents the average level of perceived influence of each stakeholder group across urbanicity, charter status, quartiles of the proportion of free or reduced lunch approved students, quartiles of the proportion of racial minority students, and quartiles of the frequency of problem behaviors.The most pronounced differences in perceived influence are apparent when comparing charter and non-charter schools.In Table 2, statistically significant differences in   perceived influence for charter as compared to non-charter schools are shown with the superscript "b" beside the stakeholder name.While principals in charter schools do not rate their influence or that of teachers at a higher level than that of non-charters, charter school principals do rate the influence of higher levels of governance (state boards of education, local school boards, and district staff) as significantly lower than similar ratings by principals in non-charter schools.
In addition to differences across charter status, principals in schools serving higher proportions of traditionally disadvantaged groups consistently reported higher levels of influence over discipline from state board of educations and parent associations.As shown in Table 2, schools with higher proportions of students approved for free or reduced lunch or higher proportions of racial minorities reported higher levels of influence by states and parent associations on setting school discipline policy.In particular, statistically significant differences between schools in the 4 th quartile as compared to the 1 st quartile for each of these groups are shown by superscripts "c" (FRPL) and "d" (racial minority composition) by the stakeholder name.
Results from the ordered logistic regression models (Table 3) allow for an exploration of the relationships after controlling for the other relevant predictors.As shown, the relationship between charter status and level of influence remains consistent with the results of the conditional means.In particular, charter school principals report less influence of state boards of education, local school boards, and district staff over discipline policy.
With regard to the relationships observed for schools with different proportions of students approved for FRPL, the results suggest that the relationships seen in the conditional means may be driven in large part by differences across schools serving different proportions of racial minorities.In particular, after controlling for the other covariates, the proportion of students approved for FRPL is no longer a significant predictor of higher levels of influence from state boards of education or parent associations.In contrast, the proportion of minority student enrollment continues to predict higher levels of state boards of education, district staff, and parent associations influence over discipline policy.

Changes over Time
In addition to documenting the relative influence of each stakeholder in 2007-08, the results also suggest changes in relative influence over time.Descriptively, principals' perceptions of influence over discipline policy have increased across a number of stakeholders between the period of 1990 and 2007.(see Table 1 and Figure 2).In particular, the perceived influence of principals and teachers have increased by around 0.4 points on the 0-3 scale between 1990 and 2007 while the influence of states and district staff have increased by about 0.2 points (see Table 1).The change for local school boards and parent associations is practically insignificant.These same trends are shown in Table 4 which presents results from models predicting influence of each stakeholder from a continuous measure of year.As shown in the first row, principals' perceptions of influence over discipline policy have increased significantly across almost all stakeholders between the period of 1990 and 2007 with the largest increases being seen for principal and teacher influence.Table 4 also  shows results of models interacting the continuous year measure with several school characteristics that were consistently available across waves of data.I find that the perceived influence reported for principals and teachers in urban schools has increased over time while the relative influence of state boards of education and local school boards has shown less of an increase over time in urban ascompared to non-urban settings.A similar trend is apparent in schools serving higher proportions of racial minority students or students approved for FRPL though the relative size of such differential trends is quite small.In particular, given that the models use the proportion rather than percentage of racial minority students or those approved for the coefficients on the interaction terms represent the differential relationship for schools with 0% as compared to 100% minority/FRPL enrollment.In terms of differences per percentage point increase in either of these groups, the differential trend is practically negligible.

Discussion
The findings of this study provide a descriptive overview of the perceived influence of various stakeholders over setting school discipline policy across a number of contexts and across time.The findings both extend prior research while also pointing to important considerations for policy and practice.In this section, I put the findings in context of the extant literature and discuss the implications of the work.
The finding that the relative level of influence over discipline policy varies significantly across stakeholders and levels of governance extends prior research that had only considered influence of a select few stakeholders or at select time points (Bloom & Owens, 2011;Ni, Yan, & Pounder, 2017;Shen, 1998;Ware & Kitsantas, 2011).The findings of this study suggest that perceived influence generally increases as one moves to more local levels of formal governance.In particular, influence is highest for local actors such as principals and teachers and lowest for state boards of education while local school boards and district staff fall in the middle.Parent associations, while local, are an exception to this trend, having lower perceived influence perhaps due to their being outside the formal bureaucracy of schooling.Such trends in the perceived level of influence are consistent with prior findings using these SASS measures (Ni, Yan, & Pounder, 2017).
The finding that principals and teachers have the highest level of perceived influence aligns with the conception of these individuals as street level bureaucrats welding great influence over disciplinary decisions.The policy implementation literature has documented the role of such individuals who work on the frontline interfacing with clients as a critical influencer of policy affecting the end client (Lipsky, 1980).In the context of school discipline, this suggests that discipline policy reform must consider the decision-making and policy setting roles of principals and teachers.Indeed, prior work has demonstrated that schools, principals, and teachers often implement discipline differently, despite working under common state or district policy environments (Curran, 2017;Feuerborn & Tyre, 2016;Payne & Welch, 2010, 2015).This finding suggests that professional development activities and pre-service training focused on the role of discipline in education may hold some promise for empowering principals and teachers to best use their influence to positively impact discipline policy.Indeed, prior work has demonstrated the promise of teacher professional development to reduce the use of exclusionary discipline (Gregory, Allen, Mikami, Hafen, & Pianta, 2015).
Importantly, the way in which school personnel, particularly principals and teachers, make decisions around discipline has the potential to alleviate racial disparities in exclusionary discipline rates.Given that much of the Black-White discipline gap is explained by differences across schools rather than within schools (Anderson & Ritter, 2017), school level staff (i.e.principals and teachers) wield important leverage for addressing such disparities.In particular, principals might reconsider the use of zero tolerance and other severe or exclusionary practices that have been shown to be more common in schools serving higher percentages of racial minority students (Payne & Welch, 2010).At the same time, principals may also look to adopt more positive behavioral solutions such as positive behavior supports and interventions or restorative justice practices as these practices have been shown to be less commonly used in schools serving student bodies consisting of greater percentages of minority students (Payne & Welch, 2015).Positive behavior interventions and restorative practices have shown promise at reducing exclusionary discipline (Baker, 2008;Bradshaw, Mitchell, & Leaf, 2010;Horner et al., 2009;Gonzalez, 2015;Morrison, 2007).Though some studies have questioned whether such approaches reduce racial disparities in discipline within schools (Vincent, Sprague, Pavel, Tobin & Gau, 2015), the increased use of these practices in schools serving greater proportions of minority students, where such practices are currently less likely to be used, does hold potential to reduce disparities (Kline, 2016;Payne & Welch, 2015).Along with such policy shifts, principals may also look for ways to train their staff to actively reduce the use of exclusionary discipline and racial disparities therein through continued professional development.Prior work has shown that professional development which focused on teachers' emotional support, instructional capacity, and classroom practices may improve behavioral outcomes and reduce disparities in discipline (Gregory, Allen, Mikami, Hafen, & Pianta, 2016).
In addition to school personnel, parents, particularly organized through parent associations and other collectives, may hold some potential for addressing racially inequitable discipline.The findings suggest that parent associations have significantly more influence over discipline policy in schools with higher proportions of racial/ethnic minority enrollment than in schools with lower percentages of minority enrollment.While their influence is significantly lower than that of all of the other stakeholders, parent associations may represent one mechanism for creating positive external pressure on schools to reform discipline.Since parent associations appear to have more influence in schools serving higher proportions of minority students, they may be able to advocate for reductions in the presence of exclusionary practices and the institution of more restorative approaches in schools serving greater percentages of these student groups.Indeed, recent work outlining a framework for reducing disparities in school discipline include family connections and input on policy as one component for successful disciplinary reform, a suggestion that is supported by research on the involvement of family in resolving disciplinary infractions (Gregory, Skiba, & Mediratta, 2017;Sheridan et al., 2012).
Next, the finding that the perceived influence of principals and teachers has increased over the last several decades is perhaps surprising when considering changes in discipline policies from the federal and state governments.The 1990s saw a rapid increase in the application of zero tolerance discipline policies (Heaviside, Rowand, Williams, & Farris, 1998;Skiba & Knesting, 2001;Skiba et al., 2006).These policies were driven in part by legislation at the federal and state level that ostensibly removed some autonomy from local actors by dictating mandatory responses to certain disciplinary infractions (Curran, 2016a).Anecdotally, accounts of the implementation of such policies include numerous accounts of principals and teachers describing their hands as tied when it comes to disciplinary decisions (i.e.Fox, 2014, March 21).
Such trends in federal and state legislation as well as anecdotal evidence suggesting reduced autonomy of local actors would have suggested a decrease in the influence of local actors over discipline.That the results point to the opposite suggests that such anecdotal cases may not represent the general feeling of local actors as a whole or that principals and teachers have increased influence on other margins of discipline despite losing autonomy under zero tolerance policies.These findings extend that of Shen (1998) who had documented increases in the perception of principal influence over discipline policy in the late 1980s and early 1990s.The results of this study suggest that this trend has continued at least into the early 2000s, though perhaps being somewhat stagnant between 2003 and 2007.It is possible that the continuance of this trend through the 1990s may reflect general pushes during that time period toward site-based management, a trend that could have countered increased federal and state legislation on discipline (Cross & Reitzug, 1996).
Finally, relationships between school characteristics and perceived influence over discipline policy point to important considerations for ensuring equity in school discipline.The finding that principals of charter schools report significantly less influence from higher levels of governance (states, school boards, district staff) is expected given the unique organizational structure of charter schools.In particular, charters are by design granted greater autonomy than traditional public schools (Finnigan, 2007;Gawlik, 2007;Wohlstetter & Chau, 2004).Interestingly, however, the perceived lower influence of higher levels of governance among charter school principals does not translate to perceived higher influence of themselves as principals in both charters and non-charters rate themselves similarly on influence.
Whether such autonomy on the part of charter schools with regards to discipline policy is a positive or negative is more of an open question.On the one hand, many prominent charter schools have achieved high levels of student academic success while using strict "no excuses" disciplinary approaches (Angrist, Pathak, & Walters, 2013).On the other hand, some charters do have excessively high levels of exclusionary discipline, and others are criticized for utilizing suspension and expulsion as mechanisms to remove certain students from the charter environment (Joseph, 2016;Losen et al., 2016;Zimmer & Guarino, 2013).While the evidence suggests that, on average, charter schools do not use exclusionary discipline at rates much different than their traditional counterparts (Losen, Keith, Hodson, & Martinez, 2016;Malkus, 2016), such concerns nevertheless point to the possibility for problematic discipline in contexts that lack as much influence from higher levels of governance.In short, it is unclear whether the lower level of influence by higher levels of governance over school discipline policy in charter schools is a positive or negative.
While less influence by higher levels of governance is observed in charter schools, the opposite trend is apparent in schools serving traditionally disadvantaged students.This relationship is particularly pronounced for schools serving higher proportions of racial/ethnic minority students.In fact, in the fully specified models, the proportion of minority students in a school remains a significant predictor of increased influence of states and district staff on school discipline policy while similar relationships for proportion of students approved for FRPL lose statistical significance.
That higher levels of governance are, in the view of principals, wielding greater influence on discipline policy in schools serving higher proportions of minority students raises several possibilities.It is well documented that Black and Hispanic students are disproportionately subjected to exclusionary discipline such as suspensions or expulsions (Balfanz et al., 2015;Curran, 2016aCurran, , 2016b;;Gregory et al., 2010;Skiba et al., 2002) and that schools serving larger proportions of minorities are more likely to employ punitive discipline policies and practices while being less likely to use positive, restorative approaches (Blumer, 1958;Curran, 2017;Payne & Welch, 2010, 2015;Welch & Payne, 2010).It is possible, then, that states and district staff are imposing greater disciplinary control over these environments.Such a scenario would be consistent with the racial threat hypothesis or the idea that institutions put in place social controls to maintain power in the presence of greater proportions of racial minorities (Crawford, Chiricos, & Kleck, 1998;Payne & Welch, 2010;Welch, 2017;Welch & Payne, 2010).Alternatively, it could be that principals in these environments feel more bound to policies from higher levels of governance.If principals in schools serving higher proportions of minorities feel more bound by the regulations of higher levels of governance, they may feel less capable of exercising discretion in the case of potential violations of zero tolerance policies or in setting school discipline policy.Such feelings, then, could potentially explain part of the higher rates of use of these policies in schools serving minorities, even when such schools fall under the same state or district laws and policies.
On the other hand, it is also possible that principals serving in schools with higher proportions of minority students feel greater pressure from higher levels of governance with regard to decreasing disproportionate rates of discipline.Given increasing media and public attention to disparate rates of discipline by race, states and districts may be leveraging their influence more in schools serving higher minority proportions as a means of trying to address disparities in disciplinary infractions.
Whatever the explanation, it is nevertheless interesting that student racial composition appears to be the driving force behind differences in perceived influence, rather than the frequency of problem behaviors.With the exception of teachers, who principals perceive to have less influence in schools with higher levels of problem behaviors, the frequency of problem behaviors is largely unrelated to relative influence of various stakeholders.This suggests that differences in influence over discipline policy may not be a large driver of differences in misbehavior in schools.At the same time, however, it also suggests that states, school boards, and districts may not be crafting discipline policy designed to address the unique behavioral situations at particular schools.In applying influence evenly across schools with and without high levels of misbehavior, stakeholders may be overlooking opportunities to tailor policy that aligns with the needs of schools facing high levels of misbehavior.Instead, by focusing influence on schools based on racial composition, stakeholders may be contributing to disparate rates of punitive discipline.

Limitations
As with any study, the analysis described here includes certain limitations.First, it should be highlighted that the levels of influence explored reflect the perceptions of principals.It is possible that principals overestimate their own influence or misestimate the influence of others entities.Consequently, results of this study should be interpreted as reflecting principals' perceptions of influence rather than objective metrics of influence.Future work could replicate the findings of this study using data in which influence is measured by reports from each stakeholder.
Next, for results analyzing trends over time, results should be tempered by the changing nature of the survey questions included in SASS.As previously mentioned, the number of response options on questions asking about influence over discipline policy changed across waves of the SASS.Consequently, it is possible that some of the time trends documented may be attributable to changes in survey item design.
Finally, as the primary results of this study rely on data from the 2007-08 school year, the data used in the analyses is approaching ten years old.While it would be ideal to have more recent data, the 2007-08 iteration of the Schools and Staffing Survey was the last to include questions regarding the perceived influence of all stakeholders over school discipline policy.Nevertheless, there are reasons to believe that the results may extend to years beyond 2007-2008.For example, while the 2011-12 iteration of the SASS only asked about the influence of the principal on discipline policy, the average principal reported influence over discipline policy in 2011-12 (2.870) was almost exactly the same as that for 2007-08 (2.872).Additionally, based on estimates from the Office for Civil Rights, national rates of suspension of Black, White, and Latino students have been fairly stable from 2003-04 through the 2011-12 data, and similar disparities are seen in the latest data from 2013-14 (U.S.Department of Education, 2016;Losen, Hodson, Keith, Morrison, & Belway, 2015).While there is evidence that overall rates of suspension have dropped somewhat since then and that many states and school districts have begun implementing discipline reform, many of the issues surrounding school discipline remain similar to the conversations at the time of the 2007-08 data collection (Steinberg & Lacoe, 2017).Consequently, these limitations notwithstanding, the results of this study nevertheless further our understanding of stakeholder influence over discipline policy in schools.

Conclusion
School discipline policy promises to remain a prominent area for discussion and action in the field of education policy.As evidence around and pressure to change disparate rates of discipline by race/ethnicity continue to mount, the impetus to revisit discipline policy remains high.As outlined in this paper, however, such changes must navigate a multi-tiered system of governance as well as pressures external to the formal system of education.This paper provides initial evidence on the relative influence of a number of the stakeholders whose actions shape discipline policy.
While descriptive in nature, the findings point to the importance of engaging with particular stakeholders who appear to weld great influence on discipline, such as principals and teachers.At the same time, the findings suggest that discipline reforms should take into account differing characteristics of schools.In particular, the racial composition of schools as well as whether the school is a charter school appear to represent particularly salient characteristics for the way in which stakeholders weld influence over discipline policy.That the relative influence of different stakeholders has changed over time, also suggests that attention should be given to the way in which the different stakeholders' influence changes temporally.That said, such changes do not appear to be a zero sum game, as principal and teacher influence have increased despite parallel increases in state laws governing school discipline.
The findings of this paper extend the academic literature on the locus of control around school discipline policy.By looking at multiple stakeholders, across a number of contexts, and over time, this study expands our understanding of the landscape of discipline policy.In doing so, it lays a foundation for future work that can further unpack the influence and potential policy levers of various entities over school discipline policy.

Note.
Standard errors in parentheses.All estimates weighted by administrator weight.Superscript letters represent statistically significant difference (p<0.05) for the following comparisons: a = urban and non-urban; b = charter and non-charter; c = FRPL qrt. 1 and 4; d = minority percentage qrt. 1 and 4; e = problem behaviors qrt. 1 and 4.

Figure 1 .
Figure 1.Relative influence of various stakeholders over school discipline policy as perceived by principals for year 2007 (n = 6,430)

Figure 2 :
Figure 2: Principal reported perceived influence on discipline policy of various stakeholders over time

Table 1
Means and Standard Errors for Descriptive Statistics on Key Independent and Dependent Variables by Year Note.Standard errors in parentheses.All estimates weighted by administrator weight.

Table 2
Principal Rated Influence Over Setting School Discipline Policy for the Full Sample and by Select Subgroups forYear 2007Year  -2008

Table 3
Ordered Logistic Regression Coefficients, Standard Errors, and Odds Ratios from Models Predicting Principal Perceptions of Influence Over Setting Discipline Policy from Independent Variables of Interest Note.Standard errors in parentheses are clustered by district.All estimates are weighted by the administrator weight.

Table 4
Coefficients and Standard Errors from Ordered Logistic Regression Models Predicting Principal Perceptions of Influence over School Discipline Policy from a Continuous Year Measure with Interactions by Urbanicity, Racial Minority Percentage, and Free and Reduced Price Lunch Percentage