ANNOTATED BIBLIOGRAPHY

article11.pdf

Home >English homework help >Article writing homework help >ANNOTATED BIBLIOGRAPHY

Vulnerable Decision Points for Disproportionate Office Discipline Referrals: Comparisons of Discipline for African American and White

Elementary School Students

Keith Smolkowski Oregon Research Institute

Erik J. Girvan, Kent McIntosh, Rhonda N. T. Nese, and Robert H. Horner University of Oregon

ABSTRACT: Racial disparities in rates of exclusionary school discipline are well documented and seemingly intractable. However, emerging theories on implicit bias show promise in identifying effective interventions. In this study, we used school discipline data from 1,666 elementary schools and 483,686 office discipline referrals to identify specific situations in which disproportionality was more likely. Results were largely consistent with our theoretical model, indicating increased racial and gender disproportionality for subjectively defined behaviors, in classrooms, and for incidents classified as more severe. The time of day also substantially affected disproportionality. These findings can be used to pinpoint specific student–teacher interactions for intervention.

▪ In the United States, racial disparities in rates of exclusionary discipline for students of color have been well documented, with differ- ences most pronounced for African American students in particular (Losen & Gillespie, 2012; Losen, Hodson, Keith, Morrison, & Belway, 2015). For example, in the 2011–2012 school year, nationally, administrators used out-of- school suspensions to discipline 8% of African American elementary students and 23% of Afri- can American secondary students, compared to 2% of White elementary students and 7% of White secondary students (Losen et al., 2015). In addition, converging research provides evi- dence that disproportionality cannot be wholly attributed to structural factors associated with students or schools. Even controlling for poverty, participation in gifted-and-talented programs, student–teacher ratio, attendance rates, and oth- er factors, African American students continue to be disciplined at higher rates than White stu- dents (Anyon et al., 2014; Fabelo et al., 2011; Skiba, Poloni-Staudinger, Simmons, Feggins, & Chung, 2005; Wallace, Goodkind, Wallace, & Bachman, 2008). As a result, although disparate treatment of any students by race is concerning, the disparities for African American students are most severe. Moreover, there is no evidence that disproportionality results from differences in levels of student behavior by race. To the

contrary, research has shown that teachers are more likely to issue office discipline referrals (ODRs) to African American students even after controlling for their own ratings of the students’ behaviors (Bradshaw, Mitchell, O’Brennan, & Leaf, 2010).

This research base provides a clear descrip- tion of the extent of disproportionality in school discipline. By comparison, there is much less empirical research on interventions to reduce disproportionality, or even what variables should be targeted for intervention (Martinez, 2013; Staats, 2014). Thus, to lay a crucial foundation for addressing disproportionality, it is necessary to focus on developing and validating a theoreti- cal framework that explains when and why disproportionality is most likely to occur and, more importantly, identifies malleable variables that can be used to reduce it.

A Model for Explaining Disproportionality in School Discipline

To address this need, McIntosh, Girvan, Horner, and Smolkowski (2014) proposed the Vulnerable Decision Points model. This model draws on psychological research to describe the conditions under which racial bias is most likely to influence decisions in the school

178 / August 2016 Behavioral Disorders, 41 (4), 178–195

discipline context and highlights specific ave- nues for intervention. As such, it focuses on teacher and administrator perceptions and judgments within specific discipline decisions. Although explaining the entire model is beyond the scope of this article, two critical aspects of it are described here: (a) explicit versus implicit bias and (b) vulnerable decision points in school discipline.

Explicit Versus Implicit Bias

A large body of psychological research suggests that there are two distinct types of bias, explicit and implicit, which operate differently and can influence different types of decisions (Girvan, 2015; Girvan, Deason, & Borgida, 2015). This distinction is particularly important because interventions that have been shown to reduce the effects of explicit bias are not necessarily effective for implicit bias and vice versa (Lai, Hoffman, Nosek, & Greenwald, 2013; Pettigrew & Tropp, 2006). Explicit bias is what we typically think of as prejudice: ethnocentrism, racism, and other consciously endorsed attitudes or beliefs, such as the belief that African Americans are inher- ently criminal or lazy (Pearson, Dovidio, & Gaertner, 2009). By comparison, implicit bias is the automatic, often unconscious impact that stereotypic associations with racial and other groups can have on perceptions, judg- ments, decision-making, and behavior (Devine, 1989; Greenwald & Banaji, 1995; Pearson et al., 2009). Rather than conscious endorse- ment of beliefs or feelings, it has its roots in generalized associations formed from systemat- ically repetitious or unique and limited experi- ence or exposure. Thus, for example, regularly seeing images of African American but not White criminal offenders in the media or know- ing only one person who was the victim of a carjacking, an incident in which the perpetrator happened to be African American, may lead even people with egalitarian values to automat- ically assume that a racially unidentified gang member is African American, presume that an area in which there are many African Ameri- cans living must have a crime problem, or lock their car door when seeing an African American man (Greenwald & Banaji, 1995). From a behavioral standpoint, implicit bias can be conceptualized as inappropriate stimulus control over an individual’s responses to others’ behavior that is based on irrelevant features of the behavior, as opposed to an objective view

of the behavior. In school settings, implicit bias may be seen in staff decisions to send stu- dents of color to the office for relatively minor incidents of unwanted behavior.

An individual’s levels of explicit and implicit biases are relatively independent of one another (Greenwald, Poehlman, Uhlmann, & Banaji, 2009). As a result, several combina- tions are possible: Those without explicit or implicit bias (the “truly nonprejudiced”), those with both explicit and implicit bias (the “truly prejudiced”), and those whose explicit and implicit biases are not aligned (Fazio, Jackson, Dunton, & Williams, 1995). Of the combina- tions, given evidence of both a substantial decline in explicit racial bias and the relative pervasiveness of implicit racial bias (Bobo, Charles, Krysan, & Simmons, 2012; Greenwald & Pettigrew, 2014; Nosek, Greenwald, & Banaji, 2007), the majority of U.S. adults are likely to express no explicit racial bias (i.e., have and report a belief in the value of diversi- ty, equity, and inclusion in society) but have implicit racial biases favoring Whites over African Americans, a combination known as aversive racism (Dovidio & Gaertner, 2000; Pearson et al., 2009).

Under aversive racism theory, people are assumed to be highly motivated not to be, or appear not to be, racially biased. As such, when confronted with decisions in which incor- rect responses are clear or particular responses would be seen as discriminatory, they will most likely select correct or nondiscriminatory responses. But in discretionary decisions and those with an unclear “right answer,” decisions do not directly implicate or threaten decision- makers’ egalitarian values or self-concepts. In those circumstances, values notwithstanding, the majority of people may act in ways that are discriminatory (Pearson et al., 2009). Research across a range of contexts outside of education supports these predictions. For example, White people are generally willing to offer help at the same rates to African Americans as to Whites, except when there are difficult circumstances, such as when helping is risky or time consum- ing, or in the presence of other factors that can be used to justify the failure to help (Saucier, Miller, & Doucet, 2005); people tend to dis- criminate in hiring recommendations against moderately but not highly or poorly qualified candidates (Dovidio & Gaertner, 2000); and incriminating evidence that is found to be inad- missible tends to influence jury decisions about

Behavioral Disorders, 41 (4), 178–195 August 2016 / 179

African American but not White defendants (Johnson, Whitestone, Jackson, & Gatto, 1995).

Translating this work into the school disci- pline context, disproportionality may come from an individual educator’s explicit biases, but we posit it is more likely that it comes from implicit biases. School discipline data patterns can help identify which is more at play. If ex- plicit bias is prominent, school discipline data would likely represent a consistent pattern of disproportionality across many circumstances. For example, analysis of discipline data might demonstrate that African American students are sent out of class regularly for behavior incidents, regardless of the situation. From the literature, effective top-down policies, such as evaluating administrators and teachers based on levels of disproportionality, are more likely to mitigate the effects of explicit bias (Lerner & Tetlock, 1999; Pettigrew & Tropp, 2006).

In contrast, an indicator of implicit bias in school discipline data would reflect peaks and valleys in disproportionality from the same teach‐ ers across different situations, with relative equi- ty in some situations and high disproportionality in others, as predicted by psychological theory. The data might demonstrate, for example, that African American students receive dispropor- tionately more ODRs for defiance or disrespect than White students because identifying these behaviors involves a discretionary decision for teachers (e.g., whether student behavior is acceptable or unacceptable to the teacher). Or it might show that consequences for the same behavior are more severe for African American students during times of the day when teachers are tired. The model indicates that effects of implicit bias can be reduced by making disci- pline procedures for these types of behaviors as objective as possible, and by examining staff expectations and providing training in how to respond instructionally to unexpected student behavior in these specific situations, without resorting to an ODR.

Vulnerable Decision Points in School Discipline

We use the term vulnerable decision points (VDPs) to describe specific situations in which increased disproportionality tends to occur. VDPs are contextual events or ele- ments, such as those that increase the likeli- hood of implicit bias affecting discipline decision making, including a teacher’s deci- sion to issue an ODR or an administrator’s decision to suspend the student.

Emerging research indicates the presence of some VDPs in education. The VDP with the strongest research support is a situation in which the student behavior is inherently subjec- tive (i.e., when staff have to make a judgment call regarding whether the behavior is a viola- tion; Skiba, Michael, Nardo, & Peterson, 2002). For example, defiance and disrespect are more ambiguously defined and allow more staff discretion than more objectively defined behaviors (Greflund, McIntosh, Mercer, & May, 2014). In these circumstances, educators must decide whether a student’s behavior (e.g., a student sharing an opinion about an assignment) is disrespectful, whereas behaviors such as smoking or theft are far more easily determined. An analysis of school discipline outcomes for every ninth-grade student in Texas for three academic years showed that, after con- trolling for 83 student- and school-level factors, African American students had a 31% higher likelihood than White students of being disci- plined for discretionary violations but a 23% lower likelihood of being disciplined for man- datory violations (Fabelo et al., 2011).

Other VDPs have theoretical but no empir- ical support to date. A previously untested extension of the subjectivity of VDPs is the dis- cretionary judgment involved in classifying similar or borderline student behavior as severe (e.g., “fighting” warranting a major ODR) or less severe (e.g., “physical contact” warranting a minor ODR or warning slip). In keeping with our theory, educators may be more likely to overreact to minor behavior by African American students by classifying it as a more severe (major) incident. Another hypothesized but untested VDP is location. The nature of student–teacher interactions across different contexts may lead to different behaviors observed or varying risk for biased responding. For example, previous theories implicate a relation between discipline dispro- portionality and the academic achievement gap (Gregory, Skiba, & Noguera, 2010), which suggests that classrooms are themselves a potential VDP, especially during periods with a strong academic focus (e.g., literacy). There are several reasons why classrooms could be a VDP. For example, in classrooms, teachers provide more directions to complete tasks, which some students may perceive as control- ling, too difficult, or irrelevant to their lives. There may also be less engaging instructional techniques used, a mismatch of teacher and student goals (e.g., instructional time versus

180 / August 2016 Behavioral Disorders, 41 (4), 178–195

socialization), or a fear of “losing control” of a classroom when minor noncompliance occurs (Fenning & Rose, 2007; Okonofua, Walton, & Eberhardt, 2016). Evidence for other VDPs, such as time of day (Kouchaki & Smith, 2014; Linder et al., 2014) or decision-maker fatigue or hunger (Danziger, Levav, & Avnaim-Pesso, 2011; Gailliot, Peruche, Plant, & Baumeister, 2009; Kouchaki & Smith, 2014) align with the behavioral principle of a motivating (or estab- lishing) operation (Laraway, Snycerski, Michael, & Poling, 2003). These VDPs are not related to student behavior, but rather the internal state of the decision maker. For example, when a teacher is exhausted, sending a student from a particular group out of the classroom becomes particularly reinforcing. The student’s behavior is no different, but the internal state of the edu- cator may make unexpected behavior more likely to be categorized as defiance. These VDPs have been found in other fields (e.g., law, medicine) but have yet to be explored in education.

Finally, research shows significant gender differences and possible race–gender interac- tions in rates of student discipline (Fabelo et al., 2011; Losen et al., 2015). For example, disci- pline rates for African American females can be much higher than for White females and in some cases are more similar to those of African American males (Blake, Butler, Lewis, & Dar- ensbourg, 2011; Crenshaw, Ocen, & Nanda, 2015), suggesting potential interactions between students’ behaviors and the salience of either their race or gender to teachers (Sinclair & Kunda, 1999, 2000). Such interactions would indicate that student gender may influence a teacher’s decision to send a student to the office, making it a potential VDP as well.

Understanding the mechanisms by which implicit bias emerges and affects school disci- pline decision making will be important for developing effective interventions. If those conditions that are most likely to be influenced by implicit bias are also those most likely to produce disproportionality, research from oth- er fields would indicate the following interven- tion approaches as promising: (a) making the specific decision more objective (e.g., creating operational definitions of behavior violations that should and should not result in an ODR), (b) teaching individuals to recognize these VDPs (including personal motivating opera- tions), and (c) practicing and using alternative responses to behavior that are instructive and

nonexclusionary (McIntosh, Girvan, Horner, & Smolkowski, 2014).

The Present Study

The main purpose of this study was to iden- tify patterns in actual school discipline data that would support or disprove the VDP model. Our conceptual model predicts that, within the context of adult decisions about disciplinary actions, certain situations are more vulnerable to the impacts of implicit biases. This study test- ed two specific research hypotheses consistent with previous research and our VDP model:

1. Compared to White students, African Amer- ican students receive ODRs at a higher rate for subjectively defined behaviors than for objectively defined behaviors.

2. The relative odds of receiving a subjective versus objective ODR will be greater for African Americans when associated with four potential VDPs: (a) at the end of the day versus earlier in the day, (b) in class- room settings versus nonclassroom settings, (c) for classifying incidents as severe versus minor, and (d) for African American females as opposed to White females.

To the extent patterns in discipline data sug- gest that disproportionality tends to be concen- trated in particular situations, such as those discussed above, it indicates that the VDP approach may be an important strategy to improve equity in school discipline. If the model was supported, the results could then serve as a guide for understanding the larger problem of disproportionality as one that is, in fact, com- posed of smaller, more specific, and potentially more manageable situations that can be targeted for the development of effective interventions.

Method

Participants and Settings

The sample included 483,686 ODRs issued in the 2011–2012 academic year to 235,542 students by 53,030 educators in 1,666 elemen- tary schools that were using the School-Wide Information System (SWIS; May et al., 2013), an online computer application for tracking and analyzing ODRs. Schools came from 45 states across the U.S. The average enrollment was 493 (SD5 184), the average percent of stu- dents receiving free or reduced price meals was

Behavioral Disorders, 41 (4), 178–195 August 2016 / 181

55% (SD 5 24%), and the average percent of non-White students was 47% (SD 5 26%).

Consistent with our goals, this sample included a number of restrictions. We included only elementary schools to examine relations in a relatively consistent student–teacher con- text, as opposed to middle and high schools, which tend to operate differently (e.g., different teachers each period). We included only schools that coded race or ethnicity for at least 80% of ODRs and with at least 10 African American and 10 White students to avoid using estimates with schools that have little or no racial diversity. For the present analysis, we included only ODRs delivered to African American or White students to narrow our focus to the most common comparison for dis- proportionality (Skiba et al., 2011). Because the analysis focused on subjectively versus objec- tively defined ODRs, the sample excluded ODRs for behavior types that could not reliably be classified as one or the other by a panel of educational experts, as described below (Greflund et al., 2014). The sample included 424,840 subjective ODRs and 58,846 objec- tive ODRs.

Measures

Office Discipline Referrals

ODRs are standardized forms used to doc- ument incidents of problem behavior (Sugai, Sprague, Horner, & Walker, 2000). School per- sonnel issue ODRs to students for a defined set of behavior violations (e.g., fighting). When the process and specific behaviors are operational- ly defined (as is required for the use of SWIS), ODRs can be reliable and valid indicators of problem behavior (Irvin, Tobin, Sprague, Sugai, & Vincent, 2004; McIntosh, Campbell, Carter, & Zumbo, 2009; Walker, Cheney, Stage, & Blum, 2005). SWIS ODRs include a range of fields (e.g., location, time of day, student, staff) that can be used to identify specific situations with elevated problem behavior. In this study, these fields were used to identify and test whether and to what extent specific situations had increased disproportionality (i.e., theorized VDPs).

Subjectivity of ODRs

Each ODR behavior type was classified as either subjective (e.g., defiance, disrespect, disruption) or objective (e.g., fighting, theft, tru- ancy) by an expert panel, composed of four

researchers in school discipline, racial/ethnic disproportionality, and/or culturally responsive behavior support, which rated the specific SWIS behavior definitions used for ODRs (Greflund et al., 2014). ODRs for behaviors in which the expert panel did not agree on a clas- sification (e.g., dress code violation) or were not clearly attributable to student actions in ele- mentary school (e.g., attendance, which is related to caregiver behavior) were removed from analyses.

Time of Day

Based on the time of the incident, ODRs were coded as occurring in 15-min intervals throughout the school day. According to our hypotheses regarding fatigue, ODRs issued in approximately the last hour of the school day (2:00 to 3:00 p.m.) were compared to ODRs issued earlier in the day (8:30 a.m. to 1:45 p.m.).

Severity of ODRs

In SWIS, school personnel determine whether each behavior incident is major (i.e., requiring administrator action) or minor (i.e., expected to be handled in the classroom). Except for certain major behaviors (e.g., arson, bomb threat), ODRs can be classified by staff as either major (i.e., severe) or minor (i.e., less severe).

School-Level Variables

School characteristics included proportion of students receiving free and reduced-price lunch, the proportion of African American stu- dents, and the proportion of minority students other than African American. These data were collected from the National Center for Educa- tional Statistics (NCES) and were used as covar‐ iates to control for their influence on ODR patterns.

Procedure

ODR data from each school were extracted from the SWIS database. Each school in the study had signed a data-sharing agreement that allowed their data to be used for research purposes. Upon extraction, information entered for each ODR was used to identify student race, student gender, referring staff member, and hypothesized VDPs (e.g., subjectivity, time of day, location, severity).

182 / August 2016 Behavioral Disorders, 41 (4), 178–195

Analytic Plan

Consistent with the VDP model, we hypothesized that the teacher decision to issue ODRs was subject to implicit bias, and thus that biases increased the odds of subjective compared to objective ODRs for African Amer- ican students during specific VDPs. According- ly, the dependent variable is the relative odds of an ODR for a subjectively defined behavior versus an objectively defined behavior. For ease of reference, in describing the analysis and results, we refer to this outcome as the odds of a subjective ODR or just use the term “subjective ODR.”

To assess these hypotheses, we fit a series of multilevel logistic regression models with differ- ent predictors of the odds of a subjective ODR: (a) unconditional model with no predictors, (b) school-level covariates only, (c) African Ameri- can recipient, (d) African American and end of day, (e) African American and classroom, (f) African American and major ODR, and (g) Afri- can American and female. We then fit three final models that included each of the three hypothesized VDPs with both African American and female as predictors. Models with multiple predictors also included all relevant interac- tions. Because ODRs were collected from dif- ferent educators in different schools, we also included these two sources of random variation into all models (i.e., ODRs were nested within educators and schools).

For certain analyses, we further restricted the sample, so the sample size varied by model. First, missing data on covariates reduced the sample to 455,527 ODRs (94.2%) and 1,595 schools (95.7%). The model that tested major versus minor ODRs was restricted to schools that use both majors and minors, 368,692 sub- jective ODRs (76.2%). Models that tested time of day included only the 402,724 ODRs (83.3%) that occurred when classes are typical- ly in session, from 8:30 a.m. to 3:00 p.m.

Interpretation of coefficients. Logistic regressions are particularly useful because the results allow the calculation of odds ratios, a form of effect size, from the raw parameter esti- mates (Judge & Cable, 2004). The odds ratio is an estimate of the increase in odds per unit change of the predictor, so if the model pro- duced a raw coefficient for African Americans of 0.405, then the odds ratio 5 e0.405 5 1.5 and implies that African American students are 50% more likely to receive a subjective ODR as White students (holding all other predictors,

covariates, and random effects constant). Con- sider, for example, a school with 100 African American and 100 White students. An odds ratio of 1.5 would indicate that, if the 100White students received a total of 20 subjective ODRs, then the African American students would have received a total of 30, controlling for covariates. An odds ratio of 1 indicates that African Ameri- can and White students are equally likely to receive a subjective ODR. An odds ratio of 0.5 indicates that the outcome is half as likely for that group. As such, the odds ratio for African American students is our primary indicator of disproportionality in this set of analyses.

Because the large sample made even trivial differences statistically significant, we focused on whether odds ratios in the hypothesized VDPs represent substantial increases in risk of disproportionality in the expected direction. In the United States, each state is left to determine its own criterion for significant racial dis- proportionality in education (U.S. Government Accountability Office, 2013), leaving little guid- ance for those seeking a benchmark formeaning- ful differences. To determine this threshold, we used the “four-fifths” or “80% rule” used by the Department of Justice and Equal Employment Opportunity Commission (EEOC) to identify employment practices that result in “serious dis- crepancies” based on race or other protected classes (Equal Employment Opportunity Com- mission, Civil Service Commission, Department of Labor, & Department of Justice, 1978). The rule translates to an odds ratio of 1.25 or greater or, equivalently, 0.80 or less. Referring again to the hypothetical school with 100 White and 100 African American students, an odds ratio of 1.25 implies that if 20 White students were sent to the office for a subjective offense, such as inap- propriate language, 25 African American stu- dents would have been sent to the office, which would be problematic if the actual rate of behav‐ iors were similar for both White and African American students (e.g., Bradshaw et al., 2010). This threshold for odds ratios is conservative and is thus not intended to suggest that odds ratios of, say, 1.10 to 1.25 are unimportant. To the contrary, the EEOC, for example, considers evidence of racial discrepancies that do not satisfy the “four-fifths” rule to constitute an adverse impact if the discrepancies are based on large (e.g., nationwide) samples and other- wise practically significant. Consistent with this, we considered those odds ratios that reveal dis- proportionality within the [0.80, 1.25] interval to be worth examination and odds ratios equal

Behavioral Disorders, 41 (4), 178–195 August 2016 / 183

to or outside of the [0.80, 1.25] interval to identify a situation in which disproportionality may be especially problematic.

Results

Table 1 describes the complete set of mod- els and reports estimates, standard errors, and statistical significance indicators. This table provides coefficients in the log odds scale, which are not easily interpretable but fully describe our statistical models. We interpreted the results in terms of odds and odds ratios, which we present in Tables 2 through 6.

The analysis included five predictors asso- ciated with ODRs: African American (AA), female, end of day, classroom, and major ODR. The names denote the event (coded “1”) compared to the converse (coded “0”; i.e., White, male, earlier in the day, and minor ODR). The sample contained 38% AA, 24% female, 19% end of day, 57% in the classroom setting, and 33% major ODRs. The sample ranged from 368,692 to 483,686 depending on the measure and model. For analyses of major ODRs, for example, schools were excluded if they did not use minor ODRs, leaving 368,692 ODRs for the analysis. End of day versus earlier ODRs included only those given during the regular portion of the day, from 8:30 a.m. to 3:00 p.m. (N 5 402,724). The analyses includ- ed data from 1,100 to 1,595 schools, which should offer robust, generalizable results.

Subjective ODRs

The dependent variable for all analyses was the subjectivity of ODRs: subjective versus objective ODRs (subjective ODR). Approximate- ly 88% (424,840) of the ODRs were subjective (58,846 were objective). Thus, we calculated 7.2:1 odds of a subjective ODR across educators and schools: For every one objective ODR in the dataset, there were 7.2 subjective ODRs. Once we account for clustering within educators and schools, the odds of a subjective ODR for a spe- cific individual educator and school, on average, is 13.8:1 or simply 13.8.

Random (Clustering) Effects

The variances in Table 1 represent random effects from the multilevel logistic regression models. We interpreted the educator- and school-level variances in terms of the median odds ratio (MOR), which is “the median value

of the odds ratio between the [cluster] at high- est risk and the [cluster] at lowest risk when randomly picking out two [clusters]” (Merlo et al., 2006, p. 292). The MOR for educator was 2.81, 95% CI [2.75, 2.86], and the MOR for school was 2.75, 95% CI [2.63, 2.86]. These MORs suggest considerable variability across teachers and schools in the odds of sub- jective ODRs compared to objective ODRs.

Fixed Effects

The fixed effects describe the odds ratios for race, gender, and the three VDPs, control- ling for school-level variables and accounting for variability at the educator and school levels. The odds ratio for African American students was 1.20. With these controls, on average, Afri- can American students were 1.2 times more likely to receive a subjective ODR than White students from the same teacher in the same school. As an illustration, if a teacher issued 10 subjective and 10 objective ODRs to White students, we would expect that same teacher to have issued 12 subjective and 10 objective ODRs to African American students.

AA and VDPs: Two-Way Interactions

We estimated interactions between African American (AA) and end of day versus other times, classroom versus nonclassroom settings, major versus minor ODRs, and student gender. The results are presented in Tables 2 through 6 and shown in Figure 1.

End of day.We classified end-of-day ODRs as those delivered between 2:00 and 3:00 p.m. and compared them to ODRs delivered between 8:30 a.m. and 1:45 p.m. Table 2 parti- tions the effects from the fourth multilevel logis- tic regression model, which include predictors AA, end of day, and their interaction, in terms of odds ratios. The first four rows of the table give the odds (not odds ratios) of a subjective ODR for each subgroup. The odds represent the raw likelihood of a subjective referral, before comparing it to the odds of another situ- ation (i.e., odds ratios). As shown in Table 2, the odds increase for White students at the end of the day but remain consistently high for African American students.

The next set of four rows, Rows 5 through 8, provide odds ratios associated with AA: White students for either early in the day or the end of day, and vice versa. Earlier in the day, African American students were 1.25

184 / August 2016 Behavioral Disorders, 41 (4), 178–195

TA B LE

1 M ul til ev

el Lo

gi st ic

R eg

re ss io n M od

el s Pr ed

ic tin

g th e R ec

ei pt

of a Su

bj ec

tiv e R ef er ra l by

D em

og ra ph

ic an

d V ul ne

ra bl e D ec

is io n Po

in t (V D P)

In di ca

to rs

(1 )

U nc

on di tio

na l

(2 )

C ov

ar ia te s

O nl y

(3 )

A fr ic an

A m er ic an

(A A )

(4 )

A A &

En d

of D ay

(5 )

A A &

C la ss ro om

(6 )

A A &

M aj or

R ef er ra l

(7 )

A A &

Fe m al e

(8 )

A A ,F

em al e,

& En

d of

D ay

(9 )

A A ,F

em al e,

& C la ss ro om

(1 0)

A A ,F

em al e,

& M aj or

R ef er ra l

Fi xe d ef fe ct s

In te rc ep

t 2. 63

** * (.0

5) 2. 59

** * (.0

5) 2. 52

** * (.0

5) 2. 49

** * (.0

5) 2. 38

** * (.0

5) 2. 59

** * (.0

5) 2. 59

** * (.0

5) 2. 57

** * (.0

5) 2. 41

** * (.0

5) 2. 66

** * (.0

A A

0. 18

** * (.0

2) 0. 23

** * (.0

2) 0. 14

** * (.0

2) 0. 09

** * (.0

2) 0. 14

** * (.0

2) 0. 18

** * (.0

2) 0. 11

** * (.0

2) 0. 07

** (.0

Fe m al e

− 0. 30

** * (.0

2) − 0. 33

** * (.0

2) − 0. 16

** * (.0

2) − 0. 31

** * (.0

V D P

0. 19

** * (.0

2) 0. 29

** * (.0

2) − 0. 22

** * (.0

2) 0. 17

** * (.0

2) 0. 36

** * (.0

2) − 0. 23

** * (.0

A A 6

Fe m al e

0. 20

** * (.0

3) 0. 22

** * (.0

3) 0. 13

** (.0

4) 0. 14

** * (.0

A A 6

V D P

− 0. 20

** * (.0

3) 0. 10

** * (.0

2) 0. 20

** * (.0

3) − 0. 16

** * (.0

4) 0. 06

* (.0

3) 0. 16

** * (.0

Fe m al e 6

V D P

0. 08

(.0 4)

− 0. 28

** * (.0

3) 0. 01

(.0 4)

A A 6

Fe m al e 6

V D P

− 0. 15

* (.0

7) 0. 13

* (.0

5) 0. 19

** (.0

FR L %

0. 09

** * (.0

1) 0. 09

** * (.0

1) 0. 10

** * (.0

1) 0. 09

** * (.0

1) 0. 07

** * (.0

2) 0. 09

** * (.0

1) 0. 10

** * (.0

1) 0. 09

** * (.0

1) 0. 08

** * (.0

B la ck

% 0. 03

(.0 2)

0. 01

(.0 2)

0. 01

(.0 2)

0. 00

(.0 2)

− 0. 01

(.0 2)

0. 01

(.0 2)

0. 01

(.0 2)

0. 00

(.0 2)

− 0. 01

(.0 2)

N on

-B la ck

% − 0. 02

(.0 2)

− 0. 03

* (.0

2) − 0. 03

(.0 2)

− 0. 03

* (.0

2) − 0. 03

(.0 2)

− 0. 03

* (.0

2) − 0. 03

(.0 2)

− 0. 03

* (.0

2) − 0. 03

(.0 2)

R an

do m

ef fe ct s

Ed uc

at or

1. 17

(.0 2)

1. 16

(.0 2)

1. 16

(.0 2)

1. 14

(.0 2)

1. 16

(.0 2)

1. 17

(.0 3)

1. 15

(.0 2)

1. 14

(.0 2)

1. 16

(.0 2)

1. 17

(.0 3)

Sc ho

ol 1. 12

(.0 5)

1. 04

(.0 5)

1. 04

(.0 5)

1. 03

(.0 5)

1. 04

(.0 5)

1. 04

(.0 5)

1. 05

(.0 5)

1. 03

(.0 5)

1. 04

(.0 5)

1. 04

(.0 5)

N ot e.

Th e ta bl e re po

rts fix

ed -e ffe

ct co

ef fic

ie nt s, β,

in te rm

s of

th e lo g of

th e od

ds ra tio

,a nd

eβ w ou

ld be

in te rp re te d as

an od

ds ra tio

.T he

co ef fic

ie nt s ar e re po

rte d w ith

th ei r st at is tic

al si gn

ifi ca nc

e (a st er is ks )a

nd st an

da rd

er ro r (in

pa re nt he

se s) . A lth

ou gh

w e co

nd uc

te d t te st s to

te st

pa ra m et er s,

th e de

gr ee s of

fre ed

om ra ng

ed fro

m 57

9 an

d 16

65 , w hi ch

do no

t di ffe

r ap

pr ec ia bl y fro

m th e st an

da rd

no rm

al di st rib

ut io n

(e .g ., Z te st s) .V

ar ia nc

es w er e no

tt es te d fo r st at is tic

al si gn

ifi ca nc

e. A A 5

A fri ca n A m er ic an

st ud

en ts ;F

R L 5

Pe rc en

to fs tu de

nt s re ce iv in g fre

e an

d/ or

re du

ce d pr ic e lu nc

he s.

* p ,

.0 5.

** p ,

.0 1.

** *p

, .0 01

Behavioral Disorders, 41 (4), 178–195 August 2016 / 185

times as likely as White students to receive a subjective ODR. Near the end of the day, how- ever, African American students were 1.02 times as likely as White students to receive a subjective ODR. The odds ratio for earlier in the day meets the criteria for disproportionality according to the four-fifths rule.

Classroom. Continuing with the same approach to interpretation, Table 2 shows that subjective ODRs are 1.26 times more likely to have been given to African American students than White students within classrooms. The model also shows that subjective ODRs were more likely in classrooms than other settings for both African American students, odds ratio 5 1.47, and White students, odds ratio 5 1.33.

Major ODRs. African American students were also much more likely to receive major subjective ODRs compared to White students, odds ratio 5 1.34. For minor ODRs, the odds ratio associated with AA was 1.1, also greater than 1.0, likely representing important levels of disproportionality but below our four-fifths criterion level for indicating a serious problem.

Gender.Males received three fourths of the ODRs in the sample. Table 3 presents the parti- tioned effects as odds ratios from multilevel logistic regression model, which included the predictors AA, female, and their interactions. The results suggest that African American males were more likely to receive subjective ODRs than White males, odds ratio 5 1.15, but the

TABLE 2 Odds and Odds Ratios (ORs) of a Subjective Referral for Specific Contrasts between African

American (AA) versus White and Three Vulnerable Decision Points (VDPs) from Multilevel Logistic Regression

Analysis Focus Student Race VDP Odds or OR

95% CI Four-Fifths

RuleLower Upper

End of Day White Earlier 12.11 11.38 12.89

White End of day 14.62 13.64 15.67

AA Earlier 15.17 14.22 16.18

AA End of day 14.97 13.86 16.16

AA:White Earlier 1.25 1.21 1.30 .

AA:White End of day 1.02 0.96 1.09 ↔

White End of day:Earlier 1.21 1.16 1.25 ↔

AA End of day:Earlier 0.99 0.94 1.04 ↔

AA:White End of day:Earlier 0.82 0.77 0.87 ↔

Classroom White Other setting 10.80 10.14 11.50

White Classroom 14.38 13.5 15.32

AA Other setting 12.37 11.58 13.21

AA Classroom 18.16 16.99 19.41

AA:White Other setting 1.15 1.10 1.19 ↔

AA:White Classroom 1.26 1.21 1.31 .

White Classroom:Other 1.33 1.29 1.37 .

AA Classroom:Other 1.47 1.41 1.53 .

AA:White Classroom:Other 1.10 1.05 1.16 ↔

Major White Minor 13.28 12.32 14.31

White Major 10.69 9.91 11.55

AA Minor 14.55 13.46 15.73

AA Major 14.37 13.25 15.58

AA:White Minor 1.10 1.05 1.14 ↔

AA:White Major 1.34 1.28 1.41 .

White Major:Minor 0.81 0.78 0.83 ↔

AA Major:Minor 0.99 0.94 1.04 ↔

AA:White Major:Minor 1.23 1.16 1.30 ↔

Note. This table provides the odds or odds ratio from specific contrasts created from the models in Table 1. For rows that contain singular terms (e.g., White, AA, or Earlier), the table reports to odds of a subjective referral. For rows that contain comparison (e.g., AA:White, End of Day:Earlier), the rows provide ORs. Confidence intervals (CIs) that exclude 1.0 indicate a statistically significant result. The four-fifths rule indicates whether a particular OR equals or exceeds (.) four-fifths (1.25), its reciprocal (,; 0.80), or falls within the bounds of the four-fifths rule (↔).

186 / August 2016 Behavioral Disorders, 41 (4), 178–195

risk was much higher for African American females compared to White females, odds ratio 5 1.40. Females also tended to receive fewer subjective ODRs than males, and more so for White females, odds ratio 5 0.74, than African American females, odds ratio 5 0.90.

Gender, AA, and VDPs: Three-Way Interactions

Although the two-way interactions demon- strated disproportionality among the delivery of subjective ODRs to African American students, the differences by gender suggested a more complex interaction. We therefore fit the data to three additional models to tease out differ- ences by gender.

End of day. Table 4 provides the odds ratios associated with the three-way AA6 Female 6 End-of-Day Interaction and all specific sub- group contrasts. The first eight rows give the odds of a subjective ODR by ODR subgroup. The middle 12 rows give odds ratios for one contrast (e.g., AA:White or female:male) within subgroups defined by the other two predictors, and the first four of these rows represent the contrasts of most interest for the present analy- sis. The final seven rows provide odds ratios that represent the two-way interactions and, lastly, the three-way interaction.

This analysis demonstrated that African American females were 1.49 times more likely to receive a subjective ODR before the end of the day than White females. African Ameri- can males were also more likely to receive

subjective ODRs than White males, odds ratio 5 1.19, earlier in the day. African American females, odds ratio 5 1.10, and males, odds ratio 5 1.01, were both more likely to receive subjective ODRs at the end of the day as well, but these risks were much smaller.

Classroom. African American females were more likely to receive subjective ODRs in the classroom, odds ratio 5 1.54, and in other set- tings, odds ratio 5 1.27, than White females. Again, African American males were also at more risk of subjective ODRs in both settings, but at a lower level (see Table 5).

Major ODRs. The final analyses specifically examined risk by rated severity of behavior (major vs. minor). As noted above, African American students were more likely to receive subjective major ODRs compared to White stu- dents, odds ratio 5 1.34. The disaggregation of effects by gender in Table 6 shows that the risk for subjective majors was particularly promi- nent for African American females, odds ratio 5 1.73, although still problematic for African American males, odds ratio 5 1.25.

Supplemental Analyses

We failed to confirm our hypothesis that the end of the day represented a VDP for African American students. Instead, end of day appeared to relate to increased odds of subjective ODRs for White students relative to African American students, which had the effect of reducing disproportionality itself. To further explore the impact of time of day, we tested a competing

TABLE 3 Odds and Odds Ratios (ORs) of a Subjective Referral for Specific Contrasts between African

American (AA) versus White and Females versus Males

Student Race Student Gender Odds or OR

95% CI Four-Fifths

RuleLower Upper

White Male 13.27 12.47 14.11

White Female 9.83 9.21 10.50

AA Male 15.28 14.33 16.30

AA Female 13.80 12.87 14.81

AA:White Male 1.15 1.11 1.19 ↔

AA:White Female 1.40 1.34 1.47 .

White Female:Male 0.74 0.72 0.76 ,

AA Female:Male 0.90 0.87 0.94 ↔

AA:White Female:Male 1.22 1.16 1.29 ↔

Note. This table provides the odds or odds ratio from specific contrasts created from the models in Table 1. For rows that contain singular terms (e.g., White, AA, or Male), the table reports to odds of a subjective referral. For rows that contain comparison (e.g., AA:White, Female:Male), the rows provide ORs. Confidence intervals (CIs) that exclude 1.0 indicate a statistically significant result. The four-fifths rule indicates whether a particular OR equals or exceeds (.) four-fifths (1.25), its reciprocal (,; 0.80), or falls within the bounds of the four-fifths rule (↔).

Behavioral Disorders, 41 (4), 178–195 August 2016 / 187

VDP hypothesis: the start of the school day. Early mornings represent a stressful time of day for teachers, as many struggle to organize their students for academic instruction. Hence, we explored the beginning of the day as an alterna- tive VDP, defining it as the first 90 minutes of the school day (i.e., ODRs delivered from 8:30 to 10:00 a.m.) compared to those delivered later (i.e., 10:15 a.m. to 3:00 p.m.).

These results showed that African Ameri- can students were considerably more likely to receive subjective ODRs in the first 90 min compared to White students, odds ratio 5 1.40. After the school day is underway (i.e., after the first 90 min), African American stu- dents continued to receive more subjective ODRs than White students, odds ratio 5 1.19,

but at a reduced level of disproportionality. The three-way interactions with gender showed that the level of disproportionality was high for both males and females. African American males received more subjective referrals than White males in the first 90 min, odds ratio 5 1.32, but the risk was smaller throughout the rest of the day, odds ratio 5 1.13. The risks were more troublesome for African American females, who were much more likely to receive subjective ODRs than White girls in the first 90 min, odds ratio 5 1.72. African American females, compared to White females, also con- tinued to exceed the four-fifths criterion for dis- proportionality during the remainder of the day, odds ratio 5 1.35.

TABLE 4 Odds and Odds Ratios (ORs) of a Subjective Referral for Specific Contrasts between African American (AA) versus White, Females versus Males, and End of Day versus Earlier in the Day

Student Race

Student Gender VDP

Odds or OR

95% CI Four-Fifths

RuleLower Upper

White Male Earlier 13.01 12.22 13.85

White Male End of day 15.38 14.31 16.53

White Female Earlier 9.38 8.77 10.05

White Female End of day 12.07 11.01 13.23

AA Male Earlier 15.53 14.53 16.60

AA Male End of day 15.60 14.37 16.94

AA Female Earlier 14.02 13.01 15.11

AA Female End of day 13.26 11.93 14.73

AA:White Male Earlier 1.19 1.15 1.24 ↔

AA:White Male End of day 1.01 0.94 1.09 ↔

AA:White Female Earlier 1.49 1.41 1.58 .

AA:White Female End of day 1.10 0.98 1.23 ↔

White Female:Male Earlier 0.72 0.70 0.75 ,

White Female:Male End of day 0.78 0.73 0.85 ,

AA Female:Male Earlier 0.90 0.86 0.95 ↔

AA Female:Male End of day 0.85 0.77 0.94 ↔

White Male End of day:Earlier 1.18 1.13 1.23 ↔

White Female End of day:Earlier 1.29 1.19 1.39 .

AA Male End of day:Earlier 1.00 0.94 1.07 ↔

AA Female End of day:Earlier 0.95 0.86 1.04 ↔

AA:White Female:Male Earlier 1.25 1.18 1.33 .

AA:White Female:Male End of day 1.08 0.95 1.23 ↔

AA:White Male End of day:Earlier 0.85 0.79 0.92 ↔

AA:White Female End of day:Earlier 0.73 0.65 0.83 ,

White Female:Male End of day:Earlier 1.09 1.00 1.19 ↔

AA Female:Male End of day:Earlier 0.94 0.84 1.05 ↔

AA:White Female:Male End of day:Earlier 0.86 0.75 1.00 ↔

188 / August 2016 Behavioral Disorders, 41 (4), 178–195

Discussion

The results demonstrated that, overall, Afri- can American students were more likely to receive subjective ODRs than White students (Figure 1). In addition, African American stu- dents were at greater risk for subjective ODRs thanWhite students in the classroom compared to other settings and when the ODR was per- ceived as a major offense rather than minor. Rather than finding that disproportionality was stable and pervasive throughout all situations, we found specific situations in which dispro- portionality in subjective ODRs was more pro- nounced, as well as situations in which the provision of subjective ODRs was not inequita- ble. These patterns lend support to the VDP

model (McIntosh,Girvan,Horner,& Smolkowski, 2014) and are consistent with research suggesting that some decisions and decision contexts are more influenced by implicit racial biases. Our findings highlight the staff decision to issue major ODRs in the classroom, particularly in the first 90minof the day, as particularly disproportionate and worthy of focused equity intervention.

The results for the end of the day versus earlier, in particular, were not consistent with our hypotheses. The end of the day may have been a poor choice as a VDP given its compe- tition with other times that may have also been VDPs, such as the early morning. Subjective ODRs became more likely at the end of the day for White students, which is consistent with the possibility that teachers accumulate information about particular students over the

TABLE 5 Odds and Odds Ratios (ORs) of a Subjective Referral for Specific Contrasts between African American (AA) versus White, Females versus Males, and Classroom versus Other Settings

Student Race

Student Gender VDP

Odds or OR

95% CI Four-Fifths

RuleLower Upper

White Male Other setting 11.14 10.45 11.87

White Male Classroom 15.97 14.97 17.03

White Female Other setting 9.51 8.85 10.23

White Female Classroom 10.35 9.65 11.11

AA Male Other setting 12.44 11.62 13.32

AA Male Classroom 19.03 17.76 20.39

AA Female Other setting 12.05 11.12 13.05

AA Female Classroom 15.92 14.72 17.21

AA:White Male Other setting 1.12 1.07 1.17 ↔

AA:White Male Classroom 1.19 1.14 1.25 ↔

AA:White Female Other setting 1.27 1.18 1.36 .

AA:White Female Classroom 1.54 1.44 1.64 .

White Female:Male Other setting 0.85 0.82 0.89 ↔

White Female:Male Classroom 0.65 0.62 0.68 ,

AA Female:Male Other setting 0.97 0.91 1.03 ↔

AA Female:Male Classroom 0.84 0.79 0.89 ↔

White Male Classroom:Other setting 1.43 1.39 1.48 .

White Female Classroom:Other setting 1.09 1.03 1.15 ↔

AA Male Classroom:Other setting 1.53 1.46 1.60 .

AA Female Classroom:Other setting 1.32 1.23 1.42 .

AA:White Female:Male Other setting 1.13 1.05 1.22 ↔

AA:White Female:Male Classroom 1.29 1.20 1.39 .

AA:White Male Classroom:Other setting 1.07 1.01 1.13 ↔

AA:White Female Classroom:Other setting 1.21 1.11 1.33 ↔

White Female:Male Classroom:Other setting 0.76 0.71 0.81 ,

AA Female:Male Classroom:Other setting 0.86 0.80 0.94 ↔

AA:White Female:Male Classroom:Other setting 1.14 1.03 1.26 ↔

Behavioral Disorders, 41 (4), 178–195 August 2016 / 189

course of the day, an effect that actually attenu- ated the substantial racial differences seen ear- lier in the day. Instead, we found support in our exploratory analysis for the first 90 min of the school day as a possible VDP. Although specu- lative, it is possible that early morning teacher stress, a stronger academic focus at the start of the day, increased disorder from transitioning between home and school environments and corresponding changes in behavioral expecta- tions, or some combination of these features may explain this finding.

Our gender analyses indicated that subjec- tive ODR rates differed by gender and the intersection of race and gender. The risk of dis- proportionate ODRs associated with VDPs reached problematic levels among African American males versus White males. However,

evidence of disproportionality was particularly strong for African American females compared to White females. One example is in major ODRs, an area with particularly important implications for educational outcomes, because their receipt is more likely to have been associ- ated with the student’s removal from the class- room, leading to less instructional time spent with peers. African American students were 1.34 times as likely to receive a major subjec- tive ODR. African American males were given major ODRs at a rate of 1.25 times as often as their White counterparts, but African American females in particular were much more likely, odds ratio 5 1.73, to receive major ODRs when compared to White females, consistent with previous research (Blake et al., 2011). This finding implies that African American

TABLE 6 Odds and Odds Ratios (ORs) of a Subjective Referral for Specific Contrasts between African

American (AA) versus White, Females versus Males, and Major Referrals versus Minor Referrals

Student Race

Student Gender VDP

Odds or OR

95% CI Four-Fifths

RuleLower Upper

White Male Minor 14.30 13.26 15.43

White Male Major 11.40 10.54 12.32

White Female Minor 10.45 9.64 11.32

White Female Major 8.38 7.66 9.17

AA Male Minor 15.30 14.12 16.58

AA Male Major 14.26 13.12 15.50

AA Female Minor 12.79 11.71 13.98

AA Female Major 14.50 13.12 16.02

AA:White Male Minor 1.07 1.02 1.12 ↔

AA:White Male Major 1.25 1.18 1.32 .

AA:White Female Minor 1.22 1.15 1.31 ↔

AA:White Female Major 1.73 1.58 1.89 .

White Female:Male Minor 0.73 0.70 0.76 ,

White Female:Male Major 0.74 0.69 0.78 ,

AA Female:Male Minor 0.84 0.79 0.89 ↔

AA Female:Male Major 1.02 0.94 1.10 ↔

White Male Major:Minor 0.80 0.77 0.83 ,

White Female Major:Minor 0.80 0.75 0.86 ↔

AA Male Major:Minor 0.93 0.88 0.99 ↔

AA Female Major:Minor 1.13 1.04 1.23 ↔

AA:White Female:Male Minor 1.14 1.06 1.23 ↔

AA:White Female:Male Major 1.38 1.25 1.53 .

AA:White Male Major:Minor 1.17 1.10 1.25 ↔

AA:White Female Major:Minor 1.41 1.27 1.57 .

White Female:Male Major:Minor 1.01 0.94 1.08 ↔

AA Female:Male Major:Minor 1.22 1.10 1.34 ↔

AA:White Female:Male Major:Minor 1.21 1.07 1.36 ↔

190 / August 2016 Behavioral Disorders, 41 (4), 178–195

females were nearly twice as likely as White females to be removed from the classroom dur- ing instruction, which could negatively affect academic achievement.

The pattern of disproportionality we observed at the intersection of race and gender also suggests that the most influential biases may involve paternalistic gender bias (i.e., overlooking violations of female students) and in-group preferences rather than deliberate hos- tility toward students who look less like teach‐ ers (Greenwald & Pettigrew, 2014). Specifically, the result of our analysis suggest that a substan- tial proportion of the disproportionality in ele- mentary schools is a function of teachers having very low odds, relatively speaking, of making referrals of White female students for subjectively defined compared to objectively defined behaviors. Other researchers using different criterion measures have reported similar overall patterns of rates of student disci- pline by race and gender, suggesting the

pattern is fairly robust (Fabelo et al., 2011; Losen et al., 2015).

The observed gender pattern is also consis- tent with the weight of research on decisions in legal settings involving adults. Female criminal defendants and grievants in labor disputes have been found to be treated more favorably than their male counterparts, an effect attributable by psychologists directly to benevolent forms of implicit and explicit sexism (Girvan et al., 2015). Further, although one often thinks of dis- crimination in terms of members of a dominant group seeking to harm members of a minority group, recent examinations of social psycho- logical research on bias suggest that most discrimination occurs because of explicit or implicit motivations to favor in-group members (Greenwald & Pettigrew, 2014; Smith, Levin- son, & Robinson, 2014). Indeed, many of the interventions that have been shown to be the most effective at reducing implicit bias work primarily because they reduce positive implicit

Figure 1. Comparisons of potential VDPs. The columns at the bottom represent the average odds of an ODR for a subjectively versus objectively defined behavior by student race for each clustered condition. The diamonds linked by lines represent the odds ratio within each cluster (data for odds and odds ratios come from the first column of data in Tables 2 and 3). Odds ratios of 1 indicate no disproportionality in the clustered condition. The darker horizontal line at an odds ratio of 1.25 indicates the threshold above which we interpret the magnitude of disproportionality to be particularly problematic.

Behavioral Disorders, 41 (4), 178–195 August 2016 / 191

in-group associations (Lai et al., 2013). Because roughly 76% of teachers are female and 82% areWhite (NCES, 2014), we estimate that about 62% of teachers were White females and 20% White males, which contrasts sharply with the estimated 7% African American teachers, or 5% African American female teachers and 2% African American male teachers. Due to their group membership or paternalistic attitudes towards certain groups in certain contexts (Sin- clair & Kunda, 1999, 2000), teachers may be less inclined to categorize the behavior of White female students in particular as meriting a disciplinary response than they would African American female students or male students in general. In any case, the operation of these biases appeared to be strongest in situations that aligned with VDPs. Further, these findings point to another possible VDP, that teachers may need to be particularly careful when asses- sing the behavior of students who are further from their in-group, in terms of race, gender, or other characteristics.

Limitations and Future Research

This study was limited by extant data from a nonrandom sample of elementary schools. Analyses should be repeated with samples of middle and secondary schools to assess the consistency of VDPs across settings. Although schools contributed data from nearly all the U.S. states, we suggest caution in generalizing results, particularly beyond schools using a standardized system for tracking ODRs. Like- wise, our analyses were limited to recorded ODRs, and therefore we could not measure behavior that met criteria for ODR but did not result in ODRs. Future research could include direct observation of behavior and analyses of whether students received ODRs at all. The analysis was also correlational in nature and does not allow for a causal interpretation. Nonetheless, the research hypotheses pro- posed specific patterns that were, within the precision of our VDP measures, falsifiable and subsequently supported. In addition, the VDPs represented in this research consist of proxies for potentially more accurately defined VDPs. Subsequent research should address these lim- itations and include more information about teachers’ race and gender and the school day (e.g., exactly when teachers may be most fatigued or hungry). Schools may also have unique patterns of VDPs that do not con- form to these general VDPs (e.g., different

schedules). Finally, the promise of the VDP model is limited until intervention research can confirm its intervention validity. Such research is currently underway.

Implications for Practice

Despite these limitations, the results provide tentative support for the importance of consider- ing VDPs as important variables in understand- ing and reducing school discipline disparities. The findings suggest two avenues. First, improv- ing the specificity of definitions of subjective ODRs such as defiance and disrespect (i.e., pro- viding definitions that reduce ambiguity as much as possible) could attenuate the influence of implicit bias on discipline decisions. School personnel can decrease (but not eliminate) subjectivity by creating and using operational definitions of each behavior, as well as the thresholds for no ODR, a minor ODR, and a major ODR. Second, these general VDPs (i.e., first 90 min of the day, classroom, assessing severity) could be used to help educators identi- fy specific decisions that are vulnerable to bias and use alternative responses in place of issuing ODRs that perpetuate disproportionality. Once they are aware of these VDPs, teachers may be trained in responses that are more instructional than exclusionary, such as teaching or reteach- ing expectations or visibly modeling cool- down strategies for students (McIntosh, Ellwood, & McCall, 2016; McIntosh, Girvan, Horner, Smolkowski, & Sugai, 2014). Administrators can be encouraged to use more instructional or restorative alternatives to suspension (Nese, Massar, & McIntosh, 2015) and use interven- tions such as Check-in Check-out, which have been shown to be effective with African Ameri- can students (Vincent, Tobin, Hawken, & Frank, 2012). Schools can also use their own discipline data to identify school-specific VDPs (McIntosh, Barnes, Morris, & Eliason, 2014), which could be even more effective. In addition, professional development can help educators identify and counteract their own personal VDPs. For exam- ple, if individuals make less equitable discipline decisions when they are stressed in the early morning, they can use this knowledge as a cue to slow the decision-making process during these moments. Finally, preventive approaches, such as proactively teaching classroom rou- tines, using acknowledgement systems equita- bly, and enhancing the level of student engagement in classroom instruction, may prevent VDPs in the first place (Chaparro,

192 / August 2016 Behavioral Disorders, 41 (4), 178–195

Nese, & McIntosh, 2015; Tobin & Vincent, 2011). Although promising, these implications should be tested through rigorous intervention research.

REFERENCES Anyon, Y., Jenson, J. M., Altschul, I., Farrar, J.,

McQueen, J., Greer, E., … Simmons, J. (2014). The persistent effect of race and the promise of alternatives to suspension in school discipline out- comes. Children and Youth Services Review, 44, 379–386. doi:10.1016/j.childyouth.2014.06.025

Blake, J. J., Butler, B. A., Lewis, C., & Darensbourg, A. (2011). Unmasking the inequitable discipline experiences of urban Black girls: Implications for urban educational stakeholders. The Urban Review, 43, 90–116. doi:10.1007/s11256-009- 0148-8

Bobo, L., Charles, C., Krysan,M.,& Simmons,A. (2012). The real record on racial attitudes. In P. V.Marsden (Ed.), Social trends in American life: Evidence from the General Social Survey since 1972 (pp. 38–83). Princeton, NJ: Princeton University Press.

Bradshaw, C. P., Mitchell, M. M., O’Brennan, L. M., & Leaf, P. J. (2010). Multilevel exploration of factors contributing to the overrepresentation of black students in office disciplinary refer- rals. Journal of Educational Psychology, 102, 508–520. doi:10.1037/a0018450

Chaparro, E. A., Nese, R. N. T., & McIntosh, K. (2015). Examples of engaging instruction to increase equity in education. Eugene, OR: Cen- ter on Positive Behavioral Interventions and Sup- ports, University of Oregon.

Crenshaw, K. W., Ocen, P., & Nanda, J. (2015). Black girls matter: Pushed out, overpoliced, and under- protected. New York: African American Policy Forum.

Danziger, S., Levav, J., & Avnaim-Pesso, L. (2011). Extraneous factors in judicial decisions. Proceed- ings of the National Academy of Sciences, 108, 6889–6892. doi:10.1073/pnas.1018033108

Devine, P. G. (1989). Stereotypes and prejudice: Their automatic and controlled components. Journal of Personality and Social Psychology, 56, 5–18. doi:10.1037/0022-3514.56.1.5

Dovidio, J. F., & Gaertner, S. L. (2000). Aversive rac- ism and selection decisions: 1989 and 1999. Psy- chological Science, 11, 315–319. doi:10.1111/ 1467-9280.00262

Equal Employment Opportunity Commission, Civil Service Commission, Department of Labor, & Department of Justice. (1978). Uniform guide- lines on employee selection procedures. Federal Register, 43, 38290–38315.

Fabelo, T., Thompson, M. D., Plotkin, M., Carmichael, D., Marchbanks, M. P. I., & Booth, E. A. (2011). Breaking schools’ rules: A statewide study of how school discipline relates to students’ success and juvenile justice involvement. New York: Council of State Governments Justice Center.

Fazio, R. H., Jackson, J. R., Dunton, B. C., & Wil- liams, C. J. (1995). Variability in automatic acti- vation as an unobtrusive measure of racial attitudes: A bona fide pipeline? Journal of Person- ality and Social Psychology, 69, 1013–1027. doi:10.1037/0022-3514.69.6.1013

Fenning, P., & Rose, J. (2007). Overrepresentation of African American students in exclusionary disci- pline: The role of school policy. Urban Education, 42, 536–559. doi:10.1177/0042085907305039

Gailliot, M. T., Peruche, B. M., Plant, E. A., & Bau‐ meister, R. F. (2009). Stereotypes and prejudice in the blood: Sucrose drinks reduce prejudice and stereotyping. Journal of Experimental Social Psychology, 45, 288–290. doi:10.1016/j.jesp. 2008.09.003

Girvan, E. J. (2015).Onusing the psychological science of implicit bias to advance anti-discrimination law. GeorgeMason University Civil Rights Law Journal, 26, 1–86.

Girvan, E. J., Deason, G., & Borgida, E. (2015). The gen- eralizability of gender bias: Testing the effects of contextual, explicit, and implicit theories of sexism on labor arbitration decisions. Law and Human Behavior, 39, 525–537. doi:10.1037/lhb0000139

Greenwald, A. G., & Banaji, M. R. (1995). Implicit social cognition: Attitudes, self-esteem, and stereo- types. Psychological Review, 102, 4–27. doi: http://dx.doi.org/10.1037/0033-295X.102.1.4

Greenwald, A. G., & Pettigrew, T. F. (2014). With mal- ice toward none and charity for some: Ingroup favoritism enables discrimination. American Psy- chologist, 69, 669–684. doi:http://dx.doi.org/10. 1037/a0036056

Greenwald, A. G., Poehlman, T. A., Uhlmann, E. L., & Banaji, M. R. (2009). Understanding and using the Implicit Association Test: III. Meta-analysis of pre- dictive validity. Journal of Personality and Social Psychology, 97, 17–41. doi:http://dx.doi.org/10. 1037/a0015575

Greflund, S., McIntosh, K., Mercer, S. H., & May, S. L. (2014). Examining disproportionality in school discipline for Aboriginal students in schools implementing PBIS. Canadian Journal of School Psychology, 29, 213–235. doi:10.1177/ 0829573514542214

Gregory, A., Skiba, R. J., & Noguera, P. A. (2010). The achievement gap and the discipline gap: Two sides of the same coin? Educational Researcher, 39, 59–68. doi:10.3102/0013189x09357621

Irvin, L. K., Tobin, T. J., Sprague, J. R., Sugai, G., & Vincent, C. G. (2004). Validity of office discipline referral measures as indices of school-wide behavioral status and effects of school-wide behavioral interventions. Journal of Positive Behavior Interventions, 6, 131–147. doi:10.1177/ 10983007040060030201

Johnson, J., Whitestone, E., Jackson, L., & Gatto, L. (1995). Justice is still not colorblind: Differential racial effects of exposure to inadmissible evi- dence. Personality and Social Psychology Bulletin, 21, 893–898. doi:10.1177/0146167295219003

Behavioral Disorders, 41 (4), 178–195 August 2016 / 193

Judge, T. A., & Cable, D. M. (2004). The effect of physical height on workplace success and income: preliminary test of a theoretical model. Journal of Applied Psychology, 89, 428–441. doi: http://dx.doi.org/10.1037/0021-9010.89.3.428

Kouchaki, M., & Smith, I. H. (2014). The morning morality effect: The influence of time of day on unethical behavior. Psychological Science, 25, 95–102. doi:10.1177/0956797613498099

Lai, C. K., Hoffman, K. M., Nosek, B. A., & Green- wald, A. G. (2013). Reducing implicit prejudice. Social and Personality Psychology Compass, 7, 315–330. doi:10.1111/spc3.12023

Laraway, S., Snycerski, S., Michael, J., & Poling, A. (2003). Motivating operations and terms to describe them: Some further refinements. Journal of Applied Behavior Analysis, 36, 407–414. doi:10.1901/jaba.2003.36-407

Lerner, J. S., & Tetlock, P. E. (1999). Accounting for the effects of accountability. Psychological Bulletin, 125, 255–275. doi:http://dx.doi.org/10.1037/0033- 2909.125.2.255

Linder, J. A., Doctor, J. N., Friedberg, M. W., Nieva, H. R., Birks, C., Meeker, D., & Fox, C. R. (2014). Time of day and the decision to prescribe antibiotics. JAMA Internal Medicine, 174, 2029– 2031. doi:10.1001/jamainternmed.2014.5225

Losen, D. J., & Gillespie, J. (2012). Opportunities sus- pended: The disparate impact of disciplinary exclusion from school. Los Angeles, CA: Center for Civil Rights Remedies at The Civil Rights Proj- ect at UCLA.

Losen, D. J., Hodson, C., Keith, M. A., Morrison, K., & Belway, S. (2015). Are we closing the school disci- pline gap? Los Angeles, CA: Center for Civil Rights Remedies at The Civil Rights Project at UCLA.

Martinez, T. (2013, January). Difference-making: Why education policymakers should attend to frequent and disparate disciplinary exclusion from school. Symposium presented at the Clos- ing the School Discipline Gap: Research to Prac- tice Conference, Washington, DC.

May, S. L., Ard, W. I., Todd, A. W., Horner, R. H., Glasgow, A., Sugai, G., & Sprague, J. R. (2013). School-wide information system. Eugene, OR: Educational and Community Supports, University of Oregon. Retrieved from http:/www.pbisapps.org

McIntosh, K., Barnes, A., Morris, K., & Eliason, B. M. (2014). Using discipline data within SWPBIS to identify and address disproportionality: A guide for school teams. Eugene, OR: OSEP Technical Assistance Center on Positive Behavioral Inter- ventions and Supports, University of Oregon.

McIntosh, K., Campbell, A. L., Carter, D. R., & Zumbo, B. D. (2009). Concurrent validity of office discipline referrals and cut points used in schoolwide positive behavior support. Behavior- al Disorders, 34, 100–113.

McIntosh, K., Ellwood, K., & McCall, L. (2016). Using discipline data within a PBIS framework to enhance equity in school discipline. Manuscript submitted for publication.

McIntosh, K., Girvan, E. J., Horner, R. H., & Smol‐ kowski, K. (2014). Education not incarceration: A conceptual model for reducing racial and eth- nic disproportionality in school discipline. Jour- nal of Applied Research on Children, 5(2), 1–22.

McIntosh, K., Girvan, E. J., Horner, R. H., Smolkowski, K., & Sugai, G. (2014). Recommendations for addressing discipline disproportionality in educa- tion. Eugene, OR: OSEP Technical Assistance Center on Positive Behavioral Interventions and Supports.

Merlo, J., Chaix, B., Ohlsson, H., Beckman, A., Johnell, K., Hjerpe, P.,… Larsen, K. K. (2006). A brief con- ceptual tutorial of multilevel analysis in social epi- demiology: Using measures of clustering in multilevel logistic regression to investigate contex- tual phenomena. Journal of Epidemiology and Community Health, 60, 290–297. doi:10.1136/ jech.2004.029454

National Center for Education Statistics. (2014). Digest of education statistics. Washington, DC: U.S. Department of Education.

Nese, R. N. T., Massar, M. M., & McIntosh, K. (2015). Alternatives to suspension: Strategies for school principals. Principal Leadership, 16, 52–56.

Nosek, B. A., Greenwald, A. G., & Banaji, M. R. (2007). The Implicit Association Test at age 7: A methodological and conceptual review. In J. A. Bargh (Ed.), Social psychology and the uncon- scious: The automaticity of higher mental process- es (pp. 265–292). New York: Psychology Press.

Okonofua, J. A., Walton, G. M., & Eberhardt, J. L. (2016). A vicious cycle: A social–psychological account of extreme racial disparities in school dis- cipline. Perspectives on Psychological Science, 11, 381–398. doi:10.1177/1745691616635592

Pearson, A. R., Dovidio, J. F., & Gaertner, S. L. (2009). The nature of contemporary prejudice: Insights from aversive racism. Social and Personality Psychology Compass, 3, 314–338. doi:10.1111/ j.1751-9004.2009.00183.x

Pettigrew, T. F., & Tropp, L. R. (2006). A meta-analytic test of intergroup contact theory. Journal of Per- sonality and Social Psychology, 90, 751–783. doi: http://dx.doi.org/10.1037/0022-3514.90.5.751

Saucier,D.A.,Miller, C. T.,&Doucet,N. (2005).Differ- ences in helping whites and blacks: A meta-analy- sis. Personality And Social Psychology Review, 9, 2–16. doi:10.1207/s15327957pspr0901_1

Sinclair, L., & Kunda, Z. (1999). Reactions to a Black professional: Motivated inhibition and activa- tion of conflicting stereotypes. Journal of Per‐ sonality and Social Psychology, 77, 885–904. doi: http://dx.doi.org/10.1037/0022-3514.77.5.885

Sinclair, L., & Kunda, Z. (2000). Motivated stereotyp- ing of women: She’s fine if she praised me but incompetent if she criticized me. Personality and Social Psychology Bulletin, 26, 1329–1342. doi:10.1177/0146167200263002

Skiba, R. J., Michael, R. S., Nardo, A. C., & Peterson, R. L. (2002). The color of discipline: Sources of racial and gender disproportionality in school

194 / August 2016 Behavioral Disorders, 41 (4), 178–195

punishment. The Urban Review, 34, 317–342. doi:10.1023/A:1021320817372

Skiba, R. J., Poloni-Staudinger, L., Simmons, A. B., Feggins, L. R., & Chung, C. G. (2005). Unprov- en links: Can poverty explain ethnic dispropor- tionality in special education? Journal of Special Education, 39, 130–144. doi:10.1177/ 00224669050390030101

Smith, R. J., Levinson, J. D., & Robinson, Z. (2014). Implicit white favoritism in the criminal justice system. Alabama Law Review, 66, 871–1169.

Staats, C. (2014). State of the science: Implicit bias review 2014. Columbus, OH: Kirwan Institute.

Sugai, G., Sprague, J. R., Horner, R. H., & Walker, H. M. (2000). Preventing school violence: The use of office discipline referrals to assess and monitor school-wide discipline interventions. Journal of Emotional and Behavioral Disorders, 8, 94–101. doi:10.1177/106342660000800205

Tobin, T. J., & Vincent, C. G. (2011). Strategies for preventing disproportionate exclusions of Afri- can American students. Preventing School Fail- ure, 55, 192–201. doi:10.1080/1045988X.2010. 532520

U.S. Government Accountability Office. (2013). Stan- dards needed to improve identification of racial and ethnic overrepresentation in special educa- tion. Report to the chairman, Committee on Health, Education, Labor, and Pensions, U.S. Senate (Report GAO-13-137). Washington, DC: Author.

Vincent, C. G., Tobin, T. J., Hawken, L. S., & Frank, J. L. (2012). Discipline referrals and access to secondary level support in elementary and middle schools: Patterns across African-American,

Hispanic-American, and white students. Educa- tion and Treatment of Children, 35, 431–458. doi:10.1353/etc.2012.0018

Walker, B., Cheney, D., Stage, S. A., & Blum, C. (2005). Schoolwide screening and Positive Behavior Supports: Identifying and supporting students at risk for school failure. Journal of Positive Behavior Interventions, 7, 194–204. doi:10.1177/ 10983007050070040101

Wallace, J. M. J., Goodkind, S., Wallace, C. M., & Bachman, J. G. (2008). Racial, ethnic, and gen- der differences in school discipline among U.S. high school students: 1991–2005. Negro Educa- tional Review, 59, 47–62.

AUTHOR’S NOTE

The development of this chapter was supported by the Office of Special Education Programs, U.S. Department of Education (#H326S130004). The opinions expressed are those of the authors and do not represent views of the Office or U.S. Department of Education.

Address correspondences to Kent McIntosh, 1235 University of Oregon, Eugene, OR, 97403; E-mail: [email protected].

MANUSCRIPT

Initial Acceptance: 6/11/2016 Final Acceptance: 6/23/2016

Behavioral Disorders, 41 (4), 178–195 August 2016 / 195

Copyright of Behavioral Disorders is the property of Council for Children with Behavioral Disorders and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.