Citation
Disproportionality of Minority Students Identified with an Emotional/behavioral Disorder

Material Information

Title:
Disproportionality of Minority Students Identified with an Emotional/behavioral Disorder Examining Teachers' Ratings of Students' Behavior and Factorial Equivalence of a Behavior Rating Scale
Creator:
Peters, Christina
Place of Publication:
[Gainesville, Fla.]
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (109 p.)

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
School Psychology
Special Education, School Psychology and Early Childhood Studies
Committee Chair:
Kranzler, John H.
Committee Members:
Joyce, Diana
Algina, James J.
Smith, Stephen W.
Graduation Date:
8/7/2010

Subjects

Subjects / Keywords:
African American culture ( jstor )
African Americans ( jstor )
Minority group students ( jstor )
Multilevel models ( jstor )
Schools ( jstor )
Social skills ( jstor )
Special education ( jstor )
Students ( jstor )
Teachers ( jstor )
Variable coefficients ( jstor )
Special Education, School Psychology and Early Childhood Studies -- Dissertations, Academic -- UF
assessment, cab, ebd, efficacy, ese, evaluation, representation
Genre:
Electronic Thesis or Dissertation
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
School Psychology thesis, Ph.D.

Notes

Abstract:
The disproportionate representation of racial/ethnic minority students identified with special education needs has been a topic of concern for 40 years. Disproportionate representation is the extent to which membership in a given ethnic group affects the probability of being placed in a specific special education disability category (Oswald, Coutinho, Best, & Singh, 1999, p.198). An assumption is that the percentage of a particular group identified with a special education need equals the percentage of that group in the student population. This study examined several hypotheses to better understand disproportionate representation of racial/ethnic minority students identified with an emotional/behavioral disorder in the school system, including teacher bias. Previously, researchers have examined teacher expectancy in addition to the completion of behavior rating scales. The purpose of this study was to build upon current literature focusing on the completion of behavior rating scales. Sixty-five teachers completed the Clinical Assessment of Behavior Teacher Form (CAB-T) for a sample of 982 Caucasian, African American, and Hispanic American students. Four outcome variables from the CAB-T were assessed (i.e., internalizing, externalizing, social skills, competence). Hierarchical linear modeling techniques were used to (1) analyze variance components across three levels (2) examine mean-group differences across outcome variables for student gender, race/ethnicity, and free/reduced price lunch status, (3) examine differences across school factors, and (4) examine whether teacher variables (i.e., age, years experience, gender, race/ethnicity, self-efficacy) predicted teacher-specific differences in ratings. Results indicated that a significant amount of variance in ratings was attributable to the teacher- and school- levels. Several mean-group differences emerged (e.g., African Americans were rated by teachers as exhibiting more externalizing behaviors than their Caucasian peers). Some teacher specific differences in ratings across groups were predicted by teacher self-efficacy for behavior management and teacher age, but not teacher race/ethnicity, gender, or years experience. A final purpose was to examine the psychometric properties of the CAB-T across racial/ethnic groups to ensure that the same factors are measured across groups. The results from Confirmatory Factor Analyses of Caucasian and African American samples indicated similar goodness-of-fit indices for both groups although a poor fit of the model to the data. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (Ph.D.)--University of Florida, 2010.
Local:
Adviser: Kranzler, John H.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2011-02-28
Statement of Responsibility:
by Christina Peters.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Embargo Date:
2/28/2011
Resource Identifier:
004979663 ( ALEPH )
769019241 ( OCLC )
Classification:
LD1780 2010 ( lcc )

Downloads

This item has the following downloads:


Full Text





Social Skills

Table 3-16 displays the coefficients for the Level 1 variables in the hierarchical

linear model with the Social Skills scale as the outcome variable. The Social Skills

intercept coefficient was 52.502. Students who were eligible for free or reduced-priced

lunch received lower social skills ratings than their peers who were not eligible for free

or reduced-priced lunch, although the coefficient was not significant. African

Americans received significantly lower social skills ratings than their Caucasian peers.

Hispanic Americans received higher social skills ratings than their Caucasian peers,

although this difference was not significant. Finally, males received significantly higher

social skills ratings than their female peers.

Table 3-17 displays the coefficients for the Level 2 variables in the hierarchical

linear model with the Social Skills scale as the outcome variable. Teachers with more

years experience produced significantly lower social skills ratings. Teachers with a

higher belief that external influences beyond teachers' control affect student outcomes

produced significantly higher social skills ratings. Conversely, teachers with a higher

sense of personal self-efficacy produced significantly lower social skills ratings. The

coefficients for teacher age, self-efficacy for behavior management, gender, and

race/ethnicity were not significant.

Table 3-18 displays the coefficients for the cross-level interactions of the teacher-

level variables and the difference between African American and Caucasian students on

the Social Skills scale. There is a significant interaction for teacher self-efficacy for

behavior management. From Table 3.16, the mean social skills difference between

African American and Caucasian students is -2.886 points (African Americans rated as

exhibiting fewer social skills) for teachers whose management self-efficacy is at the









behavior management plans, behavioral interventions, mental health services, anger

management training, substance abuse counseling, and social work services. However,

researchers have identified several unintended consequences associated with being

identified as EBD. Students with EBD as a group, compared to their peers, experience

less school success and poorer outcomes than any other group of children, with or

without disabilities (Blackorby & Wagner, 1996; Bradley, Henderson, & Monroe, 2004;

Landrum, Tankersley, & Kaufman, 2001). Students with EBD have lower levels of

academic achievement than students in general education and special education

students with SLD (Lane, Barton-Arwood, Nelson, & Wehby, 2008; Wagner & Cameto,

2004). In addition, the academic deficits of students with EBD persist and, in some

cases, worsen over time (Bradley, Doolittle, & Bartolotta, 2008; Nelson, Benner, Lane, &

Smith, 2004). Bradley et al. (2008) found that 75% of students with EBD were below

expected grade level in reading and 97% were below expected grade level in math.

Further, students with EBD are more likely than general education and other special

education students to drop out of high school. In 2003, only 35.4% of students with EBD

aged 14 or older graduated from high school with a regular diploma, compared to 36.9%

of students with mental retardation, 45.3% of children with multiple disabilities, and

57.4% of children with specific learning disabilities (U. S. Department of Education,

2005). Finally, students with EBD have high rates of suspension and expulsion (Bradley

et al., 2008)

In addition to these sequela, students with EBD also experience negative social

and emotional outcomes. Students with EBD exhibit lower social and communication

skills than their peers (Wagner, 1995; Wagner, Kutash, Duchnowski, Epstein, & Slum,









(e.g., Hillsborough County, FL), the student's response to the interventions) is

monitored. If the student does not make adequate progress in response to the

interventionss, or if it is determined that the intensity of the interventions) is beyond

what can be maintained in the general education classroom, the student is referred for a

psychological evaluation (utilizing formal data collection techniques such as behavior

rating scales). Students who are referred by their teachers have a high probability of

identification for special education services, including those services under the EBD

label (Algozzine, Christenson, & Yesseldyke, 1982). However, as several steps

(described above) occur and several members are associated with the identification

process, the probability of being identified as EBD involves many more variables than

teacher referral.

Prevalence of Emotional/Behavioral Disorders

Lane et al. (2008) estimate that 2 20% of students are likely to have difficulties

with emotional or behavioral functioning, depending on what criteria are examined.

Each year, the Department of Education collects data on the number of students

receiving special education services under IDEA (U. S. Department of Education, 2005).

In 2003, 9.1% of students in the U.S.A. aged 6 through 21 received special education

services under part B of IDEA.

Figure 1-1 displays the percentage of students receiving special education

services under the label of a Specific Learning Disability (SLD), Speech or Language

Impairment (S/L), Intellectual Disability (IND), Emotional/Behavioral Disorder (EBD),

Other Health Impairment (OHI), and other disabilities out of all students eligible for

exceptional student education services. Of students aged 6 through 21 who received

special education services, 8.0% received services under the label of EBD. As Figure 1









CHAPTER 3
RESULTS

The results of this study are presented in four sections. The first section presents

descriptive statistics for the variables of interest in this study (i.e., CAB-T and TECMD).

The second section includes Random ANOVA models for the four CAB-T outcome

variables (INT, EXT, SOC, and COM). The third section includes Intercepts and Slopes

as Outcomes models for the four CAB-T outcome variables. Finally, results for the

confirmatory factor analyses by ethnicity are presented. A discussion of these results is

presented in Chapter 4.

Descriptive Statistics

Clinical Assessment of Behavior

Table 3-1 displays descriptive statistics for the four outcome variables measured in

this study. The mean T scores for internalizing and externalizing were lower than the

norm group mean of 50. The mean T scores for social skills and competence were

higher than the norm group mean of 50.

Teacher Efficacy in Classroom Management and Discipline Scale

Table 3-2 displays descriptive statistics for the three self-efficacy variables

measured via the Teacher Efficacy in Classroom Management and Discipline Scale

(TECMD). Teachers displayed the highest self-efficacy scores for self-efficacy in

behavior management. The next highest score was self-efficacy for personal teaching

efficacy. Finally, teachers demonstrated the lowest self-efficacy for external influences.

Random ANOVA Hierarchical Linear Models

Table 3-3 displays the results from the three-level random ANOVA models (i.e., no

predictors) for the four outcome variables in the study. The random ANOVA models









for those reasons. The authors concluded that the mean group differences in teachers'

ratings of students were better explained by teacher bias since mean group differences

in behavior ratings were evident as early as kindergarten (Downey & Pribesh, 2004).

Additional variables that may have predicted mean group differences in behavior were

not examined.

Ainsworth-Darnell and Downey (1998) also examined the oppositional culture

theory by testing each of 4 hypotheses associated with the theory: (a) that involuntary

minority students (e.g., African Americans) value education less because they are lack

optimism about their future opportunities, (b) involuntary minority students display more

resistance to school than their immigrant minority (e.g., Asian Americans) peers, (c)

involuntary minority students who are high achieving are looked down upon by their

peers, and (d) the achievement gap noticed between involuntary minorities and their

immigrant minority and Caucasian peers is due to the resistance to school experienced

by the involuntary minorities. They found little support for three out of four hypotheses of

the theory, but found mixed results on the fourth hypothesis. Their results found that

teachers viewed African American students as putting forth significantly less effort and

being significantly more disruptive than Caucasian students. However, African American

students themselves reported more positive attitudes towards school than Caucasian

students, were significantly less likely to agree that it is OK to break rules, and were

significantly more likely to report satisfaction from doing what is expected of them in

class (Ainsworth-Darnell & Downey, 1998). Because the results of these studies do not

support Ogbu's (1991) oppositional culture theory, his theory is not appropriate to

describe the overrepresentation of minority students identified for special education.









students on the subscales were not predicted by an interaction of student and teacher

race/ethnicity. Finally, Abidin and Robinson (2002) assessed teachers' perceptions of

students' behaviors and referral judgments of students with behavioral difficulties. The

authors asked teachers to complete the Achenbach Teacher Report Form (Achenbach,

1991) and the Social Skills Rating System (Gresham & Elliot, 1990) and the results

indicated that there were no statistically significant differences among students' gender,

race/ethnicity, SES, and age on either of the behavioral rating scales.

Most studies examining the interaction effects of teacher race/ethnicity and

student race/ethnicity analyzed data using multivariate analysis of variance (MANOVA)

or analysis of covariance (ANCOVA). In these studies, teachers only rated one student

or a few students in their classes. Little research has examined the effects when

teachers rate all students in their classes. When teachers rate all students in their

classes, it is possible to examine whether variance in mean-group differences (e.g.,

African American students being rated as behaving worse than Caucasian students)

can be attributed to the teacher or the classroom. Mashburn, Hamre, Downer, and

Pianta (2006) set out to explore within-class and between-class sources of variance in

pre-kindergarten teachers' ratings of students' social competence. In addition, they

explored the effects of teacher variables related to professional background (e.g., years

of experience), psychological characteristics (e.g., depression, self-efficacy), and

characteristics of pre-K classrooms (e.g., child-teacher ratio). In order to explore these

topics, the authors utilized hierarchical linear modeling (HLM, Raudenbush & Bryk,

2002). In HLM, an intraclass correlation coefficient is a measure of how much variance

is accounted for by the inclusion of a particular variable. In addition, analyses can be









behaviors (e.g., conduct problems) using the Conners Teacher Rating Scale (Epstein,

March, Conners, & Jackson, 1998). Sbarra and Pianta (2001) examined teachers'

ratings of 540 children during kindergarten and first grade while controlling for mothers'

education (as a measure of socioeconomic status) and students' gender. Their results

revealed that African American students were rated as having worse behavior problems

and less competence than their Caucasian peers (Sbarra & Pianta, 2001). Finally, using

the Achenbach Behavior Rating Scales, Lau, Garland, Yeh, McCabe, Wood, and Hough

(2004) found that teachers rated Caucasian students as having more internalizing

problems and African American students as having more externalizing problems, even

though the students' ratings of themselves indicated little variability by race/ethnicity.

Downey and Pribesh (2004) mentioned that African American students may be

rated as displaying worse behavior than their Caucasian peers because their actual

behavior is worse. This phenomenon leads researchers to examine variables that may

explain why African American students are rated less favorably. Some researchers

have examined the interaction effects of both teacher and student race/ethnicity on

teachers' ratings of students' behaviors (Cullinan & Kauffman, 2005; Pigott & Cowen,

2000; Rayfield, 1997; Rong, 1996). The findings of these research studies have been

mixed, with some research supporting a claim of teacher bias towards racial/ethnic

minority students (because teacher ratings vary by teacher race/ethnicity) and other

research that shows no differences in teacher ratings between Caucasian and

racial/ethnic minority teachers.

Utilizing the norm sample from the Behavioral Assessment System for Children

(BASC; Reynolds & Kamphaus, 1992), Rong (1996) examined the interaction of









variable of interest in order to examine interaction effects between teachers' years of

teaching experience and teacher-specific differences in behavior ratings.

Self-Efficacy

Teachers' self-efficacy was measured via the Teacher Efficacy in Classroom

Management and Discipline Scale (TECMD; Emmer & Hickman, 1991). Three

components of self-efficacy including classroom management and discipline, external

influences, and personal teaching efficacy were included to examine the interaction

effects between teachers' self-efficacy and teacher-specific differences in behavior

ratings.

School-Level Data

School-level data were included to assess variance components at the school

level and to determine whether differences in ratings are evident across school

variables. School-level variables were not included as predictors of differences in

ratings as three-level interactions among student, teacher, and school variables become

very difficult to interpret.

School Size

The size of each school was obtained from each school district. School size is

included as a school-level variable as research has indicated school size to predict

other important outcomes (Koth, Bradshaw, & Leaf, 2008; Schalock, Holl, Elliott, &

Ross, 1992). I was interested in whether differences in ratings varied across school

size.

Average Family Income

The percentage of students eligible for free or reduced-price lunch in the school

was included as a school-level variable. Average family income is included as a school-









students eligible for free/reduced price lunch on internalizing behaviors. I was surprised

to find that Hispanic American students were rated as significantly less internalizing

than their Caucasian peers. In a study examining ratings for Hispanic and Caucasian

students related to Attention Deficit/Hyperactivity Disorder, Dominguez and Shapiro

(1998) found a consistent, yet non-significant trend across several measures and

subscales indicating that Hispanic students received lower ratings than their Caucasian

peers. They hypothesized that Hispanic families socialize children by instilling values

related to obedience and rule following.

Conversely, Glover, Pumariega, Holzer, Wise, and Rodriguez (1999) conducted

a study in which adolescents completed self-report measures. Two groups in Texas

were compared: one group consisting of 94% Mexican Americans and another group

consisting of Mexican Americans, Caucasian Americans, and African Americans. They

found that the predominately Mexican American group indicated significantly higher

levels of anxiety and that students born outside of the United States also indicated

significantly higher levels of anxiety than those born in the States. They hypothesized

that Hispanic Americans, particularly females, who are experiencing acculturation

express more feelings of internalized distress. It may be a possibility that our teachers

are not picking up on some indicators of anxiety and depression in Hispanic American

students and that may contribute to underrepresentation of Hispanic American youth

identified with EBD (Cullinan & Kauffman, 2005).

Regarding externalizing behaviors, I found that African American students were

rated as significantly more externalizing than their Caucasian peers while controlling for

gender and FRL status. This is consistent with previous research (e.g., Epstein, March,









students received significantly lower competence ratings than their Caucasian peers

while controlling for gender and FRL status. Finally, I found that males received

significantly higher competence ratings than their female peers controlling for their

race/ethnicity and FRL status. The link between lower competence ratings and FRL

status is no surprise. Barbarin et al. (2006) found that students from poor families

displayed significantly lower receptive language and math scores than their peers.

Additionally, they found that students who were not from poor families scored higher on

all preacademic tasks when assessed at the beginning of pre-kindergarten.

Finally, larger schools were associated with higher ratings of internalizing

behaviors, higher ratings of externalizing behaviors, and lower ratings of competence.

These differences were small but significant. School size was not significantly related to

ratings of social skills, and percentage of minority students and percentage of students

eligible for FRL in the school were not significantly related to any of the four outcome

variables.

Research Question 3: Predictors of Teacher-Specific Differences in Ratings

Although Hispanic American students and males were rated as significantly less

internalizing than their peers, none of the teacher variables (age, years experience, self-

efficacy, race/ethnicity, or gender) significantly predicted the size of the teacher-specific

differences for these groups. Considering that approximately 35% of variance in

teacher-rated internalizing problems lies at the teacher/classroom level and

approximately 9% at the school level, additional research is needed to explore

additional potential predictors of teacher-specific mean differences in internalizing

behavior ratings. Some teachers may not be aware of some of the less-noticeable









Table 3-24. Hierarchical linear model with competence as outcome: Coefficients for
cross-level interaction of teacher-level variables and differences between
African American and Caucasian students
Fixed Effect Coefficient Standard T-Ratio Approx df p-Value
Error
Teacher Age, B21ab -0.096 0.106 -0.908 57 0.368
Years Experience, B22ab 0.035 0.159 0.222 57 0.825
Management, B23ab -1.021 2.285 -0.447 57 0.656
External, B24ab -1.677 1.706 -0.982 57 0.330
Personal, B25ab 0.142 1.898 0.075 57 0.941
Teacher Gender, B26ab 1.630 2.269 0.718 57 0.475
Teacher Race/Ethnicity, -0.079 2.473 -0.032 57 0.975
B27ab
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.

Table 3-25. Hierarchical linear model with competence as outcome: Coefficients for
cross-level interaction of teacher-level variables and differences between
Hispanic American and Caucasian students
Fixed Effect Coefficient Standard T-Ratio Approx df p-Value
Error
Teacher Age, B31ab -0.156 0.127 -1.225 57 0.226
Years Experience, B32ab 0.178 0.219 0.811 57 0.421
Management, B33ab -0.568 3.175 -0.179 57 0.859
External, B34ab -2.420 2.080 -1.164 57 0.250
Personal, B35ab 1.697 2.967 0.572 57 0.569
Teacher Gender, B36ab -0.629 2.459 -0.256 57 0.799
Teacher Race/Ethnicity, -6.596 3.635 -1.814 57 0.074
B37ab
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.









LIST OF REFERENCES


Abidin, R. R. & Robinson, L. L. (2002). Stress, biases, or professionalism: What drives
teachers' referral judgments of students with challenging behaviors? Journal of
Emotional and Behavioral Disorders, 10, 204-212.

Achenbach, T. M. (1991). Manual for the Teacher's Report Form and 1991 Profile.
Burlington: University of Vermont, Department of Psychiatry.

Ainsworth-Darnell, J. W. & Downey, D. B. (1998). Assessing the oppositional culture
explanation for racial/ethnic differences in school performance. American
Sociological Review, 63, 536-553.

Algozzine, B., Christenson, S., & Yesseldyke, J. E. (1982). Probabilities associated with
the referral to placement process. Teacher Education and Special Education, 5,
19-23.

Artiles, A. J. & Trent, S. C. (1994). Overrepresentation of minority students in special
education: A continuing debate. The Journal of Special Education, 27, 410-437.

Ashton, P. T. & Webb, R. B. (1986). Making a difference: Teachers'sense of efficacy
and student achievement. White Plains, NY: Longman, Inc.

Bahr, M. W. & Fuchs, D. (1991). Are teachers' perceptions of difficult-to-teach students
racially biased? School Psychology Review, 20, 599-610.

Bandura, A. (1977). Self-Efficacy: Toward a unifying theory of behavioral change.
Psychology Review, 84, 191-215.

Barbarin, O., Bryant, D., McCandies, T., Burchinal, M., Early, D., Clifford, R., Pianta, R.,
& Howes, C. (2006). Children enrolled in public pre-k: The relation of family life,
neighborhood quality, and socioeconomic resources to early competence.
American Journal of Orthopsychiatry, 76, 265-276.

Blackorby, J., & Wagner, M. (1996). Longitudinal post-school outcomes of youth with
disabilities: Findings from the National Longitudinal Transition Study. Exceptional
Children, 62, 399-413.

Bracken, B. A. & Keith, L. K. (2004). ClinicalAssessment of Behavior. Lutz, FL:
Psychological Assessment Resources.

Bradley, R., Doolittle, J., & Bartolotta, R. (2008). Building on the data and adding to the
discussion: The experiences and outcomes of students with emotional
disturbance. Journal of Behavioral Education, 17, 4-23.

Bradley, R., Henderson, K., & Monfore, D. A. (2004). A national perspective on children
with emotional disorders. Behavioral Disorders, 29, 211-223.









In an unpublished dissertation, Rayfield (1997) examined teachers' ratings of

elementary school students' externalizing behaviors via the Sutter-Eyeberg Student

Behavior Inventory (SESBI, Sutter & Eyeberg, 1984). The teacher sample consisted of

52 teachers: 44 were Caucasian, 8 were African American, 49 were female, and 3 were

male. Each teacher was asked to complete the SESBI for 2 Caucasian females, 2

Caucasian males, 2 African American females, and 2 African American males, resulting

in data for 415 students (206 Caucasian students, 205 African American students, 2

Hispanic students, and 2 multi-racial students). The results indicated that teachers rated

boys of their own race/ethnicity as exhibiting fewer behavior problems. In addition,

Caucasian female students were rated as having less behavior problems when rated by

Caucasian teachers. African American female students were rated similarly by both

Caucasian and African American teachers. Rayfield (1997) also examined observational

data and found that African American students were observed to exhibit more behavior

problems in the classroom than their Caucasian peers.

Using data from the Early Childhood Longitudinal Study Kindergarten class of

1998 1999, Downey and Pribesh (2004) examined teachers' perceptions of

externalizing problems and approaches to learning (a prosocial variable) for 2,707

African American and 10,282 Caucasian students who had either an African American

or Caucasian teacher in Kindergarten. They found that African American students were

rated as having more externalizing problems and fewer approaches to learning skills

than their Caucasian peers. These differences held constant when teachers'

race/ethnicity was added into the models. The authors examined teachers' ratings when

the students were in 8th grade, and African American students were rated as having









that poverty is a weak predictor of disproportionate representation in the EBD category

(Skiba, Poloni-Staudinger, Simmons, Feggins-Azziz, & Chung, 2005) and the

discussion of poverty contributing to disproportionate representation is beyond the

scope of this study. However, it has also been shown that poverty magnifies already

existing racial differences in the identification of EBD (Skiba et al., 2005) and so the

effects of socio-economic status of families and children cannot be ignored.

Teacher Bias Hypotheses

In order to examine teacher bias as a hypothesis contributing to

overrepresentation of some racial/ethnic minority students identified with EBD, we need

to understand the identification process and teachers' role in that process. In many

school systems the identification of EBD begins primarily with teacher referral and it is

typically the students who annoy or bother their teachers the most who get referred

(Kauffman, 2001). Kauffman (2001) notes that when we examine individuals from a

different cultural background, our own cultural background may lead to biases in our

perception of their behavior. Further, he states that almost all standards and

expectations regarding behavior, and thus, judgments about them, are bound by

culture. Therefore, when we ask Caucasian teachers to rate racial/ethnic minority

students' behaviors, their own culture may influence their ratings of those students.

However, it is important to note that if teachers do engage in cultural serotype bias (e.g.,

biases against students from a different racial/ethnic background) it is most likely of a

subtle nature as opposed to blatant discrimination (Donovan & Cross, 2002).

In Irvine's (1990) explanation of why some African American students are failing

in school, he asserts that a cultural misunderstanding between teachers and students

leads to conflict, distrust, hostility, and potential school failure. In addition, he claims









variables of the CAB-T-T (INT, EXT, SOC, and COM). In the models, Level 1 represents

child i in class j, Level 2 represents class j, and Level 3 represents school k. The

following questions were answered.

1. What proportion of variance in teacher's ratings of students' behaviors is
associated with the teacher/classroom and the school?

2. What are the mean differences in teacher's ratings of students' behaviors across
students' gender, race/ethnicity, and family income?

3. What are the interaction effects of teacher's gender, race/ethnicity, age, years of
experience, and perceived self-efficacy on teacher-specific differences for
race/ethnicity and gender in teacher-rated behaviors?

Structural Equation Modeling

Factor analysis techniques were employed in order to determine whether

construct equivalence exists on the CAB-T for students of various racial/ethnic groups.

The four-factor model displayed in figure 3-1 was compared across Caucasian and

African American samples. Data for Hispanic students were insufficient to perform a

confirmatory factor analysis at the item level. The latent factor structure of the 70 CAB-T

item scores was examined via CFA using the Mplus version 5.1 statistical software

program (Muthen & Muthen, 1998-2004). The estimation method was diagonally

weighted least squares with robust estimation of standard errors. This method uses

data from students who have scores on all items and from students who have

incomplete data and is appropriate for ordinal item scores. Mean and variance adjusted

goodness of fit chi-square statistics were calculated.

I used the chi-square and degrees of freedom goodness-of-fit test to determine

goodness-of-fit for the two models. In addition, I used the following goodness-of-fit

indices to compare models: Bentler's comparative fit index (CFI), the Tucker-Lewis









CHAPTER 2
METHODS

Data for this study were compiled from a large data set collected by the Prevention

Research Team (PRT). The PRT examined the effectiveness of a cognitive-behavioral

intervention, Tools for Getting Along (see Daunic, Smith, Brank, & Penfield, 2006), and

collected pre- and post-test data on students and teachers from 2006 2009 in North

Central Florida. Pre-test data were used in the current study to avoid any effects that

TFGA may have had on the students' behavior or teachers' ratings.

Participants

The PRT recruited schools from a pool of approximately 70 elementary schools in

North Central Florida (suburban and rural districts) that contained a high percentage of

students who were eligible for free or reduced-price lunch (from 60-95%). Overall, 18

schools and 140 4th and 5th grade classrooms participated in the study. Data on all

students in participating classrooms were collected, although only data for students

whose parents returned a signed informed consent form were analyzed. The rate for

returned informed consent forms was approximately 70% of participating students. Each

teacher completed pre-test measures on their students over a two to three week period

prior to the beginning of the intervention. Teachers knew their students for

approximately 3 months prior to completing the pre-test measures about their students.

Teachers completed their self-report measures at the same time. Measures were

delivered to teachers in random order to avoid order effects of teachers completing

multiple surveys in the same order. For one variable of interest in this study, teacher

efficacy, a smaller set of data are available. The Teacher Efficacy in Classroom

Management and Discipline Scale (TECMD; Emmer & Hickman, 1991) was added









behaviors and how much may be attributed to other variables, such as teachers and

schools.

Using three-level random ANOVA models, I examined the proportion of variance

in teacher-rated behaviors at the student-, teacher/classroom-, and school-levels for

four outcome variables. I found that 56.71% of variance in internalizing behavior ratings,

70.73% of variance in externalizing behavior ratings, 73.53% of variance in social skills

ratings, and 79.27% of variance in competence ratings was attributable to the student-

level. Additionally, I found that 34.56% of variance in internalizing behavior ratings,

22.63% of variance in externalizing behavior ratings, 22.24% of variance in social skills

ratings, and 19.97% of variance in competence ratings was attributable to the

teacher/classroom level. These results are consistent with Mashburn et al. (2006)'s

findings that 15% to 30% of the variance in teacher rated behavior is attributable to

differences in teachers or classrooms.

The differences between teachers or classrooms could be a variety of variables,

five of which were examined in this study. Regardless, that 20% to 35% of variance lies

at the teacher/classroom level is highly important to consider when decisions are made

about students based on behavior rating scales. For example, some teachers may not

be as perceptive as others on identifying behaviors associated with internalizing

disorders (e.g., depression, anxiety). Conversely, some teachers may tend to rate a

student globally negatively if they do not like that particular student. Some behavior

rating scales (e.g., the BASC-2) have attempted to resolve this issue by including a

validity scale in their results. These findings support the best practices model to obtain









result of actual differences or teacher bias (Epstein et al., 1996). Conversely, Reid et al.

(2001) examined factors of the IOWA Conners Rating Scale across African American

and Caucasian students. Their results revealed that the same 2-factor structure was

appropriate for both African American and Caucasian students. Similar to findings by

Epstein et al. (1996), Reid et al. (2001) found that African American students received

significantly higher ratings than Caucasian students and they found a significant teacher

race/ethnicity by student race/ethnicity interaction. Finally, Walthall, Konold, and Pianta

(2005) identified the factor structure of the social skills rating system across gender and

race/ethnicity. They found factorial invariance between Caucasian and non-Caucasian

groups and near factorial invariance between males and females.

Purpose

The purpose of the current study was to build upon the current literature by

evaluating student, teacher, and school variables in order to determine whether there

are differences in teacher-rated behaviors across racial/ethnic groups, and whether

those differences are related to a variety of teacher and school level variables. The use

of HLM techniques were employed in order to partition the variances that may be

associated with teacher ratings of students' behaviors at the student, teacher, and

school level. Four outcome variables from a standardized, nationally normed,

representative behavior rating scale were examined including students internalizing

behaviors, externalizing behaviors, competence, and social skills among students of

multiple racial/ethnic groups. Specifically, I examined whether mean-group differences

exist in teacher-rated behaviors for internalizing, externalizing, social skills, and school

competence variables across Caucasian, African American, and Hispanic students. I

explored the partition of variance components across the student-, teacher-, and school-









Caucasian teachers may be biased against racial/ethnic minority students, or may not

understand their behavior (e.g., Downey & Pribesh, 2004). Given that the proportion of

minority students has been increasing while the proportion of minority teachers has

been decreasing (Donovan & Cross, 2002; Irvine, 1990), understanding teachers' views

of students' behaviors is imperative.

The focus in this literature review will be narrowed to theories and hypotheses of

disproportionate representation that are most common in the literature and those that

are specifically related to this study. Hypotheses relating to cultural mismatches in the

classroom, Ogbu's (1991) oppositional culture theory, and biases in teachers' referrals

are commonly cited and are discussed briefly in the section that follows. Hypotheses

related to teacher bias, including early research on teacher expectancy and research

using behavior rating scales, is subsequently discussed more thoroughly as the current

study builds upon that literature. The limitations of current research are discussed, and

additional hypotheses of mean-group differences in behavior are proposed (i.e., teacher

self-efficacy, school level variables, factorial equivalence).

Commonly Cited Hypotheses

Cultural mismatch hypothesis

An idea that is commonly represented in the literature is that difficulties

associated with academic failure of some racial/ethnic minority students and

overrepresentation of some racial/ethnic minority students in special education are due

to a cultural mismatch. The cultural mismatch may be between racial/ethnic minority

students and Caucasian teachers (Hilliard, 1992; Ladson-Billings, 2005; Sleeter, 2001;

Valles, 1998) or racial/ethnic minority students and the institution of education (Boykin,

Tyler, & Miller, 2005). According to this hypothesis, cultural differences between









Landrum, T. J., Tankersly, M., & Kauffman, J. M. (2003). What is special about special
education for students with emotional or behavioral disorders? The Journal of
Special Education, 37, 148-156.

Lau, A. S., Garland, A. F., Yeh, M., McCabe, K. M., Wood, P. A., & Hough, R. L. (2004).
Race/ethnicity and inter-informant agreement in assessing adolescent pathology.
Journal of Emotional and Behavioral Disorders, 12, 145-156.

MacMillan, D. L. & Reschly, D. J. (1998). Overrepresentation of minority students: The
case for greater specificity or reconsideration of the variables examined. The
Journal of Special Education, 32, 15-24.

MacMillan, D. L., Gresham, F. M., Lopez, M. F., & Bocian, K. M. (1996). Comparison of
students nominated for prereferral interventions by ethnicity and gender. The
Journal of Special Education, 30, 133-151.

Mashburn, A. J., Hamre, B. K., Downer, J. T., & Pianta, R. C. (2006). Teacher and
classroom characteristics associated with teachers' ratings of prekindergartners'
relationships and behavior. Journal of Psychoeducational Assessment, 24, 367-
380.

Moller-Leimkuhler, A. M. & Yucel, M. (2009). Male depression in females? Journal of
Affective Disorders, 121, 22-29.

Monroe, C. R. & Obidah, J. E. (2004). The influence of cultural synchronization on a
teacher's perceptions of disruption: A case study of an African American middle-
school classroom. Journal of Teacher Education, 55, 256-268.

Moretti, M. M., Catchpole, R. E. H., & Odgers, C. (2005). The dark side of girlhood: Risk
factors and trajectories to aggression and violence. Canadian Child and
Adolescent Psychiatry and Review, 14, 21-25.

Muthen, B. 0. (1991). Multilevel factor analysis of class and student achievement
components. Journal of Educational Measurement, 28, 338-354.

Muthen, B. 0. (1994). Multilevel covariance structure analysis. Sociological Methods
and Research, 22, 376-398.

National Center for Education Statistics (2000). ECLS K Base Year Data Files and
Electronic Codebook. Rockville, MD: Westat.

Nelson, J. R., Benner, G. J., Lane, K., Smith, B. W. (2004). Academic achievement of k-
12 students with emotional and behavioral disorders. Exceptional Children, 71,
59-73.


105









Table 3-4. Hierarchical linear model with internalizing as outcome: coefficients for
student-level independent variables
Fixed Effect Coefficient Standard T-Ratio Approx df p-V
Error


alue


Intercept, P0b 45.068 0.981 45.947 7 0.000
FRL Eligible Slope, 1.448 0.733 1.975 64 0.052
Pla
African American 0.229 0.899 0.255 57 0.800
Slope, P2a
Hispanic American -3.026 1.194 -2.535 57 0.014*
Slope, P3a
Male Slope, P4a -2.121 0.637 -3.327 57 0.002*
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.

Table 3-5. Hierarchical linear model with internalizing as outcome: coefficients for
teacher level independent variables
Fixed Effect Coefficient Standard T- Approx df p-Value
Error Ratio
Teacher Age, BOlab 0.057 0.40 1.412 932 0.158
Teacher Year, B02ab -0.123 0.063 -1.963 932 0.050*
Management, B03ab -3.987 0.845 -4.717 932 0.000*
External, B04ab -2.720 0.627 -4.336 932 0.000*
Personal, B05ab 3.312 0.743 4.460 932 0.000*
Male Teacher, BO6ab 0.442 0.872 0.507 932 0.612
Non-White Teacher, B07ab 0.034 1.049 0.032 932 0.974


parameter


aThis variable has been centered around its group mean. bThe residual
variance for the parameter has been set to zero.









Hypotheses and research related to teacher bias are discussed later, and two remaining

commonly cited hypotheses are discussed below (i.e., Obgu's Oppositional Culture

Theory, Biases in Teacher Referrals).

Ogbu's (1991) oppositional culture theory

A second hypothesis is that African American students may actually behave worse

when placed with Caucasian teachers than with African American teachers. Downey

and Pribesh (2004) linked this concept to Ogbu's (1991) oppositional culture theory.

Ogbu (1991) distinguishes between voluntary minorities, or those minority groups who

immigrated to the United States (U.S.) by choice and for hope of opportunity, and

involuntary minorities or caste like minorities who were brought to the U.S. against their

will. Obgu (1991) identifies American Indians, African Americans, southwestern Mexican

Americans, and Native Hawaiians as involuntary minorities or groups of people who

were immigrated into the U.S. through slavery, conquest, and colonization. American

Indians and Native Hawaiians clearly did not immigrate to the United States, although it

can be argued that they have experienced oppression similar to what has been

experienced by African Americans and Mexican Americans.

When describing why some African American students tend to fail in school,

Ogbu (1991) said that a history of oppression and lack of job opportunity can lead

young African American students to feel hopeless about their future, and thus, to exhibit

less effort than their Caucasian peers. He goes on to explain how African American

students may have animosity against the school system and, in an effort to maintain

their culture, will engage in behaviors thought to be consistent with African Americans

and avoid acting Caucasian. This theory may help explain why African American

students exhibit more problem behaviors in the classroom than Caucasian students, if









sample of Caucasian students' data and a sample of African American students' data.

My findings suggested that models for both groups of students did not fit well according

to three indicators of goodness-of-fit (e.g., the X2, CFI, and RMSEA). Findings for the TLI

indicated adequate fit for both groups. Findings for each indicator of goodness-of-fit

were very similar, suggesting a possibility that we are measuring the same factors

across Caucasian and African American students. However, as my results indicated

poor fit of the data to the model proposed by the CAB-T, I cannot conclude factorial

invariance until strong factors emerge. Exploratory factor analyses are recommended to

identify revisions to the model proposed by the authors.

Limitations

First, this sample was limited to 4th and 5th grade students. The restriction of

range of this sample limits the ability to generalize my findings to younger and older

students. Additionally, all of these data were collected from suburban and rural districts

from a large state in the southeast, further limiting the ability to generalize these findings

to other regions of the United States.

Similar to many other studies, the teachers in my sample were predominately

Caucasian females. Of my 65 teachers, 43 were Caucasian females, 13 were

Caucasian males, 6 were African American females, one was an African American

male, one was an Asian American female, and one was an Asian American male. I did

not have any Hispanic American teachers in my study. Due to the limited number of

racial/ethnic minority teachers, I analyzed the data comparing Caucasian teachers'

ratings to non-Caucasian teachers' ratings. That I only had 9 racial/ethnic minority

teachers in my sample may have limited the power to detect significant findings in this

study. However, as previously mentioned, I ran the same HLM models with a larger









studies examining teachers' perceptions of students' behaviors utilizing more objective

methods for data collection are discussed below.

Teacher ethnicity and behavior rating scales

More recent research has examined how teachers rate students' behaviors with

the use of standardized, nationally-normed behavior rating scales. As previously

mentioned, teachers' perceptions of students' behavior as expressed via behavior rating

scales are important to the topic of disproportionate representation of racial/ethnic

minority students identified for EBD because behavior rating scales are often used

when assessing for eligibility criteria for special education. Sattler and Hoge (2006)

specifically recommend behavior rating and checklist measures, in addition to other

assessment techniques, when assessing for social, emotional, or behavioral difficulties

because they are easy to administer and compare a student's performance to a national

normative sample in an attempt to provide objective data.

Cohen, DuRant, and Cook (1988) examined the relationships between students'

age, sex, and race/ethnicity and teachers' ratings of students externalizing behaviors.

The authors identified 626 general and special education students and asked their

teachers to complete the Conners' Teacher Rating Scale (Conners, 1973) for study

participants. The findings revealed that African American students who had been

identified with a behavioral disorder or as mildly mentally handicapped were rated

significantly worse than Caucasian students in the same category on the conduct

disorders subscale. However, these differences were not found for students identified

with an emotional disturbance or learning disability (Cohen et al., 1988). More recent

research examining students in Grades 5 through 9 found that teachers rated African

American students higher than Caucasian students on factors relating to externalizing











Percentage of Students Eligible for

ESE who Received Services for EBD

S12
S10
4 8
6

+ 4
C Percent
U 2


AA AI/AN W H A/PI

Race/Ethnicity


Figure 1-2. Percentage of students eligible for ESE who received services for EBD









Table 3-13 displays the coefficients for the cross-level interactions between

teacher-level variables and the contrast of Hispanic American and Caucasian students

on the Externalizing scale. None of the cross-level interactions were significant implying

that the relationships between scores on the externalizing variable and teacher

variables are similar for Hispanic American and Caucasian students.

Table 3-14 displays the coefficients for the cross-level interactions of the teacher-

level variables and the significant difference between male and female students. There

is a significant interaction for teacher age. From Table 3.10, the mean externalizing

difference between male and female students is -3.253 (females rated more

externalizing) for teachers who are at the mean for teacher age for their school. For

teachers who are above the mean for teacher age (e.g., older teachers), the difference

is smaller than -3.253. Conversely, for teachers below the mean for teacher age (e.g.,

younger teachers), the difference is larger than -3.253. Stated differently, age

significantly predicted the teacher-specific difference between male and female

students. The size of the mean difference between male and female students on the

externalizing scale observed for specific teachers was not predicted by any additional

teacher-level variables.

Table 3-15 displays the coefficients for the Level 3 variables in the hierarchical

linear model with the Externalizing scale as the outcome variable. Larger schools had

significantly higher ratings for students' externalizing behaviors. The coefficients for

school percentage of minority students and school percentage of students eligible for

free or reduced-priced lunch were not significant.









internalizing. Additionally, teachers with a higher belief that external influences beyond

teachers' control affect student outcomes rated students significantly less internalizing.

Conversely, teachers with a higher sense of personal self-efficacy rated students as

significantly more internalizing. The coefficients for teacher age, gender, and

race/ethnicity were not significant.

Table 3-6 displays the coefficients for the cross-level interactions between

teacher-level variables and the contrast of African American and Caucasian students for

the internalizing variable. None of the cross-level interactions were significant implying

that the relationships between scores on the internalizing variable and teacher variables

are similar for African American and Caucasian students.

Table 3-7 displays the coefficients for the cross-level interactions between

teacher-level variables and the contrast of Hispanic American and Caucasian students

on the internalizing variable. None of the cross-level interactions were significant

implying that the relationships between scores on the internalizing variable and teacher

variables are similar for Hispanic American and Caucasian students. Stated differently,

the size of the mean difference between Hispanic American and Caucasian students on

the internalizing scale observed for specific teachers was not predicted by the teacher-

level variables.

Table 3-8 displays the coefficients for the cross-level interactions between

teacher-level variables and the contrast of male and female students on the internalizing

variable. None of the cross-level interactions were significant implying that the

relationships between scores on the internalizing variable and teacher variables are

similar for male and female students. Stated differently, the size of the mean difference









2005). Parents of students with EBD report that their children have difficulty getting

along with other students and teachers (Wagner & Cameto, 2004). In addition, students

with EBD are more likely than other students with disabilities to be victims of bullying,

bully others, get into fights, and be suspended (Wagner & Cameto, 2004). Furthermore,

students with EBD tend to display poor post-school outcomes, including high rates of

unemployment and arrests and low rates of post-secondary education (Bullis & Cheney,

1999; Hayling, Cook, Gresham, State, & Kern, 2008; Wagner, 1995). Wagner (1995)

noted that, three to five years after exiting high school, 58% of students with EBD had

been arrested at some point. Sacks and Kern (2008) examined the overall quality of life

for students with EBD, including how they felt about themselves, their relationships, and

their environment. They found that students with EBD experience significantly lower

levels of quality of life indicators than their non-disabled peers (Sacks & Kern, 2008). It

is important to note that the label of EBD does not cause these negative outcomes.

However, if a student is identified with EBD and placed into a secluded classroom with

peers who are more likely to experience these negative outcomes, the chance of that

particular student to experience similar outcomes is increased. Given these concerns,

identification of EBD should only occur for students who are truly in need, and

inappropriate identification should be avoided at all costs.

Disproportionate Representation of Racial/Ethnic Minority Students

Disproportionate representation is defined as "the extent to which membership in a

given ethnic group affects the probability of being placed in a specific special education

disability category" (Oswald et al., 1999, p.198). Cullinan and Kauffman (2005)

describe the disparity of some racial/ethnic groups who are identified with EBD. By

comparing the percentage of students identified with EBD with the percentage of









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

DISPROPORTIONALITY OF MINORITY STUDENTS IDENTIFIED WITH AN
EMOTIONAL/BEHAVIORAL DISORDER: EXAMINING TEACHERS' RATINGS OF
STUDENTS' BEHAVIOR AND FACTORIAL EQUIVALENCE OF A BEHAVIOR RATING
SCALE


By

Christina Diane Peters

August 2010

Chair: John H. Kranzler
Major: School Psychology

The disproportionate representation of racial/ethnic minority students identified

with special education needs has been a topic of concern for 40 years. Disproportionate

representation is "the extent to which membership in a given ethnic group affects the

probability of being placed in a specific special education disability category" (Oswald,

Coutinho, Best, & Singh, 1999, p.198). An assumption is that the percentage of a

particular group identified with a special education need equals the percentage of that

group in the student population. This study examined several hypotheses to better

understand disproportionate representation of racial/ethnic minority students identified

with an emotional/behavioral disorder in the school system, including teacher bias.

Previously, researchers have examined teacher expectancy in addition to the

completion of behavior rating scales. The purpose of this study was to build upon

current literature focusing on the completion of behavior rating scales. Sixty-five

teachers completed the Clinical Assessment of Behavior Teacher Form (CAB-T) for a

sample of 982 Caucasian, African American, and Hispanic American students. Four









or who indicated mixed ethnicity, the numbers were not large enough to include them as

a sub-group. Additional research examining ratings of additional racial/ethnic groups is

needed.

Summary and Implications of Findings

This study produced several interesting findings. First and foremost, I found that

a significant amount of variance in teacher-rated behaviors lies at the teacher/classroom

level (20% to 35%) and at the school level (5% to 9%). It is imperative to keep this in

mind when evaluating student data and making decisions about students' educational

placement and provision of services. It is always recommended to gather data from

multiple sources and compare behavior ratings across raters and environments (Sattler

& Hodge, 2006). I found that Hispanic American students were rated significantly less

internalizing than their Caucasian peers and that no teacher variables in this study

predicted the teacher-specific differences regarding internalizing behaviors. Further

research could examine whether Hispanic students really do experience less

internalizing problems or whether teachers do not pick up on symptoms exhibited by

Hispanic students. I found that African American students were rated as significantly

more externalizing than their Caucasian peers and that teacher self-efficacy related to

classroom management and discipline predicted the teacher-specific differences. This is

perhaps the most important finding of this study, as I did not find that teacher

race/ethnicity or gender predicted teacher-specific differences related to externalizing

behaviors as others have found related to mean-group differences (e.g., Mashburn et

al., 2006; Rayfield, 1997; Rong, 2006). Teacher race/ethnicity is a variable that cannot

be controlled, while teacher self-efficacy related to classroom management and

discipline is something that can be enhanced through preservice and inservice training.









were included in order to examine the proportion of variance that lies at the student,

teacher, and school levels. For teacher ratings of students' internalizing behaviors,

56.71% of the variance in ratings was attributed to the student level, 34.56% to the

teacher/classroom level, and 8.73% to the school level. All variance components were

significant at the p < 0.01 level. Regarding teacher ratings of students' externalizing

behaviors, 70.73% of the variance was attributable to the student level, 22.63% to the

teacher/classroom level, and 6.64% to the school level. All variance components were

significant at the p < 0.01 level. For teacher ratings of students' social skills, 73.53% of

the variance was attributable to the student level, 22.24% to the teacher/classroom

level, and 4.23% to the school level. The teacher/classroom level variance component

was significant at the p < 0.01 level and the variance component for the school level

was significant at p = 0.021. Finally, for teacher ratings of students' competence,

79.27% of the variance was attributable to the student level, 19.97% to the

teacher/classroom level, and 0.76% to the school level. The variance component for the

teacher/classroom level was significant at the p < 0.01 level but the variance component

for the school level was not significant (p = 0.280).

Intercepts and Slopes as Outcomes Models

The IASO models were included in order to examine the main effects of the Level

1 variables (i.e., student gender, race/ethnicity, and family income), the main effects of

the Level 2 variables (i.e., teachers' age, years experience, self-efficacy, gender, and

race/ethnicity), and the main effects of the Level 3 variables (i.e., school size,

percentage minority students per school, and average family income per school).

Additionally, interactions between some Level 1 and all Level 2 variables were

examined for significance. Significant interactions can be interpreted as a Level 2









students' and teachers' gender and race/ethnicity on social behavior ratings.

Participants included 984 teachers who completed the BASC on 984 African American

and Caucasian students between 6- and 11-years of age. Using an analysis of

covariance design, controlling for parental education as a measure of socioeconomic

status, Rong (1996) found that both gender and race/ethnicity influenced teachers'

perceptions of students' behaviors and that the magnitude and direction of the effects

were influenced by teachers' race/ethnicity. Specifically, African American teachers'

ratings were not significantly different for African American and Caucasian students on

all four subscales of the BASC (adaptability, social desirability, leadership, and social

skills). However, a significant and large effect of students' ethnicity was found for

Caucasian teachers. Caucasian teachers rated African American students significantly

lower on the social desirability, social skills, and leadership subscales (Rong, 1996).

When controlling for teacher and student gender, the differences in Caucasian teachers'

ratings of female students disappeared, but remained significant for males. More

specifically, Caucasian female teachers rated Caucasian male students as more

favorably than African American male students. Significant differences in behavior

ratings were not evident for females, regardless of teacher and student race/ethnicity

(Rong, 1996). The sample in this study was limited, however, in that no Caucasian male

teachers rated African American students so we cannot tell whether Caucasian male

and female teachers rate differently. Additionally, in this study, each teacher only rated

one student so an examination of how the same teacher would rate multiple students

was not possible.









3-27 Hierarchical linear model with internalizing as outcome: Coefficients for
school-level independent variables ............... .................. ................ 87

3-28 Summary of fit statistics for the confirmatory factor analyses by race/ethnicity... 87









students in the population, they show that some racial/ethnic groups are

disproportionately represented. Using national data on the prevalence of students with

EBD and the number of students in public schools in 2002 and 2003, the authors show

that African American students represent 27% of students with EBD but only make up

17% of public school students, reflecting an overrepresentation of African American

students identified with EBD. The trend is also noticed for Caucasian students (63% of

students with EBD and 61% of public school students; Cullinan & Kaufman, 2005). The

authors also report that some racial/ethnic groups are underrepresented, such as Asian

or Pacific Islander students (1% of students with EBD, 4% of public school students)

and Hispanic students (8% of students with EBD and 16% of public school students).

The authors point out that the disproportionality of African American students is

approximately 160% of what we would expect based on their representation in the

general population (Cullinan & Kaufman, 2005). Likewise, Oswald et al. (1999) found

that African American students were 1.5 times as likely to be identified as EBD than

their Caucasian peers.

Hypotheses associated with Disproportionate Representation

The idea of overrepresentation of minority students in special education was first

expressed by Lloyd Dunn in 1968 when he asserted that a vast majority of students

identified as educable mentally retarded were from some racial/ethnic minority groups

and/or families with a low socioeconomic status; groups who have been previously

considered by some to be of a lower social status (MacMillan & Reschly, 1998). A

widespread assumption is that the proportion of a particular ethnic group identified with

special education needs should equal the proportion of that ethnic group in the school-

aged population, given the assumption that all groups are equal. If the proportion of a









teachers and students may lead to referral and later identification of EBD (Serwatka et

al., 1995) perhaps because Caucasian teachers are biased against racial/ethnic

minority students or lack an understanding of their culture. Ladson-Billings (2005)

proposes that difficulties may begin before the teacher enters the classroom, because

the overwhelmingly majority of Caucasian faculty who teach preservice educators are

likely very far removed from the needs of the urban, minority classroom. Similarly,

Research has shown that Caucasian preservice teachers have minimal cross-cultural

experience and may lack confidence regarding their ability to teach African American

students (see Sleeter, 2001, for a review).

Hilliard (1992) argued that there is a clear difference between the school system,

which is in line with the cultural style of most Caucasian Americans, and the cultural

behavioral style of African American students. Furthermore, he stated that a

misunderstanding of students' cultural behavioral style may lead to inaccurate

estimation of intellectual, academic, and language abilities. Hilliard additionally stated

that different behavioral styles exist for African American students and, if ignored during

lesson planning and instruction delivery, negative outcomes emerge (e.g., academic

underachievement, perception of behaviors as being problematic). Boykin et al. (2005)

found a misalignment between the culture of African American students and families in

America (e.g., movement, communalism) and the mainstream culture in the classroom

(e.g., individualism, competitiveness). The authors examined mostly African American

students in classrooms with African American teachers, so the proposed cultural

misalignment was between the African American culture and the Caucasian American

classroom or school. More specifically, they proposed that the classroom promotes









ACKNOWLEDGMENTS

I would like to acknowledge several people for their help throughout my

educational career. First and foremost, I would like to thank my graduate advisor, Dr.

John Kranzler, for his endless support and facilitation of my professional development.

From my first year in my graduate program, he encouraged me to engage in additional

research interests above and beyond the requirements of the program, and greatly

assisted me from start to finish on multiple projects. Additionally, he always reminded

me to reward myself for hard work by enjoying what life has to offer.

I am also grateful to have such an incredible doctoral committee and to have

worked with several inspiring faculty. I would like to thank Dr. Diana Joyce for helping

me further develop my interest in students with emotional and behavioral difficulties,

and for consistently motivating and reinforcing me. I would like to thank Dr. James

Algina for teaching me how to understand and utilize advanced statistical concepts, and

for the countless hours he has spent helping me clean, analyze, and interpret data. I

would like to thank Dr. Stephen Smith, in addition to Dr. Ann Daunic, for offering me four

years of employment on the most wonderful graduate assistantship a student could ask

for. They helped enhance my research and manuscript writing skills, and supported my

decision to travel to many amazing places. Also, thank you for providing the data for this

study. Additionally, I would like to thank Dr. Nancy Waldron for her guidance and

supervision at P. K. Yonge and beyond.

Finally, I would like to thank my loving and encouraging family. I would not have

been able to achieve this milestone without your endless support. I am especially lucky

to have my mother and grandmother as role-models. To my best friend Sejal, thank you,

as always, for everything you do.




























To my mother, the most amazing woman I have ever met, my loving family, and the
graduate faculty who have supported me throughout the past five years I could not
have accomplished this goal without you.









(1983) found that 11 out of 24 studies that examined teacher expectancies as a function

of students' race/ethnicity (Caucasian vs. African American) revealed teacher

expectancies that favored Caucasian students over African American students.

Clifton, Perry, Parsonson, and Hryniuk (1986) examined teacher expectancies of

students representing six ethnic groups in Canada. Students identified themselves as

British, French, German, Canadian-Indian, Filipino, or Portuguese and their homeroom

teachers completed a questionnaire that measured normative (i.e., social behavior) and

cognitive (i.e., academic performance) expectations for their students. The researchers

controlled for students' socioeconomic status, intellectual ability (measured via Raven's

Progressive Matrices; Raven, 1947), academic performance, and students'

expectations of themselves. They found that the students' race/ethnicity continued to

influence the teachers' expectations for students as teachers indicated different

expectations for the students from different racial/ethnic groups in terms of their

likelihood of completing 12th grade English and Math courses and their level of

cooperation and reliability (Clifton et al., 1986). However, in this study, some of the

students from different racial/ethnic groups displayed actual differences in terms of

academic achievement as well (e.g., German students outperformed British students).

Early researchers found powerful information related to teacher expectancy and

teachers' perceptions of students behaviors. While the early research laid the

foundation for current work, several limitations are apparent in the early studies.

Specifically, much of the research (e.g., Cooper, Baron, & Lowe, 1975; DeMeis &

Turner, 1978; Gottlieb, 1964) relied on teacher interviews and assessing teachers'

perceptions of students based on vignettes rather than actual student data. Research









Index (TLI), and the Root Mean Square Error of Approximation (RMSEA). Criteria for

good fit include CFI and TLI> .95, and RMSEA .06 (Hu & Bentler, 1999).

Table 2-1. Dummy coded variables for student race/ethnicity
Group Z1 Z2
Caucasian 0 0
African American 1 0
Hispanic American 0 1









based on students' family income and socioeconomic status (e.g., Dusek & Joseph,

1983).

Teacher-Level Data

Gender

Information on teachers' gender was obtained from each teacher's perspective

school district. Gender was represented as a dummy coded variable with females

assigned a value of 0 and males assigned a value of 1. Teacher gender is included as a

variable of interest in order to examine interaction effects of teacher gender and

classroom-level mean differences in behavior ratings due to the student-level variables

(i.e., teacher-specific differences).

Race/ethnicity

Information on teachers' race/ethnicity was obtained from each teacher's

perspective school district. Ethnicity was represented as a dummy coded variable with

Caucasian teachers coded 0 and non-Caucasian teachers coded 1. Teacher

race/ethnicity is included as a variable of interest in order to examine interaction effects

between teacher race/ethnicity and teacher-specific differences in behavior ratings.

Age

Information on teachers' age was obtained from each teacher's perspective school

district. Teacher age is included as a variable of interest in order to examine interaction

effects between teacher's age and teacher-specific differences in behavior ratings.

Years of Experience

Information on teachers' years of teaching experience was obtained from each

teacher's perspective school district. Teachers' years of experience is included as a









vignette for the African American lower-class student in terms of report card grades than

the vignette for Caucasian, middle-class students, which the authors' claimed supported

their hypothesis that expectations about a student's academic achievement are

influenced by race/ethnicity and socioeconomic status (Cooper et al., 1975). However,

it is difficult to determine whether the different expectations were due to bias or the fact

that actual academic mean group differences exist between students of different

backgrounds because limited variables were studied and the models proposed were

vignettes as opposed to real student data.

DeMeis and Turner (1978) evaluated the effects of students' ethnicity, dialect,

and attractiveness on teachers' evaluations by asking 68 Caucasian elementary school

teachers to listen to a cassette tape recording of 12 different males' responses to a

standardized question: "What happened on your favorite TV show the last time you

watched it? (p. 79)". The taped responses were paired with pictures that varied in

ethnicity and attractiveness and the teachers were asked to rate the speaker's

personalities, quality of responses, current academic abilities, and future academic

abilities. The results indicated that students' race/ethnicity, dialect, and level of

attractiveness all influenced the teachers' expectations for the students and the

Caucasian students were rated higher on all 4 outcome variables than African American

students (DeMeis & Turner, 1978).

In a meta-analysis of 77 studies on teacher expectancies, Duesk and Joseph

(1983) determined that student attractiveness, conduct, cumulative folder information,

ethnicity, and socioeconomic status were significantly related to teachers' expectations

about students' performance. Specifically regarding race/ethnicity, Duesk and Joseph









test bias (i.e., the factor structure of instruments) than the teachers who complete the

instruments.

Additional Predictors of Mean-Group Differences in Behavior Ratings

Teacher Self-Efficacy

A potential variable that may predict whether teachers are likely to rate students

differently based on their race/ethnicity is associated with the idea of teacher self-

efficacy, or a belief that teachers can influence a student's learning and behavior

despite the presence of risk-factors. Theories of teacher efficacy have been based on

Bandura's (1977) theory of self-efficacy. According to Ashton and Webb (1986), teacher

self-efficacy is an important concept to facilitate understanding of teachers' self-

perceived role, attitudes towards their job, and interactions with students. Two

dimensions of teacher efficacy have been identified: sense of teaching efficacy and

sense of personal teaching efficacy (Ashton & Webb, 1986; Dembo & Gibson, 1985;

Gibson & Dembo, 1984). Teachers' sense of teaching efficacy refers to general

expectations that teaching can influence student learning. For example, teachers with

low sense of efficacy may believe that some students are unable to learn and that no

teacher will be able to help them. Teachers' sense of personal teaching efficacy refers

to a teacher's evaluation of their own teaching competence. Teachers' perceptions of

their own teaching ability have been proposed to influence choices related to classroom

management and instructional strategies (Ashton & Webb, 1986). Ideas related to

teachers' sense of efficacy can be related to biases in referral decisions and ratings of

students' behaviors. For example, a teacher with high self-efficacy may be less likely to

refer a student for special education services if they believe they can work effectively

with a student exhibiting learning or behavior problems. Conversely, a teacher with low









U.S. Department of Education. (2005). Twenty-seventh annual report to Congress on
implementation of the Individuals with Disabilities Education Act. Washington,
DC: Author.

U.S. Department of Health and Human Services (1999). Mental Health: A Report of the Surgeon
General. Rockville, MD: U.S. Department of Health and Human Services, Substance
Abuse and Mental Health Services Administration, Center for Mental Health Services,
National Institutes of Health, National Institute of Mental Health.

Valles, E. C. (1998). The disproportionate representation of minority students in special
education: Responding to the problem. The Journal of Special Education, 32, 52-54.

Wagner, M. M. (1995). Outcomes for youths with serious emotional disturbance in
secondary school and early adulthood. The Futures of Children, 5, 90-112.

Wagner, M., Kutash, K., Duchnowski, A. J., Epstein, M. H., & Sumi, W. C. (2005). The
children and youth we serve: A national picture of the characteristics of students
with emotional disturbances receiving special education. Journal of Emotional
and Behavioral Disorders, 13, 79-96.

Wagner, M. & Cameto, R. (2004). The characteristics, experiences, and outcomes of
youth with emotional disturbances. NLTS2 Data Brief, 3(2).


Walthall, J. C., Konold, T. R., & Pianta, R. C. (2005). Factor structure of the social skills
rating system across child gender and ethnicity. Journal of Psychological
Assessment, 23, 201-215.


108









LIST OF TABLES


Table page

2-1 Dummy coded variables for student race/ethnicity ................. ........ ............ 64

3-1 Descriptive statistics for the clinical assessment of behavior subscales .............76

3-2 Descriptive statistics for teacher efficacy in classroom management and
discipline scale ............ .......... .... ......... ...................... 76

3-3 Percentage of variance at student, teacher/classroom, and school levels..........76

3-4 Hierarchical linear model with internalizing as outcome: coefficients for
student-level independent variables ............. ............... ........................... 77

3-5 Hierarchical linear model with internalizing as outcome: coefficients for
teacher level independent variables ...... .................... ................ 77

3-6 Hierarchical linear model with internalizing as outcome: coefficients for cross-
level interaction of teacher-level variables and differences between African
American and Caucasian students...... ..................................... 78

3-7 Hierarchical linear model with internalizing as outcome: Coefficients for
cross-level interaction of teacher-level variables and differences between
Hispanic American and Caucasian students....................... ................ 78

3-8 Hierarchical linear model with internalizing as outcome: Coefficients for
cross-level interaction of teacher-level variables and differences between
m ale and fem ale students ............................................................ ............... 78

3-9 Hierarchical linear model with internalizing as outcome: Coefficients for
school-level independent variables ...... .................... ................ 79

3-10 Hierarchical linear model with externalizing as outcome: Coefficients for
student-level independent variables ............. ............... ........................... 79

3-11 Hierarchical linear model with externalizing as outcome: Coefficients for
teacher level independent variables .... .. .................... ................ 80

3-12 Hierarchical linear model with externalizing as outcome: Coefficients for
cross-level interaction of teacher-level variables and differences between
African Am erican and Caucasian students ..................................................... .. 80

3-13 Hierarchical linear model with externalizing as outcome: Coefficients for
cross-level interaction of teacher-level variables and differences between
Hispanic American and Caucasian students....................... ................ 81










Gerber, M. M. & Semmel, M. I. (1984). Teacher as imperfect test: Reconceptualizing the
referral process. Educational Psychologist, 19, 137-148.

Gibson, S., & Dembo, M. (1984). Teacher efficacy: A construct validation. Journal of
Educational Psychology, 76, 569-582.

Gottlieb, D. (1964). Teaching and students: The views of negro and Caucasian
teachers. Sociology of Education, 37, 345-353.

Gresham, F. M. & Elliot, S. N. (1990) Social Skills Rating System. Circle Pines: MN:
American Guidance Service.

Gresham, F. M., Reschly, D. L., & Carey, M. P. (1987). Teachers as "tests":
Classification accuracy and concurrent validation in the identification of learning
disabled children. School Psychology Review, 1987, 543-553.

Hayling, C. C., Cook, C., Gresham, F. M., State, T., & Kern, L. (2008). An analysis of
the status and stability of the behaviors of students with emotional and behavioral
difficulties. Journal of Behavioral Education, 17, 24-42.

Hightower, A. D., Work, W. C., Cowen, E. L., Lotyczewski, B.S., Spinell, A. P., Guare, J.
C., & Rohrbeck, C. A. (1986). The teacher-child rating scale: A brief elementary
measure of elementary children's school problem behaviors and competencies.
School Psychology Review, 15, 393-409.

Hilliard, A. G. (1992). Behavioral style, culture, and teaching and learning. The Journal
of Negro Education, 61, 370-377.

Jensen, A. R. (1980). Bias in mental testing. New York: The Free Press.

Kauffman, J. M. (2001). Characteristics of emotional and behavioral disorders of
children and youth (7th ed.). Upper Saddle River, NJ: Prentice-Hall, Inc.

Koff, C. W., Bradshaw, C. P., & Leaf, P. J. (2008). A multilevel studt of predictors of
student perceptions of school climate: The effect of classroom-level factors.
Journal of Educational Psychology, 100, 96-104.

Kranzler, J. H., Miller, M. D., & Jordan, L. (1999). An examination of racial/ethnic and
gender bias on curriculum-based measurement of reading. School Psychology
Quarterly, 14, 327-342.

Ladson-Billings, G. J. (2005). Is the team all right? Diversity and teacher education.
Journal of Teacher Education, 56, 229-234.


104









self-efficacy may be more likely to refer the same student for special education if he or

she believes the student's needs are beyond what the teacher can provide. This may be

the case especially in situations where the student comes from a different ethnicity than

the teacher and the teacher does not understand the student's behaviors from a cultural

perspective.

Podell and Soodak (1993) asked teachers to read a case study about a student

experiencing academic difficulties and to indicate the appropriateness of the student's

class and whether they would refer the student for special education. Six case studies

were designed to vary in the student's socioeconomic status (SES) and learning

problems. The findings revealed that teachers with low self-efficacy indicated they

would refer students from a lower-SES background and teachers with high self-efficacy

did not differentiate referrals by SES (Podell & Soodak, 1993). Similarly, Soodak and

Podell (1993) examined teacher efficacy and student problem type on teachers' referral

and placement decisions. The authors also used case studies describing a student with

learning or behavioral difficulties. The results indicated that both special and regular

educators were comfortable with the general education setting for students with

difficulties when they had high teaching and personal efficacy. However, those regular

educators with lower personal efficacy were more likely to say they would refer students

to special education (Soodak & Podell, 1993).

Frey (2002) found that teachers with higher self-efficacy with regards to classroom

management and discipline recommended less restrictive special education placements

for students with EBD than teachers with lower self-efficacy. Additionally, teachers with

high self-efficacy have been shown to be less likely to refer difficult students to special









demographic factors (e.g., race/ethnicity, gender). Further, they showed that the CAB-T

produces minimal mean score differences as a result of demographic variables. The

authors examined the reliabilities of the cluster scores across different ethnic groups

and found similar reliabilities across groups on all clusters and scales and found few

differences (e.g., externalizing scale: a = .96-.97 for Caucasian students, a = .96 .97

for African American students, and a = .93 .98 for Hispanic students). The CAB-T was

selected for use in this study based on the assertion that the scores are highly reliable

and valid for students across racial/ethnic groups.

Teacher Efficacy in Classroom Management and Discipline Scale (Emmer &
Hickman, 1991)

The Teacher Efficacy in Classroom Management and Discipline Scale (TECMD)

was developed by Emmer and Hickman in 1991. They modified a teacher efficacy scale

created by Gibson and Dembo (1984) that assessed two factors: Personal Teaching

Efficacy and Teaching Efficacy. Emmer and Hickman developed additional items to

assess teacher efficacy regarding classroom management and discipline focusing on

skills and capabilities. Additionally, they created items to further assess teachers' belief

in the strength of external factors (e.g., home life, peer influences) on students'

behaviors as opposed to teacher influences. They retained items from the Gibson and

Dembo scale that assessed personal teaching efficacy. The TECMD therefore contains

3 subscales. The classroom management and discipline factor assesses teachers' self

perception of their abilities to manage and discipline their students (e.g., "I am confident

of my ability to begin the year so that students will learn to behave well"). The external

influences factor assesses teachers' belief regarding the strength of external influences

that are beyond the teacher's control (e.g., If students aren't disciplined at home, then









Table 3-9. Hierarchical linear model with internalizing as outcome: Coefficients for
school-level independent variables
Fixed Effect Coefficient Standard T-Ratio Approx df P-Va
Error
School Size, G001a 0.014 0.006 2.529 7 0.0:


lue

39*


Percentage Minority
Students in School, G002a


0.002


0.048 0.044


0.967

0.319


Percentage Students 0.165 0.153 1.074
Eligible for FRL, G003a
aThis variable has been centered around its group mean.


Table 3-10. Hierarchical linear model with externalizing as outcome: Coefficients for
student-level independent variables
Fixed Effect Coefficient Standard T-Ratio Approx df p-Value
Error
Intercept, P0b 46.032 0.716 64.276 7 0.000*
FRL Eligible Slope, 1.326 0.684 1.938 64 0.057
Pla
African American 2.996 0.843 3.554 57 0.001*
Slope, P2a
Hispanic American -2.132 1.120 -1.904 57 0.062
Slope, P3a
Male Slope, P4a -3.253 0.599 -5.430 57 0.000*
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.










Table 3-18. Hierarchical linear model with social skills as outcome: Coefficients for
cross-level interaction of teacher-level variables and differences between
African American and Caucasian students
Fixed Effect Coefficient Standard T-Ratio Approx df p-Value
Error
Teacher Aae. B2ab -0.146 0.103 -1.424 57 0.160


Years Experience, B22ab
Management, B23ab
External, B24ab
Personal, B25ab


0.059
4.610
1.884
-3.476


0.155
2.215
1.661
1.840


0.382
2.081
1.135
1.889


0.704
0.042*
0.262
0.064


Teacher Gender, B26ab 2.189 2.199 0.995 57 0.324
Teacher Race/Ethnicity, -1.803 2.405 -0.750 57 0.456
B27ab
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.

Table 3-19. Hierarchical linear model with social skills as outcome: Coefficients for
cross-level interaction of teacher-level variables and differences between
Hispanic American and Caucasian students
Fixed Effect Coefficient Standard T-Ratio Approx df p-Value
Error
Teacher Age, B31ab -0.117 0.120 -0.976 57 0.333
Years Experience, B32ab -0.152 0.206 -0.737 57 0.464
Management, B33ab -0.272 2.995 -0.091 57 0.928
External, B34ab -1.930 1.964 -0.983 57 0.330
Personal, B35ab 0.911 2.809 -0.324 57 0.747
Teacher Gender, B36ab -0.663 2.320 -0.286 57 0.776
Teacher Race/Ethnicity, -4.869 3.403 -1.431 57 0.158
B37ab
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.









al., 1995; Sleeter, 2001; Valles, 1998) are too broad in that they claim that difficulties

experienced by African American students are due to having teachers of a different

ethnicity. As Ladson-Billings (2005) pointed out, we would then expect to see little

difficulties experienced by African American students with African American teachers,

and this is not the case. Ogbu's oppositional culture theory (Obgu, 1991) attempts

explain some difficulties experienced by African American students by stating that

African American students essentially rebel against the Caucasian mainstream

classroom. However, his theory has not been empirically supported (Ainsworth-Darnell

& Downey, 1998; Downey & Pribesh, 2004). Finally, researchers have examined

potential bias in teachers' referrals for special education (Bahr & Fuchs, 1991;

MacMillan, Gresham, Lopez, & Bocian, 1996). While this may be a good starting point

for understanding overrepresentation of racial/ethnic minority students in the EBD

category, there are several steps between referral and identification (e.g., the evaluation

process) that may play a role as well.

More narrow hypotheses and more recently proposed hypotheses that may be

better suited for understanding disproportionate representation of some racial/ethnic

minority students are discussed below. These hypotheses include the possibility of

teacher bias examined via teacher expectancy and behavior rating scales, the role of

teacher self-efficacy, school-level variables, and psychometric properties of

assessments across cultural groups.

It is important to note that many researchers have also examined the role of

poverty, or the interaction of race/ethnicity and poverty, in disproportionate

representation of minority students in special education. However, it has been shown









33% of variance in teacher-rated behaviors are due to differences between teachers,

not students, indicates a need for further examination of variance at the

teacher/classroom level. An evaluation of teachers' ratings of multiple students and use

of HLM techniques would allow for a partition of variances across the student and

teacher level and the ability to include predictors of teacher-specific differences in

behavior ratings.

Additional limitations of the previously discussed studies include lack of enough

male and minority teachers to conduct complete analyses (Rayfield, 1997; Rong, 1996)

and limited dependent variables (e.g., only measuring externalizing behaviors or social

competency variables). In addition, a large majority of the studies focus exclusively on

Caucasian and African American students. As other minority groups may be

disproportionately represented in the category of EBD, it is important to study teachers'

perceptions of students from additional minority groups as well.

Finally, typical of research in this area, the research questions were limited to

whether teachers' race/ethnicity and/or gender may contribute to mean-group

differences in behavior ratings (e.g., Abidin & Robinson, 2002; Cullinan & Kauffman,

2005; Mashburn et al., 2005; Pigott & Cowen, 2000; Rayfield, 1997; Rong, 1996).

Several additional variables are of interest as potential predictors of mean-group and/or

teacher-specific differences in behavior ratings. Variables at the teacher-level include

teacher self-efficacy and years of experience teaching. Variables at the school level

include school size, percentage of minority students, and the percentage of students on

free or reduced price lunch (e.g, socioeconomic status of the school). Finally, mean-

group differences in behavior ratings may be more related to the internal indicators of









Sacks, G. & Kern, L. (2008). A comparison of quality of life variables for students with
emotional and behavioral disorders and students without disabilities. Journal of
Behavioral Education, 17, 111-127.

Sattler, J. M. & Hoge, R. D. (2006). Assessment of children: Behavioral, social, and
clinical foundations (5th ed.). La Mesa, CA: Jerome Sattler, Publisher, Inc.

Sbarra, D. A. & Pianta, R. C. (2001). Teacher ratings of behavior among African
American and Caucasian children during the first two years of school.
Psychology in the Schools, 38, 229-238.

Schalock, R. L., Holl, C., Elliott, B., & Ross, I. (1992). A longitudinal follow-up of
graduates from a rural special education program. Learning Disability Quarterly,
15, 29-38.

Serwatka, T. S., Deering, S., & Grant, P. (1995). Disproportionate representation of
African Americans in emotionally handicapped classes. Journal of Black Studies,
25, 492-506.

Shinn, M. R., Tindal, G. A., & Spira, D. A. (1987). Special education referrals as an
index of teacher tolerance: Are teachers imperfect tests? Exceptional Children,
54, 32-40.

Skiba, R. J., Poloni-Staudinger, L., Simmons, A. B., Feggins-Azziz, L. R., & Chung, C.
(2005). Unproven links: Can poverty explain ethnic disproportionality in special
education? The Journal of Special Education, 39, 130-144.

Sleeter, C. E. (2001). Preparing teachers for culturally diverse schools: Research and
the overwhelming presence of Caucasianness. Journal of Teacher Education,
52, 94-106.

Soodak, L. C. & Podell, D. M. (1993). Teacher efficacy and student problem as factors
in special education referral. The Journal of Special Education, 27, 66-81.

Sutter, J. & Eyeberg, S. M. (1984). Sutter-Eyeberg Student Behavior Inventory.
(Available from Shelia Eyeberg, Department of Clinical and Health Psychology,
University of Florida, Gainesville, FL, 32610.

Tschannen-Moran, M., Hoy, A. W., & Hoy, W. K. (1998). Teacher efficacy: Its meaning
and measure. Review of Educational Research, 68, 202-248.

U. S. Department of Education (2006). Federal register: Assistance to states for the
education of children with disabilities and preschool grants for children with
disabilities; final rule. Washington, DC: Author.









mean for their school. For teachers with management self-efficacy above the mean for

their school, the difference is smaller than -2.886 points. For teachers with management

self-efficacy below the mean for their school, the difference is larger than -2.886 points.

Stated differently, teacher self-efficacy for behavior management significantly predicted

the teacher-specific difference between African American and Caucasian students. The

size of the mean difference between African American and Caucasian students on the

social skills scale observed for specific teachers was not predicted by any additional

teacher-level variables.

Table 3-19 displays the coefficients for the cross-level interactions between

teacher-level variables and the contrast of Hispanic American and Caucasian students

on the Social Skills scale. None of the cross-level interactions were significant implying

that the relationships between scores on the social skills variable and teacher variables

are similar for Hispanic American and Caucasian students.

Table 3-20 displays the coefficients for the cross-level interactions between

teacher-level variables and the contrast of male and female students on the social skills

variable. None of the cross-level interactions were significant implying that the

relationships between scores on the social skills variable and teacher variables are

similar for male and female students.

Table 3-21 displays the coefficients for the Level 3 variables in the hierarchical

linear model with the Social Skills scale as the outcome variable. The coefficients for

school size, school percentage of minority students, and school percentage of students

eligible for free or reduced-priced lunch were not significant.









I also found that females were rated as significantly more externalizing than

males and that teacher age predicts the teacher-specific differences related to

externalizing behaviors across gender. Research focusing on increasing rates of

externalizing behavior among females is needed before suggestions can be made

regarding inservice training for teachers regarding these concerns. Additionally, I found

that African American students received significantly lower social skills ratings than their

Caucasian peers and that teacher self-efficacy in classroom management and discipline

predicted the teacher-specific differences. This finding is similar to the finding that

teacher self-efficacy predicted the teacher-specific differences for African American and

Caucasian students on externalizing behavior. I found several mean group differences

for school competence: students eligible for FRL, African American students, and

females received significantly lower competence ratings. No teacher variables included

in this study significantly predicted any teacher-specific differences related to these

variables. This finding may indicate that there are real group differences in school

competence. However, as 19% of the variance in competence ratings was attributable

to the teacher/classroom level, additional research examining possible predictors is

warranted. Finally, I found that goodness-of-fit indices for Caucasian and African

American students rated via the CAB-T were similar, suggesting that we may be

measuring the same factors across both groups. However, the results from this factor

analyses indicated a poor fit of the model to the data, indicating that additional research

is needed on psychometric properties of the CAB-T.


100









reveals, EBD was the fourth largest category of students receiving services. The

percentage of students with an identified emotional disturbance has remained

unchanged since 1993 (U. S. Department of Education, 2005).

Although the Department of Education reports the percentage of all students who

receive services under some specific categories (e.g., in 2003, 4.1% of all students

received services for SLD) these data are not available for EBD. From the data

provided, it can be estimated that 0.73% of all students aged 6 through 21 received

services for an emotional disturbance in 2003 (U. S. Department of Education, 2005).

The Department of Education also provides statistics related to race/ethnicity of

students receiving special education services. Figure 1-2 displays the percentage of

special education students in each race/ethnic category that received services under

the label of EBD. The Department of Education collects data on the following

racial/ethnic categories: African American (AA), American Indian/Alaskan Native

(Al/AN), White (W), Hispanic (H), and Asian/Pacific Islander (A/PI). Overall, 11.2% of all

African American students who were eligible for special education received services for

EBD (compared to 44.9% of African American students who received services for SLD).

The statistics for other race/ethnic categories are as follows: 8.0% of American

Indian/Alaskan Native students, 7.9% of White students, 4.9% of Hispanic students, and

4.6% of Asian/Pacific Islander students.

Outcomes for Students with Emotional/Behavioral Disorders

Receiving special education services is generally regarded as positive for students

who experience educational difficulties. For example, Wagner and Cameto (2004)

document that students with EBD receive several types of services and supports to

assist students in managing their emotional and behavioral issues at school including









information regarding students' behaviors across multiple locations and multiple raters

(Sattler & Hodge, 2006).

Finally, I also found that a significant proportion of variance in teacher-rated

behaviors is attributable to the school level. Specifically, 8.73% of variance in

internalizing behavior ratings, 6.64% of variance in externalizing behavior ratings, and

4.23% of variance in social skills ratings was attributable to the school level. The

variance component for competence ratings was not significant at the school level.

These findings suggest that variables at the school level contribute to either behavior

ratings or actual behavior of students in the school.

Research Question 2: Mean Group Differences

Another aim of this study was to examine whether mean group differences in

behavior ratings exist across gender, race/ethnicity, and family income. Several studies

have shown that African American students are rated as exhibiting more problematic

behavior than their Caucasian peers (e.g., Cohen, DuRant, & Cook, 1988; Downy &

Pribesh, 2004; Epstein, March, Conners, & Jackson, 1998; Sbarra & Pianta, 2001;

Rayfield, 1997; Rong, 1996). Additionally, research typically shows that males receive

higher ratings indicating more problem behavior than females even when controlling for

race/ethnicity (e.g., Rayfield, 1997; Rong, 1997). I examined mean group differences

while holding other variables constant.

I found that Hispanic American students were rated as significantly less

internalizing than their Caucasian peers while controlling for gender and family income,

or free/reduced price lunch status (FRL status). I also found that males were rated

significantly less internalizing than their female peers while controlling for race/ethnicity

and FRL status. Significant differences were not found for African American students or









and an Executive Function Cluster. Finally, the CAB-T produces scores for the following

additional clusters: Learning Disability Cluster, Mental Retardation Cluster, Autistic

Spectrum Behaviors Cluster, and an Attention-Deficit/Hyperactivity Cluster. Both the

scale scores and the cluster scores are highly reliable and interpretable (Bracken &

Keith, 2004). For the purposes of this study, the Internalizing, Externalizing, Social

Skills, and Competence (a measure of adjustment and adaptive strength in an area that

is closely related to cognitive and academic functioning scales) were used as outcome

variables.

The CAB-T utilizes standard T scores with a mean of 50 and a standard deviation

of 10. For the Internalizing and Externalizing scales, higher Tscores indicate higher

risk. Scores 5 59 are considered to be in the normal range, scores between 60 and 69

are considered to be mild clinical risk, scores between 70 and 79 are considered

significant clinical risk, and scores 80 are considered very sig nificant clinical risk. The

Social Skills and Competence scales are adaptive scales and, as such, higher Tscores

indicate better adjustment. TScores 5 19 are considered a very significant adaptive

weakness, scores between 20 29 are a significant adaptive weakness, scores

between 30 and 39 are a mild adaptive weakness, scores between 40 59 are

considered to be in the normal range, scores between 60 and 69 are considered a mild

adaptive strength, scores between 70 and 79 are considered a significant adaptive

strength, and scores 80 are a very significant adaptive strength.

Bracken and Keith (2004) suggested that scores on the CAB-T are highly reliable

and cluster interpretations are valid for students across racial/ethnic groups. They

showed that minimal variance in both clinical and adaptive behaviors is due to









in the manual to lie on one of the four factors was entered into the model. The

Internalizing scale consists of 16 questions while the three remaining scales each

consist of 18 questions. Figure 3-1 shows the model that was proposed for samples of

Caucasian and African American students.

Table 3.28 displays a summary of fit statistics from the CFA on the models run on

samples of Caucasian students and African American Students. Model 1, the model run

with a Caucasian student sample, yielded a X2 (135) = 6750.983, indicating poor fit. In

addition, CFI = 0.651 and RMSEA = 0.215, also indicating poor fit. However, TLI =

0.949 indicated good fit. Results for Model 2, the model run with an African American

student sample, also indicated poor fit across most goodness-of-fit indices. For this

model, X2 (113) = 13258.348, CFI = 0.784, and RMSEA = 0.211. Similar to the model

run with a Caucasian American sample, TLI = 0.966, indicating good fit.

Table 3-1. Descriptive statistics for the clinical assessment of behavior subscales
Variable N M SD Range
Internalizing 974 44.80 10.57 22.0 -87.0
Externalizing 973 45.47 9.86 28.0 87.0
Social Skills 973 52.79 9.61 13.0-81.0
Competence 973 53.14 9.68 26.0 -80.0

Table 3-2. Descriptive statistics for teacher efficacy in classroom management and
discipline scale
Variable N M SD Range
Management 65 4.73 0.66 3.00 6.00
External 65 3.04 0.64 1.71 -4.50
Personal 65 4.06 0.68 2.57 5.57

Table 3-3. Percentage of variance at student, teacher/classroom, and school levels
Variable Student Level Teacher/Classroom School Level
Level
Internalizing 56.71% 34.56% 8.73%
Externalizing 70.73% 22.63% 6.64%
Social Skills 73.53% 22.24% 4.23%
Competence 79.27% 19.97% 0.76%









conducted in order to determine how much variance in an outcome variable lies at the

student level and the classroom/teacher level.

Mashburn et al. (2006)'s results indicated that 15% to 33% of the differences in

teachers' ratings of students' behaviors can be attributed to differences between the

teachers or classrooms. Given that a large portion of variance in behavior ratings lies at

the teacher level, the next step is to examine teacher-level variables that may predict

the teacher-specific differences. The authors found that teachers' race/ethnicity was

significantly related to three of the four student-level variables examined in the study.

Specifically, Caucasian teachers rated their students as having more behavior problems

and less competence than African American teachers (Mashburn et al., 2006).

Limitations of Current Research

Research on teachers' perceptions of students' behavior have been examined for

at least 40 years and research has made it clear that African American students are

viewed less favorably than Caucasian students. However, several gaps in the literature

exist and, thus, many questions are still unanswered. Regarding the teacher bias

hypothesis, some research has supported hypotheses of teacher bias by demonstrating

differential ratings of students by Caucasian and African American teachers (Mashburn

et al., 2005; Rayfield, 1997; Rong, 1996) while conflicting research concludes the

opposite (Abidin & Robinson, 2002; Cullinan & Kauffman, 2005; Pigott & Cowen, 2000).

Most of these studies are limited in that the teachers only rated one or two students,

and so a complete evaluation of teacher variables that may affect teachers' ratings is

not possible. When teachers rate multiple students in their classes, it is possible to

examine the proportion of mean group differences in behavior ratings that can be

attributed to the teachers. Findings by Mashburn et al. (2005) indicating that 15% to









Bias in teacher referral hypothesis

Gerber and Semmel (1984) proposed that, instead of using psychometric criteria

for identifying specific learning disabilities, mild mental retardation, and mild emotional

disturbance (i.e., EBD), that teachers can be used as the tests. They argued that

teachers' judgments are reliable to use as indicators of which students should be

identified as needing special education services and that we do not need psychometric

tests. Gresham, Reschly, and Carey (1987) showed that teachers' judgments may be

just as accurate in identifying students with a learning disability as psychometric tests.

Similarly, Shinn, Tindal, and Spira (1987) found support for the accuracy of teacher

judgments for identifying students with mental handicaps. However, they also found

support for gender and racial/ethnic bias in the teachers' referrals. For example, the

authors found that males and females who were referred had the same reading

performance, but compared to the population base rates of poor readers, significantly

more males were referred for reading problems by their teachers. Additionally, African

American students were referred by their teachers at a rate that was higher than

expected (Shinn et al., 1987). However, the referred African American students also

performed lower than referred Caucasian students, so this particular example may not

be indicative of teacher bias.

Bahr and Fuchs (1991) examined whether teachers' judgments about difficult-to-

teach (DTT) students (i.e., students who were considered DTT by their teachers) were

ethnically and racially biased by examining teachers' descriptions of students, teachers'

ratings of academics and behavior, student reading achievement, and observed

classroom behavior. They found that teachers rated African American DTT students as

more appropriate for referral even though the African American DTT students were










levels in order to determine whether predictors of differences should be included.

Finally, I included teacher race/ethnicity and age as predictor variables to remain

consistent with previous literature, and added teacher age, years of experience, and

self-efficacy as predictor variables to build upon the current literature.

A second purpose of this study was to examine the construct equivalence of an

omnibus behavior rating scale that is likely to be used in the assessment of EBD. I

hypothesized that mean group differences in teacher-rated behaviors may be due to a

lack of factorial equivalence across racial/ethnic groups. I utilized Confirmatory Factor

Analytic techniques to explore this hypothesis.


Percentage of ESE Students in Each

Category
50
40 --,
40
g 30
4- 20 -
20
W Percent
b 10 -


-. SLD S/L IND EBD OHI Other
Category


Figure 1-1. Percentage of ESE students in racial/ethnic categories









certain cultural themes (e.g., individualism, competitiveness) that are in contrast to

cultural themes promoted by African American students and families.

However, the difficulties seen in the classroom, including the underachievement

and behavioral problems exhibited by some African American students, cannot be fixed

just by adding more racial/ethnic minority teachers to the teaching force. Ladson-Billings

(2005) stated that, if this was the case, then school districts in Detroit and Washington

D. C. would be exemplary for educating African American students. Therefore,

researchers often call for preservice teachers to be trained in methods that are useful

and appropriate for majority and racial/ethnic minority students (Valles, 1998). Monroe

and Obidah (2004) conducted observations and interviews with an African American

teacher and concluded that, when there is cultural synchronization of teachers and

students (e.g., African American teachers with African American students), there are

effective styles of classroom and behavior management. Furthermore, the authors

noted that more culturally responsive teaching may lead to fewer discipline referrals and

less negative outcomes. A major limitation of this study was that only one classroom

was examined. Although several researchers (Boykin et al., 2005; Hilliard, 1992;

Ladson-Billings, 2005; Monroe & Obidah, 2004; Sleeter, 2001; Valles, 1998) have

discussed hypotheses related to a cultural mismatch, it has been difficult to draw a

direct line between cultural differences and disproportionate representation of minority

students identified with EBD because there are several potential confounding variables

that may play a role as well. Therefore, several researchers (e.g., DeMeis & Turner,

1978; Downey & Pribesh, 2004; Pigott & Cowen, 2000; Rong, 1996) have examined

hypotheses related to a possibility of teacher bias as a main variable of interest.










Internalizing


Questions
2, 4, 5, 12,
15, 18, 19,
24, 36, 43,
45, 50, 53,
55 R9 R7


Questions
3, 8, 14, 16,
21, 23, 27, 33,
37, 40, 41,42,
46, 48, 49, 54,
56, 58


Questions
1,6, 10, 11,
17, 20, 30, 32,
35, 39, 47, 51,
52, 59, 63, 64,
69, 70


Questions
7, 9, 13, 22,
25, 26, 28, 29,
31, 34, 38, 44,
57, 60, 61, 65,
66, 68


Figure 3-1. CAB factor structure.


Social
Skills


Externalizing









Table 3-11. Hierarchical linear model with externalizing
teacher level independent variables


as outcome: Coefficients for


Fixed Effect Coefficient Standard T- Approx df p-Value
Error Ratio
Teacher Age, BOlab 0.043 0.038 1.131 931 0.259
Years Experience, B02ab -0.102 0.058 -1.741 931 0.082
Management, B03ab -3.601 0.789 -4.560 931 0.000*
External, B04ab -2.244 0.587 -3.825 931 0.000*
Personal, B05ab 2.798 0.694 4.034 931 0.000*
Male Teacher, BO6ab 0.048 0.817 0.059 931 0.953
Non-White Teacher, B07ab 0.281 0.981 0.287 931 0.774
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.

Table 3-12. Hierarchical linear model with externalizing as outcome: Coefficients for
cross-level interaction of teacher-level variables and differences between
African American and Caucasian students
Fixed Effect Coefficient Standard T-Ratio Approx df p-Value
Error
Teacher Age, B21ab 0.099 0.101 0.986 57 0.329
Years Experience, B22ab -0.007 0.152 -0.048 57 0.963
Management, B23ab -4.419 2.175 -2.032 57 0.046*
External, B24ab -1.906 1.624 -1.174 57 0.246
Personal, B25ab 2.587 1.807 1.431 57 0.158
Teacher Gender, B26ab -2.711 2.159 -1.256 57 0.215
Teacher Race/Ethnicity, 0.914 2.353 0.388 57 0.699
B27ab
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.









3-14 Hierarchical linear model with externalizing as outcome: Coefficients for
cross-level interaction of teacher-level variables and differences between
m ale and fem ale students ............................... ......... .................. 81

3-15 Hierarchical linear model with externalizing as outcome: Coefficients for
school-level independent variables ...... .................... ................ 82

3-16 Hierarchical linear model with social skills as outcome: Coefficients for
student-level independent variables ............. ............... ........................... 82

3-17 Hierarchical linear model with social skills as outcome: Coefficients for
teacher level independent variables ...................................... .......................... 82

3-18 Hierarchical linear model with social skills as outcome: Coefficients for cross-
level interaction of teacher-level variables and differences between African
Am erican and Caucasian students........................................ .......................... 83

3-19 Hierarchical linear model with social skills as outcome: Coefficients for cross-
level interaction of teacher-level variables and differences between Hispanic
American and Caucasian students...... ..................................... 83

3-20 Hierarchical linear model with social skills as outcome: Coefficients of cross-
level interaction of teacher-level variables and differences between male and
fem ale students .......................................... ................................... 84

3-21 Hierarchical linear model with social skills as outcome: Level 3 coefficients....... 84

3-22 Hierarchical linear model with competence as outcome: Coefficients for
student-level independent variables ............. ............... ........................... 85

3-23 Hierarchical linear model with competence as outcome: Coefficients for
teacher-level independent variables...................................................... 85

3-24 Hierarchical linear model with competence as outcome: Coefficients for
cross-level interaction of teacher-level variables and differences between
African Am erican and Caucasian students ..................................................... .. 86

3-25 Hierarchical linear model with competence as outcome: Coefficients for
cross-level interaction of teacher-level variables and differences between
Hispanic American and Caucasian students............................. ....... ......... 86

3-26 Hierarchical linear model with competence as outcome: Coefficients for
cross-level interaction of teacher-level variables and differences between
m ale and fem ale students ............................... ......... .................. 87









LIST OF FIGURES


Figure page

1-1 Percentage of ESE students in racial/ethnic categories....... ......................... 51

1-2 Percentage of students eligible for ESE who received services for EBD ............52

3-1 C A B factor structure........... ..... .... ...................... .... ................ ..... 88









Dembo, M. H. & Gibson, S. (1985). Teachers' sense of efficacy: An important factor in
school improvement. The Elementary School Journal, 86, 173-184.

DeMeis, D. K. & Turner, R. R. (1978). Effects of students' race, physical attractiveness,
and dialect on teachers' evaluations. Contemporary Educational Psychology, 3,
77-86.

Donovan, M. S., & Cross, C. T. (Eds.). Minority students in special and gifted education.
Washington, DC: National Academy Press.


Downey, D. B. & Pribesh, S. (2004). When race matters: Teachers' evaluations of
students' classroom behavior. Sociology of Education, 77, 267-282.

Dusek, J. B. & Joseph, G. (1983). The bases of teacher expectancies: A meta-analysis.
Journal of Educational Psychology, 75, 327-346.

Emmer, E. T. & Hickman, J. (1991). Teacher efficacy in classroom management and
discipline. Educational and Psychological Measurement, 51, 755-765.

Epstein, J. N., March, J. S., Conners, C. K., Jackson, D. L. (1998). Racial differences on
the conners teacher rating scale. Journal of Abnormal Child Psychology, 26,
109-118.
Epstein, J. N., Willoughby, M., Valencia, E. Y., Tonev, S. T., Abikoff, H. B., Arnold, L. E.,
& Hinshaw, S. P. (2005). The role of children's ethnicity in the relationship
between teacher ratings of attention-deficit/hyperactivity disorder and observed
classroom behavior. Journal of Consulting and Clinical Psychology, 73, 424-434.

Epstein, M. H. & Cullinan ,D. (1998). Scale for assessing emotional disturbance. Austin,
TX: Pro-ed.

Florian, L., Hollenweger, J., Simeonsson, R. J., Wedell, K., Riddell, S., Terzi, L., &
Holland, A. (2006). Cross-cultural perspectives on the classification of children
with disabilities: Part I. Issues in the classification of children with disabilities. The
Journal of Special Education, 40, 36-45.

Florida Department of Education, Bureau of Exceptional Education and Student
Services. (2009). Florida Statutes and State Board of Education Rules: Excerpts
for special programs. Tallahassee, FL: Author.

Frey, A. (2002). Predictors of placement recommendations for children with behavioral
or emotional disorders. Behavioral Disorders, 27, 126-126.

Gelb, S. A. & Mizokawa, D. T. (1986). Special education and social structure: The
commonality of exceptionalityy". American Educational Research Journal, 23,
543-557.


103









factors (e.g., feelings of sadness, frequent crying, restlessness, loss of interest in

friends or school work, fears, phobias, excessive worrying) and external factors (e.g.,

inability to build or maintain interpersonal relationships, disruptive behaviors,

noncompliance, aggression, poor social skills). A student must exhibit one or more of

the above characteristics for a minimum of six months and display characteristics in at

least two settings (one setting must be school) in order to receive special education

services (Florida Department of Education, 2009).

Although a nation-wide system of identification does not exist, the major

procedures are usually similar. If and when a student exhibits emotional or behavioral

difficulties in the classroom, the teacher should attempt pre-referral interventions, such

as a behavior management plan, to ameliorate any problem behavior. If the student

continues to exhibit difficulties, the teacher refers the student to be evaluated for special

education. Once referred, a school psychologist collects informal assessment data,

including a file review and interviews, and formal assessment data, including behavior

rating scales. The school psychologist prepares a psychoeducational report outlining

the findings of the evaluation to present to a child study team. The team consists of the

student's teacher, a special education teacher, the exceptional student education

director, the school psychologist, and the student's parents, and the team decides

whether the student is eligible for special education services provided under the EBD

label.

Many states now utilize a response to intervention (RTI) model under which

small-group or individualized evidence-based interventions) are provided to students

with emotional and/or behavioral concerns prior to referral. In many school districts









determining what proportion of variance in teachers' ratings of students' behaviors is

attributable to the teachers and the schools) and for an analysis of interactions between

variables at the different levels.

The preliminary analysis consisted of a random effects ANOVA model, or a

completely unconditional three-level model, to determine whether additional variance

exists at the teacher level (Level 2) and the school level (Level 3), and whether added

teacher- and school-level variables are justified. If a significant amount of variance

exists at the teacher and school levels, the results indicate that variables at the teacher

or school level also contribute to teacher-rated behaviors. Therefore, the model was

expanded to include student-, teacher-, and school-level variables in an intercepts and

slopes as outcomes (ISAO) model. From the ISAO models, I examined the main effects

of the Level 1 variables (i.e., student gender, family income, and race/ethnicity), the

main effects of the Level 2 variables (i.e., teachers' race/ethnicity, gender, age, years of

experience, and self-efficacy), and the main effects of the Level 3 variables (i.e., school

size, average family income, and percentage of minority students). I also examined the

interactions between the Level 1 and Level 2 variables. In an ISAO model, significant

interactions can be interpreted as a Level 2 variable predicting differences among Level

1 variables. Interactions with students' gender and race/ethnicity with teachers'

race/ethnicity, gender, age, years of experience, and self-efficacy were examined,

controlling for students' family income, school size, average family income, and

percentage of minority students.

The procedure of examining the unconditional model and intercepts and slopes as

outcomes model was carried out four times in order to evaluate the four outcome









Table 3-13. Hierarchical linear model with externalizing as outcome: Coefficients for
cross-level interaction of teacher-level variables and differences between
Hispanic American and Caucasian students
Fixed Effect Coefficient Standard T-Ratio Approx df p-Value
Error
Teacher Age, B31ab 0.016 0.118 0.132 57 0.896
Years Experience, B32ab -0.091 0.204 -0.446 57 0.657
Management, B33ab 0.838 2.964 0.283 57 0.778
External, B34ab 2.987 1.941 1.539 57 0.129
Personal, B35ab -1.726 2.779 -0.621 57 0.537
Teacher Gender, B36ab 0.164 2.287 0.071 57 0.944
Teacher Race/Ethnicity, 3.343 3.350 0.998 57 0.323
B37ab
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.

Table 3-14. Hierarchical linear model with externalizing as outcome: Coefficients for
cross-level interaction of teacher-level variables and differences between
male and female students
Fixed Effect Coefficient Standard T-Ratio Approx df p-Value
Error
Teacher Age, B41ab -0.151 0.070 -2.152 57 0.035*
Years Experience, B42ab 0.156 0.104 1.500 57 0.139
Management, B43ab 1.723 1.505 1.145 57 0.258
External, B44ab -0.103 1.107 -0.093 57 0.927
Personal, B45ab -2.174 1.350 -1.611 57 0.112
Teacher Gender, B46ab -0.378 1.540 -0.246 57 0.807
Teacher Race/Ethnicity, -2.510 1.870 -1.343 57 0.185
B47ab
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.









behaviors exhibited by students with internalizing disorders (e.g., withdrawal, irritability,

nervousness).

Regarding the externalizing variable, African American students were rated as

exhibiting significantly more externalizing behaviors. Teacher age, years experience,

gender, race/ethnicity, and feelings of self-efficacy related to external influences and

personal efficacy did not significantly predict the teacher-specific differences for the

race/ethnicity variables in this study. However, I found that teacher self-efficacy related

to classroom management and discipline significantly predicts the teacher-specific

differences in ratings for Caucasian and African American students. As teachers' sense

of self-efficacy regarding their behavior management and classroom discipline skills

increased, the teacher-specific difference between African American students and

Caucasian students decreased. However, as teachers' self-efficacy decreases, teacher-

specific differences in ratings increase. There are two possible reasons for this finding.

One hypothesis is that teachers with higher self-efficacy in behavior management skills

actually employ better skills in the classroom, and therefore, their students may exhibit

less behavior problems. A second hypothesis is that teachers with higher self-efficacy

skills relating to behavior management do not perceive behaviors exhibited by African

American students as being as problematic as teachers with a lower sense of self-

efficacy. In either case, the implications for practice imply a greater need for preservice

and inservice training for teachers regarding classroom management and discipline.

Further research in this area is warranted as well.

I also found that males were rated as less externalizing than females in this

study. Teacher age, but no other variables, significantly predicted the teacher-specific









mean difference between African American and Caucasian students on the competence

scale observed for specific teachers was not predicted by any teacher-level variables.

Table 3-25 displays the coefficients for the cross-level interactions between

teacher-level variables and the contrast of Hispanic American and Caucasian students

on the competence scale. None of the cross-level interactions were significant implying

that the relationships between scores on the competence variable and teacher variables

are similar for Hispanic American and Caucasian students.

Table 3-26 displays the coefficients for the cross-level interactions between

teacher-level variables and the contrast of male and female students on the

competence scale. None of the cross-level interactions were significant implying that the

relationships between scores on the competence variable and teacher variables are

similar for male and female students. Stated differently, the size of the mean difference

between male and female students on the externalizing scale observed for specific

teachers was not predicted by any teacher-level variables.

Table 3-27 displays the coefficients for the Level 3 variables in the hierarchical

linear model with the Competence scale as the outcome variable. Larger schools

produced significantly lower ratings on the Competence scale. The coefficients for

school percentage of minority students and school percentage of students eligible for

free or reduced-priced lunch were not significant.

Confirmatory Factor Analysis Examining Factorial Equivalence

To examine the latent structure of the CAB by ethnicity, two sets of data were

analyzed using latent variable structural equation modeling with the Mplus version 5.1

program. The CAB manual suggests that each question contributes to one of four

factors: Internalizing, Externalizing, Social Skills, or Competence. Each item proposed









Competence

Table 3-22 displays the coefficients for the Level 1 variables in the hierarchical

linear model with the Competence scale as the outcome variable. The Competence

intercept coefficient was 52.914. Students who were eligible for free or reduced-priced

lunch received significantly lower competence ratings than their peers who were not

eligible for free or reduced-priced lunch. African American students received

significantly lower competence ratings than their Caucasian peers. Hispanic Americans

received higher social competence ratings than their Caucasian peers, although this

difference was not significant. Finally, males received significantly higher competence

ratings than their female peers.

Table 3-23 displays the coefficients for the Level 2 variables in the hierarchical

linear model with the Competence scale as the outcome variable. Teachers with higher

self-efficacy for behavior management rated students as significantly more competent.

Teachers with a higher belief that external influences beyond teachers' control affect

student outcomes also produced significantly higher competence ratings. Conversely,

teachers with a higher sense of personal self-efficacy produced significantly lower

competence ratings. The coefficients for teacher age, years experience, gender, and

race/ethnicity were not significant.

Table 3-24 displays the coefficients for the cross-level interactions between

teacher-level variables and the contrast of African American and Caucasian students on

the competence scale. None of the cross-level interactions were significant implying that

the relationships between scores on the competence variable and teacher variables are

similar for African American and Caucasian students. Stated differently, the size of the









that, at times, some teachers may have African American and low-income students

whom they do not want to teach and may not like. Irvine (1990) stated that the concept

of cultural aversion, the tendency for teachers and administrators to avoid discussing

race/ethnicity-related issues, leads to a greater lack of cultural synchronization and lack

of teachers' understanding of African American students. He further asserted that

teachers disregard components of African American students' culture (e.g., beliefs,

behaviors, perceptions) when they fail to acknowledge their students' race/ethnicity and

that they should be incorporating aspects of students' culture in their approaches to

teaching. This lack of cultural understanding among teachers, or what some view as

teacher bias, may lead to the overrepresentation of African American students identified

with EBD.

Researchers began thinking about how teachers might view racial/ethnic minority

students differently from their Caucasian peers as early as the 1960's (e.g., Gottlieb,

1964). The early research on teachers' potentially different views of students focused on

teacher expectations and how those differed based on students' race/ethnicity. More

recently, research has examined how teachers' views of students differ more objectively

through the use of behavior rating scales.

Early research on teacher's perceptions of Caucasian and minority students

Although the impact of a prejudicial school environment has been a topic of

interest for researchers, it is difficult to study the nature of bias and prejudice in the

schools because of its subtle nature. In earlier studies, researchers used hypothetical

situations and vignettes in order to elicit information about teachers' expectations

(Donovan & Cross, 2002). Pioneering the field, Gottlieb (1964) investigated how both

Caucasian and African American elementary school teachers viewed their work and









Ogbu, J. U. (1991). Minority coping responses and school experience. Journal of
Psychohistory, 18, 433-456.

Oswald, D. P., Coutinho, M. J., Best, A. M., & Singh, N. N. (1999). Ethnic representation
in special education: The influence of school-related economic and demographic
variables. The Journal of Special Education, 32, 194 206.

Patton, J. M. (1998). The disproportionate representation of African Americans in
special education. The Journal of Special Education, 32, 25-31.

Pigott, R. L. & Cowen, E. L. (2000). Teacher race, child race, racial congruence, and
teacher ratings of children's school adjustment. Journal of School Psychology,
38, 177-196.

Podell, D. M. & Soodak, L. C. (1993). Teacher efficacy and bias in special education
referrals. Journal of Educational Research, 86, 247-253.

Raudenbush, S. W. & Bryk, A. S. (2002). Hierarchical linear models: Applications and
data analysis methods (2nd ed.). Thousand Oaks, CA: Sage Publications.

Raven, J. C. (1947). Progressive Matrices. London: Lewis.

Rayfield, A. (1997). Concurrent validity of the sutter-eyeberg student behavior inventory
with grade school children. Unpublished doctoral dissertation, University of
Florida.

Reid, R. (1995). Assessment of ADHD with culturally different groups: The use of
behavioral rating scales. School Psychology Review, 24, 537-560.

Reid, R., Casat, C. D., Norton, H. J., Anastopulos, A. D. & Temple, P. (2001). Using
behavior rating scales for ADHD cross ethnic groups: The IOWA Conners.
Journal of Emotional and Behavioral Disorders, 9, 210-218.

Reynolds, C. R. & Kamphaus, R. W. (1992). Behavior Assessment Scale for Children:
Manual. Circle Pines, MN: American Guidance Service, Inc.

Rong, X. L. (1996). Effects of race and gender on teachers' perception of the social
behavior of elementary students. Urban Education, 31, 261-290.

Rosenthal, R. & Jacobson, L. (1968). Pygmalion in the classroom. New York: Holt,
Rinehart, & Winston.
Ryabov, I. & Van Hook, J. (2007). School segregation and academic achievement
among Hispanic children. Social Science Research, 36, 767-788.


106









that is the case. In addition, Ogbu (1991) explains that African American students and

their families may have a lack of trust for Caucasian teachers and administrators, and

so it may be that African American students do engage in more problem behaviors

when placed with Caucasian teachers.

Although Ogbu's (1991) oppositional culture theory has been frequently cited to

explain behavioral differences (e.g., Ainsworth-Darnell and Downey, 1998; Downey &

Pribesh, 2004), when it has been used to explain differences in the way teachers rate

students' behaviors, the theory has not been supported by data in the literature.

Downey and Pribesh (2004) tested the theory by examining teacher ratings of

kindergartners' and 8th graders' behaviors by compiling data from the Early Childhood

Longitudinal Study (National Center for Education Statistics, 2000). The authors

hypothesized that if Ogbu's theory was correct, teachers' ratings of students' behaviors

should be similar for Caucasian and African American students in kindergarten but that

African American students' behavior should be rated worse by teachers in 8th grade

after African American students have had substantial amount of time in an educational

setting and time to develop a learned helplessness regarding their educational potential.

Downey and Pribesh (2004) created an Externalizing Problem Behaviors Scale from

teachers' evaluations of students' externalizing behaviors (e.g., anger, aggression), and

an Approaches to Learning Scale from teachers' ratings of students' prosocial

behaviors. The authors concluded that, since African American students were rated as

exhibiting more problematic behavior than Caucasian students as early as kindergarten,

there was little support for the oppositional culture theory because kindergartners are

likely too young to understand biases against them and thus exhibit problem behaviors









their students. He administered surveys and interviewed 89 elementary school teachers

in the Midwest who taught in schools that he classified as inner-city. He found that

African American teachers were more likely than Caucasian teachers to come from

similar backgrounds of inner city children and that Caucasian teachers in these schools

were more critical of students and parents than African American teachers. In order to

identify teachers' perceptions of inner-city students, Gottlieb (1964) asked teachers to

choose adjectives to describe the "outstanding characteristics of the children (p. 352)" of

the students they were working with from a list of 33 adjectives. Caucasian teachers

selected "Talkative", "Lazy", "Fun Loving", "High Strung", and "Rebellious" with the

highest frequency. In contrast, African American teachers selected "Fun Loving",

"Happy", "Cooperative", "Energetic", and "Ambitious" with the highest frequency.

Gottlieb (1964) concluded that African American teachers were less critical of their

students than Caucasian teachers probably because many of them had come from

similar backgrounds and had overcome many of the same social barriers that faced

their students.

With the publication of Rosenthal and Jacobson's (1968) Pygmalion in the

Classroom, which asserted that teachers' expectations about student performance

contributed to a self-fulfilling prophecy and actual student performance, researchers

began to study how teachers' expectations of students' academic achievement and

behavior differed by gender and ethnicity. Cooper, Baron, and Lowe (1975) asked 128

Caucasian, female, mostly middle class undergraduate psychology and education

students to provide their expectations for students given student vignettes that varied by

ethnicity and socioeconomic status. Teachers indicated lower expectations for the









Table 3-26. Hierarchical linear model with competence as outcome: Coefficients for
cross-level interaction of teacher-level variables and differences between
male and female students
Fixed Effect Coefficient Standard T-Ratio Approx df p-Value
Error
Teacher Age, B41ab 0.024 0.074 0.324 57 0.747
Years Experience, B42ab 0.021 0.109 0.191 57 0.849
Management, B43ab -0.510 1.580 -0.323 57 0.748
External, B44ab 0.874 1.163 0.751 57 0.455
Personal, B45ab 0.365 1.412 0.258 57 0.797
Teacher Gender, B46ab 2.022 1.622 1.247 57 0.218
Teacher Race/Ethnicity, -0.193 1.965 -0.098 57 0.923
B47ab
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.

Table 3-27. Hierarchical linear model with internalizing as outcome: Coefficients for
school-level independent variables
Fixed Effect Coefficient Standard T-Ratio Approx df p-Value
Error
School Size, G001a -0.008 0.003 -2.995 7 0.021*
Percentage Minority 0.015 0.024 0.616 7 0.557
Students in School, G002a
Percentage Students -0.132 0.069 -1.900 7 0.098
Eligible for FRL, G003a
aThis variable has been centered around its group mean.

Table 3-28. Summary of fit statistics for the confirmatory factor analyses by
race/ethnicity
Race/Ethnicity X2 df CFI TLI RMSEA

Caucasian American 6750.983* 135 0.671 0.949 0.215
Students

African American 13258.348* 113 0.784 0.966 0.211
Students
aN = 1059. bN = 569









level variable as research has indicated various indicators of school socioeconomic

status to predict other important outcomes (Caldas & Bankston, 1998; Ryabov & Van

Hook, 2007). I was interested in whether differences in ratings varied across average

family income.

Racial/Ethnic Minority Students

The percentage of racial/ethnic minority students in each school was obtained

from each district and was included as a school-level variable. Schools that enroll few

minority students may have a different climate than schools that enroll a large

percentage of minority students. Percentage of minority students in each school is

included as a school-level variable to examine whether differences in teacher-rated

behaviors across racial/ethnic groups is related to the percentage of minority students in

the school.

Analyses

Hierarchical Linear Modeling

The purpose of this study was to explore the effects of students' gender, ethnicity,

and family income and teachers' gender, ethnicity, age, years of experience, and

perception of self-efficacy on teachers' ratings of students' clinical and adaptive

behavior. An additional purpose of this study was to examine whether variance in

teachers' ratings of students' behaviors are related to school-level factors including

school size, average of family income, and percentage of minority students in each

school. In this particular research design, students are nested within classrooms which

are nested within schools. Therefore, hierarchical linear modeling techniques (HLM;

Raudenbush & Bryk, 2002) were used to analyze the data. The use of HLM procedures

allows for the estimation of variance and covariance components with nested data (i.e.,









African American teachers had fewer difficulties with overrepresentation (Serwatka et

al., 1995). Similarly, Oswald et al. (1999) found that districts with fewer African

American students displayed higher probability for African American students to be

identified as EBD, especially in low-poverty communities. However, districts with more

African American students were less likely to identify disproportionate rates of African

American students with EBD. The results of these studies suggest that school and

district level factors may contribute to teachers' perceptions of students' behaviors.

Cross-Cultural Examination of Psychometric Properties

An additional hypothesis is that the assessment techniques that psychologists use

may not be valid and reliable across members from different racial/ethnic groups. As

previously mentioned, Donovan and Cross (2002) state that the standards for referral

and assessment may be biased or may be applied differently across racial/ethnic

groups. While it is difficult to conceptualize and measure their claims, we can measure

variables related to the reliability and validity of the instruments we use in the evaluation

process. Stated differently then, the hypothesis proposes that the tests we use may not

be valid for use across different racial/ethnic groups. Jensen (1980) and others

differentiate between internal criteria and external criteria when examining tests for bias

across groups. Internal criteria of test bias include the psychometric qualities of the test,

including the factor structure and differential item analyses. External sources of bias

include those variables that are not related to the test per se, but to factors that may

influence performance on the test including whether the tests are timed and the

race/ethnicity and gender of the examiner (Jenson, 1980).

The assessment process for EBD typically includes the use of behavior rating

scales (Sattler & Hoge, 2006). When using behavior rating scales with students from









CHAPTER 4
DISCUSSION

The purpose of this study was to build upon the current literature by examining

possible influences on the disproportionate representation of some racial/ethnic minority

groups in the EBD category of special education. The primary instrument I used was the

CAB-T and I examined four outcome variables: internalizing, externalizing, social skills,

and competence. First, I was interested in examining the variance components in

teacher-rated behaviors at the student, teacher/classroom, and school levels. I was also

interested in examining whether mean-group differences in teacher-rated behaviors

existed across Caucasian, African American, and Hispanic students. Additionally, I was

interested in whether teacher-specific differences in ratings could be predicted by a host

of teacher variables. Finally, I was interested in whether the CAB-T demonstrates

factorial validity across race/ethnic groups. A discussion of the most relevant findings of

each research question and their implications for research and practice is discussed

below.

Research Question 1: Proportion of Variance

One goal of this study was to examine variance components in teacher-rated

behaviors. When teachers are asked to complete behavior rating scales, they are asked

with the intent to capture information regarding how the student behaves compared to

same-age peers. While information related to student behavior is obtained, it is

important to keep in mind that the data present the teacher's perception of how the

student behaves. By examining variance components, researchers can attempt to

understand how much of behavior rating scale data are related to actual student









identified with EBD than Caucasian, American Indian/Alaskan Native, Asian/Pacific

Islander, and Hispanic students.

Although several researchers (e.g., Abidin & Robinson, 2002; Bahr & Fuchs,

1991; Rong, 1996) have examined the topic of disproportionate representation of

racial/ethnic minority students identified with special education needs, it is difficult to

apply one specific conceptual framework to understand why or how disproportionate

representation occurs (Coutinho & Oswald, 1998; Oswald et al., 1999). Serwatka,

Deering, and Grant (1995) found that the causes most commonly cited in the literature

include

cultural differences that may lead to a predisposition of a diagnosis of an
emotional handicap, a lack of uniform identification procedures, bias in the
assessment instruments used in diagnosis, the attendant problems of
poverty, and a general pattern of racial discrimination in society reflected in
school systems. (p. 493)

Artiles and Trent (1994) argued that a number of variables should be considered

when discussing issues related to overrepresentation. In their review of the history of

overrepresentation, they identified litigation cases, debate about systemic issues

relating to definitions of special education categories, debate about biased referral and

assessment procedures, the effects of socioeconomic status on educational

performance, theories addressing the school failure of minorities, and others as factors

that are linked with overrepresentation of minority students in special education.

Additional researchers have identified similar theories and hypotheses relating to

the problem of overrepresentation. Donovan and Cross (2002) hypothesized that the

standards for referral and assessment of students for special education may be biased,

or that they may not be applied consistently across racial/ethnic groups, although no

data are presented to support this claim. Other researchers have asserted that









they aren't likely to accept it at school"). The personal efficacy factor assesses teachers'

perception that they can make a difference in students' lives (e.g., "If a student masters

a new concept quickly this might be because I knew the necessary steps in teaching the

concept"). Emmer and Hickman (1991) show that their scale has adequate test-retest

reliability, internal consistency, and construct validity.

Student-Level Data

Gender

Information on students' gender was obtained from each student's perspective

school district. Gender was represented as a dummy coded variable with girls assigned

a value of 0 and boys assigned a value of 1. Gender is included as a student-level

variable in order to examine differences in teachers' perceptions of students' behaviors

due to students' gender (e.g., Clifton et al., 1986)

Race/Ethnicity

Information on students' ethnicity was obtained from each student's perspective

school district. This student-level variable is the main variable of interest. Race/ethnicity

was represented by dummy coded variables representing Caucasian, African American,

and Hispanic American students. Table 2.1 presents the dummy codes.

Family Income

Information on whether students are eligible for free or reduced-price lunch was

obtained from each student's perspective school district and was used as a measure of

family income. Family income was represented as a dummy coded variable with those

not eligible for free or reduced-price lunch coded 0 and those students who are eligible

for free or reduced-price lunch coded as 1. Family income is included as a student-level

variable in order to examine differences in teachers' perceptions of students' behaviors









SD = 240.70). The schools had a range of 5.60% to 86.50% for percentage minority

students enrolled (M = 40.75, SD = 25.78). Finally, the schools had a range of 59.30%

to 88.20% for percentage of students eligible for free or reduced-priced lunch (M =

73.08, SD = 9.61).

The larger data set was utilized for the confirmatory factor analyses as I wanted to

include all possible data for the CAB-T. One factor analysis was conducted on a

Caucasian sample that included data for 1,059 4th and 5th grade students. A second

factor analysis was conducted on an African American sample that included data for

569 4th and 5th grade students.

Measures

Clinical Assessment of Behavior Teacher Form (Bracken & Keith, 2004)

The Clinical Assessment of Behavior Teacher Form (CAB-T) is an omnibus

behavior rating scale that is designed to aid in the assessment of children who may

need behavioral, educational, or psychiatric treatment or intervention. The CAB-T is

aligned with current diagnostic criteria and includes additional clusters that assess

areas of public concern (e.g., bullying, gifted). The CAB-T contains 70 items and is

nationally normed on a representative sample and can be used with students who are

5- through 18-years old. The CAB-T produces scores for two Clinical scales including

the Internalizing Behaviors Scale (INT) and the Externalizing Behaviors Scale (EXT).

The CAB-T further produces scores for Clinical Clusters including an Anxiety Cluster,

Depression Cluster, Anger Cluster, Aggression Cluster, Bullying Cluster, and Conduct

Problems Cluster. In addition, the CAB-T produces scores for two Adaptive scales

including the Social Skills Scale (SOC) and the Competence Scale (COM). The CAB-T

further produces scores for Adaptive Clusters including a Gifted and Talented Cluster









BIOGRAPHICAL SKETCH

Christina Peters was born in Winter Park, Florida, in 1983. She spent many

wonderful years in the area and graduated from Winter Park High School in 2001. She

earned her B.S. in Psychology and her B.A. in Criminology from the University of Florida

(UF) in 2005. She earned her M.Ed. in school psychology from UF in 2007. Specializing

in Emotional and Behavioral Disorders (EBD), Christina had the opportunity to complete

many unique practicum placements to enhance professional development including the

public school system, a center day school for students with EBD, a Department of

Juvenile Justice Residential Facility, Shands Teaching Hospital, P. K. Yonge, and

PACE Center for Girls. Christina completed a year-long internship with Hillsborough

County Public Schools as a culminating experience, throughout which she learned how

to become an independent practitioner. Christina has taken the opportunity to travel to

several international locations in order to enhance cultural knowledge and awareness

and to explore the outdoors. Additionally, she has led commercial expedition wilderness

trips for college students in Alaska and several other domestic trips with friends.

Christina's future plans include obtaining licensure as a psychologist, working

collaboratively with school system professionals to enhance the lives of children,

exploring the world one country at a time, and living a fulfilling life with family and

friends.


109









outcome variables from the CAB-T were assessed (i.e., internalizing, externalizing,

social skills, competence). Hierarchical linear modeling techniques were used to (1)

analyze variance components across three levels (2) examine mean-group differences

across outcome variables for student gender, race/ethnicity, and free/reduced price

lunch status, (3) examine differences across school factors, and (4) examine whether

teacher variables (i.e., age, years experience, gender, race/ethnicity, self-efficacy)

predicted teacher-specific differences in ratings. Results indicated that a significant

amount of variance in ratings was attributable to the teacher- and school- levels.

Several mean-group differences emerged (e.g., African Americans were rated by

teachers as exhibiting more externalizing behaviors than their Caucasian peers). Some

teacher specific differences in ratings across groups were predicted by teacher self-

efficacy for behavior management and teacher age, but not teacher race/ethnicity,

gender, or years experience. A final purpose was to examine the psychometric

properties of the CAB-T across racial/ethnic groups to ensure that the same factors are

measured across groups. The results from Confirmatory Factor Analyses of Caucasian

and African American samples indicated similar goodness-of-fit indices for both groups

although a poor fit of the model to the data.









gender differences observed in this study. As teachers' age increased, the teacher-

specific gender difference in ratings grew smaller. This may be explained by recent

research discussing an increasing trend in externalizing behaviors for females (Moller-

Leimkuhler & Yucel, 2009; Moretti, Catchpole, & Odgers, 2005). It may be that older

teachers still perceive females as less externalizing than males while younger teachers

are more perceptive to changing gender differences. This is another area where

additional research is needed.

Finally, I found that African Americans received significantly lower social skills

ratings than their Caucasian peers. The only teacher variable that predicted a teacher-

specific difference in social skills ratings across these groups was self-efficacy in

classroom management and discipline. This finding suggests that, as teacher self-

efficacy for classroom management and discipline increases, the size of the teacher-

specific difference in social skills ratings decreases. However, as teacher self-efficacy

for classroom management and discipline decreases, the size of the teacher-specific

difference in social skills ratings increases. Mashburn et al. (2006) found that teachers'

self-efficacy ratings were positively associated with their reports of students' social

competence as well. Additional research on the relationship between teacher self-

efficacy in classroom management and discipline and perceptions of students' social

skills is needed.

Research Question 4: Factorial Equivalence

A final aim of this study was to examine the factorial equivalence of the CAB-T

across racial/ethnic groups. My hypothesis was that mean-group differences in ratings

on behavior rating scales may be related to the fact that different factors are measured

across different racial/ethnic groups. I conducted a confirmatory factor analysis on a










Table 3-22. Hierarchical linear model with competence as outcome: Coefficients for
student-level independent variables
Fixed Effect Coefficient Standard T-Ratio Approx df p-Value
Error
Intercept, POb 52.914 0.465 113.736 7 0.000
FRL Eligible Slope, -1.553 0.731 -2.124 64 0.037*
Pla
African American -2.235 0.886 -2.524 57 0.015*
Slope, P2a
Hispanic American 0.630 1.196 0.526 57 0.600
Slope, P3a
Male Slope, P4a 2.475 0.631 3.921 57 0.000*
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.

Table 3-23. Hierarchical linear model with competence as outcome: Coefficients for
teacher-level independent variables
Fixed Effect Coefficient Standard T- Approx df p-Value
Error Ratio
Teacher Age, BOlab 0.007 0.039 0.169 931 0.866
Teacher Year, B02ab 0.052 0.059 0.881 931 0.379
Management, B03ab 2.027 0.806 2.514 931 0.012*
External, B04ab 1.689 0.601 -2.808 931 0.006*
Personal, B05ab -1.540 0.709 -2.171 931 0.030*
Male Teacher, B06ab 0.028 0.840 0.034 931 0.973
Non-White Teacher, B07ab -0.479 1.005 -0.477 931 0.633
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.









CHAPTER 1
INTRODUCTION

The disproportionate representation of racial/ethnic minority students identified to

receive special education services has been a topic of concern in the psychological and

educational literature for approximately 40 years (Artiles & Trent, 1994). Despite the fact

that disproportionate representation is a long-standing issue, questions remain as to

whether disproportionate representation actually exists, and what causes it (Artiles &

Trent, 1994; Oswald, Coutinho, Best, & Singh, 1999). Although the federal government

recognizes several categories of possible placement for students in special education,

the difficulties with disproportionate representation only seem to appear in three main

categories (Gelb & Mizokawa, 1986).Those categories include specific learning

disabilities (SLD), mental retardation (MR), and emotional and behavioral disorders

(EBD). This study will focus on the possibility of disproportionate representation as it

relates to the EBD category of special education.

Emotional and Behavioral Disorders

The Individuals with Disabilities Education Act (IDEA) is a federal law that

mandates a free and appropriate education for all eligible children and youth with

disabilities in the United States (U.S. Department of Education, 2005). The law

recognizes 11 categories of disabilities: intellectual disability (IND, previously mental

retardation, MR), hearing impairment (including deafness), speech or language

impairment, visual impairment (including blindness), emotional/behavioral disorders

(EBD), orthopedic impairment, autism, traumatic brain injury, specific learning disability

(LD), deaf-blindness, and multiple disabilities (U. S. Department of Education, 2006).









between male and female students on the internalizing scale observed for specific

teachers was not predicted by the teacher-level variables.

Table 3-9 displays the coefficients for the Level 3 variables in the hierarchical

linear model with the Internalizing scale as the outcome variable. Larger schools had

significantly higher ratings for students' internalizing behaviors. The coefficients for

school percentage of minority students and school percentage of students eligible for

free or reduced-priced lunch were not significant.

Externalizing

Table 3-10 displays the coefficients for the Level 1 variables in the hierarchical

linear model with the Externalizing scale as the outcome variable. The Externalizing

intercept coefficient was 46.032. Students who were eligible for free or reduced-priced

lunch were rated as more externalizing than their peers who were not eligible for free or

reduced-priced lunch, although the coefficient was not significant African Americans

were rated as significantly more externalizing than their Caucasian peers. Hispanic

Americans were rated as less externalizing than their Caucasian peers, although this

was not a significant difference. Finally, males were rated as significantly less

externalizing than their female peers.

Table 3-11 displays the coefficients for the Level 2 variables in the hierarchical

linear model with the Externalizing scale as the outcome variable. Teachers with higher

self-efficacy for behavior management rated students as significantly less externalizing.

Additionally, teachers with a higher belief that external influences beyond teachers'

control affect student outcomes rated students significantly less externalizing.

Conversely, teachers with a higher sense of personal self-efficacy rated students as









Conners, & Jackson, 1998; Sbarra & Pianta, 2001) and an important finding as I found

differences even after controlling for gender and SES.

I was surprised to find that males were rated significantly less externalizing than

their female peers while controlling for race/ethnicity and FRL status. Cell sizes were

relatively similar: 109 African American females, 100 African American males, 316

Caucasian females, and 316 Caucasian males. In a recent study, Moller-Leimkuhler

and Yucel (2009) examined self-ratings of depression and other variables in a college

sample in Germany. They found that females displayed significantly higher ratings of

depression in addition to many externalizing variables including aggression and

irritability. Indeed, Moretti, Catchpole, and Odgers (2005) discussed trends on females'

involvement in aggressive behavior and called for additional research studying female

aggression as rates have been increasing for over 20 years.

Regarding social skills, I found that African American students were rated as

displaying significantly less adaptive social skills than their Caucasian peers while

controlling for gender and FRL status. I also found that males were rated as displaying

significantly more adaptive social skills than females while controlling for race/ethnicity

and FRL status. This finding related to race/ethnicity is similar to Rong (1996)'s findings

that African American students received lower social skills ratings than their Caucasian

peers. The finding related to gender opposes previous research in that Mashburn et al.

(2006) found that teachers rated females as displaying significantly better social skills.

Regarding competence, I found that students eligible for FRL received

significantly lower competence ratings than their peers who were not eligible for FRL

while controlling for gender and race/ethnicity. I also found that African American









particular ethnic group identified for a specific category of special education is higher

than the proportion of that group in the school-aged population, a common explanation

is that the disproportion is due to discrimination and that rates of disproportionate

representation are evidence of bias in the system (MacMillan & Reschly, 1998).

Given these concerns, the Committee on Minority Representation in Special

Education of the National Research Council (NRC) examined the rates of racial/ethnic

minority students identified as LD, MR, EBD, and gifted and found that the rates of

some racial/ethnic minority students (in particular African American students) with LD,

MR, and EBD are higher than expected based on rates of racial/ethnic minority students

in the population and that rates for gifted are lower than expected. They show that the

disproportionate rates of identification are evidence of an overrepresentation for LD,

MR, and EBD, and an underrepresentation for gifted (Donovan & Cross, 2002). The

authors noted that disproportion of racial/ethnic minorities identified with MR, LD, and

EBD has been evident since the passage of IDEA in 1975. Gelb and Mizokawa (1986)

mentioned that overrepresentation of racial/ethnic minority students is mostly apparent

in the cases of LD, MR, and EBD. Further, they refer to these categories as judgmental

because, from the authors' perspective, they are not clearly tied to biological

underpinnings and thus require a great deal of subjectivity when determining eligibility.

This concern is especially apparent in the identification of EBD as inappropriate

behavior is a main variable that initiates a referral for EBD and individuals have different

opinions about what behaviors are inappropriate or intolerable. In their review, Donovan

and Cross (2002) noted that African American students were at higher risk for being









S tu d e nt-L e ve l D ata ................................................................................ 5 8
G e n d e r ..................................................................................................................5 8
R ace/E ethnicity .................................. .......................... ........ 58
F a m ily Inco m e ................ ................................... 58
T eacher-Level D ata ................ ....................................... ...... 59
G e n d e r ..................................................................................................................5 9
Race/ethnicity ...................... ............... ......... ... .... ................... 59
A g e ....................... ........................... ...... ............................... 59
Y ears of Experience ............... .................. ........ .................................59
Self-Efficacy ................ ................................... ........ ...... .............. 60
School-Level Data ................ ................................... .............. .. ........ 60
School Size ....................... ................ ................... 60
Average Family Income .................. ........ ................ 60
Racial/Ethnic Minority Students ..................................................... 61
A n a lys es ....................... ...... ............................................... 6 1
Hierarchical Linear Modeling .................. ........................... 61
Structural Equation Modeling ............... .......... ............. .................... 63

3 RESULTS .................... ............................................... 65

D descriptive Statistics .................................................................... ....... ................. 65
Clinical Assessm ent of Behavior ................................ ......... ... ..... ........ 65
Teacher Efficacy in Classroom Management and Discipline Scale ................... 65
Random ANOVA Hierarchical Linear Models .................................................. ...... 65
Intercepts and Slopes as Outcomes Models ..... ......... ........ .. ................ 66
Internalizing .............. ............................ ................................................... 67
Externalizing ............... ............................. ..... ............... 69
S oc ia l S kills .............. ........................................................................... 7 2
Competence ................ ..... .................. ...... ..............74
Confirmatory Factor Analysis Examining Factorial Equivalence .............................75

4 D IS C U S S IO N .............. .................................................................................. 89

Research Question 1: Proportion of Variance .............................................. .. 89
Research Question 2: Mean Group Differences .......... ........................... 91
Research Question 3: Predictors of Teacher-Specific Differences in Ratings.........94
Research Question 4: Factorial Equivalence ...... ..... .............................. ....... .. 96
Limitations ................ ...... ... .. ............................... ... 97
Summary and Implications of Findings................ ............................... 99

LIS T O F R E F E R E N C E S ............. ........ .... ................. .................. ......................... 10 1

B IO G R A P H IC A L S K ET C H .................................................................................... 109







6









Table 3-6. Hierarchical linear model with internalizing as outcome: coefficients for cross-
level interaction of teacher-level variables and differences between African
American and Caucasian students
Fixed Effect Coefficient Standard T-Ratio Approx p-Value
Error df
Teacher Age, B21ab 0.093 0.108 0.869 57 0.389
Years Experience, B22ab 0.037 0.162 0.227 57 0.821
Management, B23ab -0.712 2.326 -0.306 57 0.760
External, B24ab 2.265 1.739 1.303 57 0.198
Personal, B25ab -1.099 1.931 -0.569 57 0.571
Teacher Gender, B26ab -0.744 2.303 -0.323 57 0.748
Teacher Race/Ethnicity, B27ab 1.480 2.520 0.587 57 0.559
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.

Table 3-7. Hierarchical linear model with internalizing as outcome: Coefficients for
cross-level interaction of teacher-level variables and differences between
Hispanic American and Caucasian students
Fixed Effect Coefficient Standard T-Ratio Approx p-Value
Error df
Teacher Age, B31ab -0.018 0.126 -0.145 57 0.885
Years Experience, B32ab -0.086 0.217 -0.397 57 0.693
Management, B33ab -1.593 3.156 -0.505 57 0.615
External, B34ab 1.910 2.068 0.924 57 0.360
Personal, B35ab -2.022 2.961 -0.683 57 0.497
Teacher Gender, B36ab -0.293 2.439 -0.120 57 0.905
Teacher Race/Ethnicity, B37ab 5.189 3.568 1.454 57 0.151
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.

Table 3-8. Hierarchical linear model with internalizing as outcome: Coefficients for
cross-level interaction of teacher-level variables and differences between
male and female students
Fixed Effect Coefficient Standard T-Ratio Approx df p-Value
Error
Teacher Age, B41ab -0.074 0.075 -0.987 57 0.328
Years Experience, B42ab 0.102 0.111 0.921 57 0.361
Management, B4ab 1.595 1.601 0.996 57 0.324
External, B44ab -0.779 1.178 -0.662 57 0.511
Personal, B45ab 0.107 1.436 0.074 57 0.942
Teacher Gender, B46ab 0.156 1.639 0.096 57 0.925
Teacher Race/Ethnicity, B47ab -1.184 1.989 -0.595 57 0.554
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.









TABLE OF CONTENTS


page

ACKNOW LEDG M ENTS......... .... .. ........ .... .......... ................... ... .. 4

LIS T O F T A B LE S ....................... ............................. ......... ....... ............... 7

L IS T O F F IG U R E S ..............................................................................................10

ABSTRACT ..................... .............................................. ................ 11

CHAPTER

1 INTRODUCTION ................................................. .............. ........... 13

Emotional and Behavioral Disorders............................................................. 13
Prevalence of Emotional/Behavioral Disorders ..................................................... 16
Outcomes for Students with Emotional/Behavioral Disorders ................................ 17
Disproportionate Representation of Racial/Ethnic Minority Students ................ 19
Hypotheses associated with Disproportionate Representation ............... ...............20
Commonly Cited Hypotheses ....... ................. ........... 23
Cultural mismatch hypothesis ........................ ....... .......... 23
Ogbu's (1991) oppositional culture theory .............................................. 26
Bias in teacher referral hypothesis ......................... .... ... ...... ......... 29
Summary of Research on Cultural Mismatch, Oppositional Culture, and
Teacher Bias in Referrals Hypotheses .............................. ............. ... 30
Teacher Bias Hypotheses ........................................................... 32
Early research on teacher's perceptions of Caucasian and minority
students ....................................................................... .. 33
Teacher ethnicity and behavior rating scales ............... ............................. 37
Lim stations of Current Research.............................................................. ................43
Additional Predictors of Mean-Group Differences in Behavior Ratings....................45
T teacher S elf-Efficacy ......... ......... .. ........................ ................ 45
School and D district Level V ariables ........................................... ........................ 47
Cross-Cultural Examination of Psychometric Properties .................................48
Purpose ............................... ................. 50

2 M E T H O D S ........................ .. ........... .. ........................................... 53

P a rtic ip a n ts ............... ..................................... .................... 5 3
M e as u res ....................... ............. ....... ............... ...... ........ .. ................5 5
Clinical Assessment of Behavior Teacher Form (Bracken & Keith, 2004) ........55
Teacher Efficacy in Classroom Management and Discipline Scale (Emmer
& H ic k m a n 1 9 9 1) ........... ............................. .................................... 5 7









Table 3-15. Hierarchical linear model with externalizing as outcome: Coefficients for
school-level independent variables
Fixed Effect Coefficient Standard T-Ratio Approx df p-Value
Error
School Size, G001a 0.012 0.004 2.850 7 0.025*


Percentage Minority
Students in School, G002a


0.017


0.036 0.468


Percentage Students 0.218 0.111 1.967
Eligible for FRL, G003a
aThis variable has been centered around its group mean.


Table 3-16. Hierarchical linear model with social
student-level independent variables


skills as outcome: Coefficients for


Fixed Effect Coefficient Standard T-Ratio Approx df p-Value
Error
Intercept, P0b 52.502 0.726 72.330 7 0.000
FRL Eligible Slope, -0.981 0.686 -1.431 64 0.157
Pla
African American -2.886 0.860 -3.357 57 0.002*
Slope, P2a
Hispanic American 1.380 1.134 1.217 57 0.229
Slope, P3a
Male Slope, P4a 2.679 0.602 4.449 57 0.000*
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.

Table 3-17. Hierarchical linear model with social skills as outcome: Coefficients for
teacher level independent variables
Fixed Effect Coefficient Standard T- Approx df p-Value
Error Ratio
Teacher Age, BOlab -0.072 0.038 -1.895 931 0.058
Years Experience, B02ab -0.121 0.058 2.077 931 0.038*
Management, B03ab 0.867 0.790 1.097 931 0.274
External, B04ab 1.936 0.587 3.297 931 0.001*
Personal, B05ab -1.648 0.694 -2.373 931 0.018*
Male Teacher, BO6ab -1.324 0.818 -1.618 931 0.106
Non-White Teacher, B07ab -0.522 0.982 0.531 931 0.595
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.


0.653

0.089









different racial/ethnic groups, a main question that emerges is whether there is

construct equivalence across groups or whether the scales assess the same constructs

across members from different racial/ethnic groups (Reid, 1995; Reid, Casat, Norton,

Anastopoulos, & Temple, 2001). Reid et al. (2001) identify construct equivalence as a

critical factor in cross-cultural assessment. The authors assert that if a behavior rating

scale consists of different factor structures when used across groups, then the scores

across the groups will not indicate the same construct and, therefore, cannot be

compared. In addition to tests exhibiting the same factor structure across groups, factor

loadings should also be very similar across groups (Jenson, 1980). Epstein et al. (2005)

note that if support is found for factorial invariance across groups that differences in

ratings of behavior may be more indicative of actual differences than biased

measurement.

Much of the research that has been conducted on factorial equivalence has been

limited to an evaluation of rating scales that assess ADHD (See Reid, 1995, for a

review). In addition, the limited research findings have been inconsistent. For example,

Epstein, March, Conners, and Jackson (1996) examined the factor structure of the

Conners Teacher Rating Scale across student gender and race/ethnicity. The authors

found that the main differences in factor structure related to the presence of an

antisocial factor in African American males and an inattention factor in Caucasian

females. However, the authors also stated that when conducting factor analyses

separately by gender similar factors were revealed across racial/ethnic groups. They

also found that African American students were rated higher on externalizing problems

than Caucasian students although they could not conclude whether differences were a










Table 3-20. Hierarchical linear model with social skills as outcome: Coefficients of
cross-level interaction of teacher-level variables and differences between
male and female students
Fixed Effect Coefficient Standard T-Ratio Approx df p-Value
Error
Teacher Age, B41ab -0.065 0.070 0.927 57 0.358
Years Experience, B42ab -0.072 0.104 -0.690 57 0.493
Management, B43ab 0.419 1.513 0.277 57 0.783
External, B44ab 0.690 1.112 0.620 57 0.537
Personal, B45ab 0.682 1.357 0.503 57 0.616
Teacher Gender, B46ab 1.222 1.548 0.789 57 0.433
Teacher Race/Ethnicity, 1.730 1.880 0.920 57 0.362
B47ab
aThis variable has been centered around its group mean. bThe residual parameter
variance for the parameter has been set to zero.

Table 3-21. Hierarchical linear model with social skills as outcome: Level 3 coefficients
Fixed Effect Coefficient Standard T-Ratio Approx df p-Value
Error
School Size, G001a -0.008 0.004 -1.982 7 0.087
Percentage Minority -0.003 0.036 -0.093 7 0.929
Students in School, G002a
Percentage Students -0.181 0.112 -1.606 7 0.152
Eligible for FRL, G003a
aThis variable has been centered around its group mean.









observed to behave just as "badly" as their Caucasian DTT peers. However, the African

American DDT students also displayed significantly lower reading skills, so it is likely

that teachers' referral decisions were made with the students' academic achievement in

mind. The authors noted that there were no significant differences between African

American and Caucasian teachers' rates of referral of African American and Caucasian

students for special education services.

MacMillan, Gresham, Lopez, and Bocian (1996) found that, among students who

were referred for special education evaluation, Caucasian students displayed

significantly higher verbal IQ and reading performance than African American and

Hispanic students when assessed with standardized tests. However, teachers' ratings

of students' academic competence were not significantly different across racial/ethnic

groups. The authors concluded that since mean group differences were not found,

teachers' judgments and referrals were not influenced by biases.

Summary of Research on Cultural Mismatch, Oppositional Culture, and Teacher
Bias in Referrals Hypotheses

Researchers have discussed several hypotheses to explain the disproportionate

representation of minority students identified with EBD or with special education needs

in general. Cultural mismatch theories, Ogbu's oppositional culture theory, and theories

related to biases of teachers' referrals have all been proposed to help explain why some

African American students exhibit poorer educational outcomes and are seen as

exhibiting more problem behaviors than their Caucasian peers. However, these theories

are not suitable for explaining why African American and other minority students appear

to be overrepresented in the EBD category of special education. The cultural mismatch

theories (Boykin, Tyler, & Miller, 2005; Hilliard, 1992; Ladson-Billings, 2005; Serwatka et




Full Text

PAGE 1

1 DISPROPORTIONALITY OF MINORITY STUDENTS IDENTIFIED WITH AN EMOTIONAL/BEHAVIORAL DISORDER: EXAMINING TEACHERS RATINGS OF STUDENTS BEHAVIOR AND FACTORIAL EQUIVAL ENCE OF A BEHAVIOR RATING SCALE By CHRISTINA DIANE PETERS A DISS ERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010

PAGE 2

2 2010 Christina Diane Peters

PAGE 3

3 To my mother, the most amazing woman I have ever met, my loving family, and the graduate faculty who have supported me throughout the past five years I could not have accomplished this goal without you.

PAGE 4

4 ACKNOWLEDGMENTS I would like to acknowledge several people for their help thr oughout my educational career. First and foremost, I would like to thank my graduate advisor, Dr. John Kranzler, for his endless support and facilitation of my professional development. From my first year in my graduate program, he encouraged me to engage in additional research interests above and beyond the requirements of the program, and greatly assisted me from start to finish on multiple projects. Additionally, he always reminded me to reward myself for hard work by enjoying what life has to offer. I am also grateful to have such an incredible doctoral committee and to have worked with several inspiring faculty I would like to thank Dr. Diana Joyce for helping me further develop my interest in students with emotional and behavioral difficulties and for consistently motivating and reinforcing me. I would like to thank Dr. James Algina for teaching me how to understand and utilize advanced statistical concepts, and for the countless hour s he has spent helping me clean, analyze, and interpret data. I wo uld like to thank Dr. Stephen Smith, in addition to Dr. Ann Daunic, for offering me four years of employment on the most wonderful graduate assistantship a student could ask for. They helped enhance my research and manuscript writing skills, and supported my decision to travel to many amazing places. Also, thank you for providing the data for this study. Additionally, I would like to thank Dr. Nancy Waldron for her guidance and supervision at P. K. Yonge and beyond. Finally, I would like to thank my loving and encouraging family. I would not have been able to achieve this milestone without your endless support. I am especially lucky to have my mother and grandmother as role models. To my best friend Sejal, thank you, as always, for everything you do.

PAGE 5

5 TABL E OF CONTENTS page ACKNOWLEDGMENTS ...................................................................................................... 4 LIST OF TABLES ................................................................................................................ 7 LIST OF FIGURES ............................................................................................................ 10 ABSTRACT ........................................................................................................................ 11 CHAPTER 1 INTRODUCTION ........................................................................................................ 13 Emotional and Behavioral Disorders .......................................................................... 13 Prevalence of Emotional/Behavioral Disorders ......................................................... 16 Outcomes for Students with Emotional/Behavioral Disorders .................................. 17 Disproportionate Representation of Racial/Ethnic Minority Students ....................... 19 Hypotheses associated with Disproportionate Representation ................................ 20 Commonly Cited Hypotheses .............................................................................. 23 Cultural mismatch hypothesis ....................................................................... 23 Ogbus (1991) oppositional culture theory .................................................... 26 Bias in teacher referral hypothesis ............................................................... 29 Summary of Research on Cultural Mismatch, Oppositional Culture, and Teacher Bias in Referrals Hypotheses ............................................................ 30 Teacher Bias Hypotheses .................................................................................... 32 Early research on teachers perceptions of Caucasian and minority students ...................................................................................................... 33 Teacher ethnicity and behavior rating scales ............................................... 37 Limitations of Current Research ................................................................................. 43 Additional Predictors of M ean-Group Differences in Behavior Ratings .................... 45 Teacher Self -Efficacy ........................................................................................... 45 School and District Level Variables ..................................................................... 47 Cross -Cultural Examination of Psychometric Properties .................................... 48 Purpose ....................................................................................................................... 50 2 METHODS .................................................................................................................. 53 Participants ................................................................................................................. 53 Measures .................................................................................................................... 55 Clinical Assessment of Behavior Teacher Form (Bracken & Keith, 2004) ........ 55 Teacher Efficacy in Classroom Management and Discipline Scale (Emmer & Hickman, 1991) ............................................................................................. 57

PAGE 6

6 Student Level Data ..................................................................................................... 58 Gender .................................................................................................................. 58 Race/Ethnicity ...................................................................................................... 58 Family Income ...................................................................................................... 58 Teacher Level Data .................................................................................................... 59 Gender .................................................................................................................. 59 Race/ethnicity ....................................................................................................... 59 Age ....................................................................................................................... 59 Years of Experience ............................................................................................. 59 Self -Efficacy ......................................................................................................... 60 School Level Data ...................................................................................................... 60 School Size .......................................................................................................... 60 Average Family Income ....................................................................................... 60 Racial/Ethnic Minority Students ........................................................................... 61 Analyses ...................................................................................................................... 61 Hierarchical Linear Modeling ............................................................................... 61 Structural Equation Modeling .............................................................................. 63 3 RESULTS .................................................................................................................... 65 Descriptive Statistics................................................................................................... 65 Clinical Assessment of Behavior ......................................................................... 65 Teacher Efficacy in Classroom Management and Discipline Scale ................... 65 Random ANOVA Hierarchical Linear Models ............................................................ 65 Intercepts and Slopes as Outcomes Models ............................................................. 66 Internalizing .......................................................................................................... 67 Externalizing ......................................................................................................... 69 Social Skills .......................................................................................................... 72 Competence ......................................................................................................... 74 Confirmatory Factor Analysis Examining Factorial Equivalence .............................. 75 4 DISCUSSION .............................................................................................................. 89 Research Question 1: Proportion of Variance ........................................................... 89 Research Qu estion 2: Mean Group Differences ....................................................... 91 Research Question 3: Predictors of Teacher Specific Differences in Ratings ......... 94 Research Question 4: Factorial Equivalence ............................................................ 96 Limitations ................................................................................................................... 97 Summary and Implications of Findings ...................................................................... 99 LIST OF REFERENCES ................................................................................................. 101 BIOGRAPHICAL SKETCH .............................................................................................. 109

PAGE 7

7 LIST OF TABLES Table page 2 -1 Dummy coded variables for student race/ethnicity ............................................... 64 3 -1 Descriptive statistics for the clinical assessment of behavior subscales ............. 76 3 -2 Descriptive statistics for teacher efficacy in classroom management and discipline scale ....................................................................................................... 76 3 -3 Percentage of variance at student, teacher/classroom, and school levels .......... 76 3 -4 Hierarchical linear model with internalizing as outcome: coefficients for student level independent variables ...................................................................... 77 3 -5 Hierarchical linear model with internalizing as outcome: coefficients for teacher level independent variables ...................................................................... 77 3 -6 Hierarchical linear model with internalizing as outcome: coefficients for cross level interaction of teacher level variables and differences between African American and Caucasian students ........................................................................ 78 3 -7 Hierarchical linear model with internalizing as outcome: Coefficients for cross-level interaction of teacher -level variables and differences between Hispanic American and Caucasian students ......................................................... 78 3 -8 Hierarchical linear model with internalizing as outcome: Coefficients for cross-level interaction of teacher -level variables and differences between male and female students ...................................................................................... 78 3 -9 Hierarchical lin ear model with internalizing as outcome: Coefficients for school level independent variables ....................................................................... 79 3 -10 Hierarchical linear m odel with externalizing as outcome: Coefficients for student level independent variables ...................................................................... 79 3 -11 Hierarchical linear model with externalizing as outcome: Coefficients for teacher level independent variables ...................................................................... 80 3 -12 Hierarchical linear model with externalizing as outcome: Coefficients for cross-level interaction of teacher -level variables and differences between African American and Caucasian students ........................................................... 80 3 -13 Hierarchical linear model with externalizing as outcome: Coefficients for cross-level interaction of teacher -level variables and differences between Hispanic American and Caucasian students ......................................................... 81

PAGE 8

8 3 -14 Hierarchical linear model with externalizing as outcome: Coefficients for cross-level interaction of teacher -level vari ables and differences between male and female students ...................................................................................... 81 3 -15 Hierarchical linear model with externalizing as outcome: Coefficients for school level independent variables ....................................................................... 82 3 -16 Hierarchical linear model with social skills as outcome: Coefficients for student level independent variables ...................................................................... 82 3 -17 Hierarchical linear model with social skills as outcome: Coefficients for teacher level independent variables ...................................................................... 82 3 -18 Hierarchical linear model with social skills as outcome: Coeffi cients for crosslevel interaction of teacher level variables and differences between African American and Caucasian students ........................................................................ 83 3 -19 Hierarchical linear model with social skills as outcome: Coefficients for cross level interaction of teacher level variables and differences between Hispanic American and Caucasian students ........................................................................ 83 3 -20 Hierarchical linear model with social skills as outcome: Coefficients of crosslevel interaction of teacher level variables and differences between male and female students ...................................................................................................... 84 3 -21 Hierarchical li near model with social skills as outcome: Level 3 coefficients ....... 84 3 -22 Hierarchical linear model with competence as outcom e: Coefficients for student level independent variables ...................................................................... 85 3 -23 Hierarchical linear model with competence as outcome: Coeff icients for teacher level independent variables ...................................................................... 85 3 -24 Hierarchical linear model with competence as outcome: Coefficients for cross-level interaction of teacher -level variables and differences between African American and Caucasian students ........................................................... 86 3 -25 Hierarchical linear model with competence as outcome: Coefficients for cross-level interaction of teacher -level variables and differences between Hispanic American and Caucasian students ......................................................... 86 3 -26 Hierarchical linear model with competence as outcome: Coefficients for cross-level interaction of teacher -level variables and differences between male and female stu dents ...................................................................................... 87

PAGE 9

9 3 -27 Hierarchical linear model with internalizing as outcome: Coefficients for school level independent variables ....................................................................... 87 3 -28 Summary of fit statistics for the confirmatory factor analyses by race/ethnicity ... 87

PAGE 10

10 LIST OF FIGURES Figure page 1 -1 Percentage of ESE students in racial/ethnic categories ....................................... 51 1 -2 Percentage of students eligible for E SE who received services for EBD ............ 52 3 -1 CAB factor structure. .............................................................................................. 88

PAGE 11

11 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of P hilosophy DISPROPORTIONALITY OF MINORITY STUDENTS IDENTIFIED WITH AN EMOTIONAL/BEHAVIORAL DISORDER: EXAMINING TEACHERS RATINGS OF STUDENTS BEHAVIOR AND FACTORIAL EQUIVALENCE OF A BEHAVIOR RATING SCALE By Christina Diane Peters August 2010 Chair: John H. Kranzler Major: School Psychology The disproportionate representation of racial/ethnic minority students identified with special education needs has been a topic of concern for 40 years. Disproportionate representation is the extent to which membersh ip in a given ethnic group affects the probability of being placed in a specific special education disability category (Oswald, Coutinho, Best & Singh, 1999, p.198). An assumption is that the percentage of a particular g roup identified with a special education need equals the percentage of that group in the student population. This study examined several hypotheses to better understand disproportionate representation of racial/ethnic minority students identified with an emotional/behavioral disor der in th e school system, including teacher bias. Previously researchers have examined teacher expectancy in addition to the completion of behavior rating scales. The purpose of this study was to build upon current literature focusing on the completi on of behavior rating scales. Sixty -five teachers completed the Clinical Assessment of Behavior Teacher Form (CAB-T) for a sample of 982 Caucasian, African American, and Hispanic American students. Four

PAGE 12

12 outcome variables from the CAB -T were assessed (i.e., internalizi ng, externalizing, social skills, competence). Hierarchical linear modeling techniques were used to (1) analyze variance components across three levels (2) examine mean group differences across outcome variables for student gender, race/ethnicity, and free/reduced price lunch status, (3) examine differences across school factors, and (4) examine whether teacher variables (i.e., age, years experience, gender, race/ethnicity, self -efficacy) predicted teacher -specific differences in ratings. Results indicated that a significant amount of variance in ratings was attributable to the teacher and school level s Several mean -group differences emerged (e.g., African Americans were rated by teachers as exhibiting more externalizing behaviors than their Caucasian peers). Some teacher specific differences in ratings across groups were predicted by teacher self efficacy for behavior management and teacher age, but not teacher race/ethnicity, gender, or years experience. A final purpose was to examine the psychometric pr operties of the CAB -T across racial/ethnic groups to ensure that the same factors are measured across groups. The results from Confirmatory Factor Analyses of Caucasian and African American samples indicated similar goodness of -fit indices for both groups although a poor fit of the model to the data.

PAGE 13

13 CHAPTER 1 INTRODUCTION The disproportionate representation of racial/ethnic minority students identified to receive special education services has been a topic of concern in the psychological and educational l iterature for approximately 40 years (Artiles & Trent, 1994). Despite the fact that disproportionate representation is a long-standing issue, questions remain as to whether disproportionate representation actually exists, and what causes it (Artiles & Trent, 1994; Oswald, Coutinho, Best, & Singh, 1999). Although the federal government recognizes several categories of possible placement for students in special education, the difficulties with disproportionate representation only seem to appear in three main categories (Gelb & Mizokawa, 1986).Those categories include specific learning disabilities (SLD), mental retardation (MR), and emotional and behavioral disorders (EBD). This study will focus on the possibility of disproportionate representation as it relat es to the EBD category of special education. Emotional and Behavioral Disorders The Individuals with Disabilities Education Act (IDEA) is a federal law that mandates a free and appropriate education for all eligible children and youth with disabilities in the United States (U.S. Department of Education, 2005). The law recognizes 11 categories of disabilities: intellectual disability (IND, previously mental retardation, MR), hearing impairment (including deafness), speech or language impairment, visual impa irment (including blindness), emotional/behavioral disorders (EBD), orthopedic impairment, autism, traumatic brain injury, specific learning disability (LD), deaf blindness, and multiple disabilities (U. S. Department of Education, 2006).

PAGE 14

14 IDEA defines a n emotional disturbance as a condition exhibiting one or more of the following characteristics over a long period of time and to a marked degree that adversely affects a childs educational performance: (A) an inability to learn that cannot be explained b y intellectual, sensory, or health factors, (B) an inability to build or maintain satisfactory relationships with peers and teachers, (C) inappropriate types of behavior or feelings under normal circumstances, (D) a general pervasive mood of unhappiness or depression, or (E) a tendency to develop physical symptoms or fears associated with personal or school problems. (U. S. Department of Education, 2006, p. 46756) Developed by Bower in the 1960s (e.g., Bower, 1969), the federal government adopted this defi nition and few changes have been made since. Federal definitions for all categories of special education are set as guidelines and should be followed to receive federal funding, and each state and district determines the specific eligibility criteria for e ach category and ensures that criteria are in line with the federal de finitions of the disabilities. Given that each state determines eligibility criteria for themselves, the process of identification for EBD varies across districts and states (Florian et al., 2006). The Florida Statutes and State Board of Education Rules outlines definitions and eligibility criteria for an emotional disturbance or emotional handicap in the state of Florida. Recently, Florida changed the term from emotional handicap to emot ional/behavioral disabilities (EBD) and defined a student with EBD as having persistent (is not sufficiently responsive to implemented evidenced based interventions) and consistent emotional or behavioral responses that adversely affect performance in the educational environment that cannot be attributed to age, culture, gender or ethnicity (Florida Department of Education, 20 09, p. 233). The criteria for eligibility state that a student must show difficulty maintaining adequate performance in the classroom that cannot be explained by other factors (e.g., socio-cultural, health sensory ). The Florida Statutes and State Board of Education Rules differentiate between internal

PAGE 15

15 factors (e.g., feelings of sadness, frequent crying, restlessness, loss of interest in friends or school work fears, phobias, excessive worrying) and external factors (e.g., inability to build or maintain interpersonal relationships, disruptive behaviors, noncompliance, aggression, poor social skills). A student must exhibit one or more of the above characteristics for a minimum of six months and display characteristics in at least two settings (one setting must be school) in order to receive special education services (Flori da Department of Education, 2009). Although a nation wide sy stem of identification does not exist, the major procedures are usually similar. If and when a student exhibits emotional or behavioral difficulties in the classroom, the teacher should attempt pre-referral interventions, such as a behavior management plan, to ameliorate any problem behavior. If the student continues to exhibit difficulties, the teacher refers the student to be evaluated for special education. Once referred, a school psychologist collects informal assessment data, including a file review an d interviews, and formal assessment data, including behavior rating scales. The school psychologist prepares a psychoeducational report outlining the findings of the evaluation to present to a child study team. The team consists of the students teacher, a special education teacher, the exceptional student education director, the school psychologist, and the students parents, and the team decides whether the student is eligible for special education services provided under the EBD label. Many states now utilize a response to intervention (RTI) model under which small -group or individualized evidence based intervention (s ) are provided to students with emotional and/or behavioral concerns prior to referral. In many school districts

PAGE 16

16 (e.g., Hillsborough Count y, FL), the students response to the intervention(s) is monitored. If the student does not make adequate progress in response to the intervention(s), or if it is determined that the intensity of the intervention(s) is beyond what can be maintained in the general education classroom, the student is referred for a psychological evaluation (utilizing formal data collection techniques such as behavior rating scales). Students who are referred by their teachers have a high probability of identification for spec ial education services, including those services under the EBD label (Algozzine, Christenson, & Yesseldyke, 1982). However, as several steps (described above) occur and several members are associated with the identification process, the probability of bein g identified as EBD involves many more variables than teacher referral. Prevalence of E motional/ B ehavioral D isorders Lane et al. (2008) estimate that 2 20% of students are likely to have difficulties with emotional or behavioral functioning, depending on what criteria are examined. Each year, the Department of Education collects data on the number of students receiving special education services under IDEA (U. S. Department of Education, 2005). In 2003, 9.1% of students in the U.S.A. aged 6 through 21 r eceived special education services under part B of IDEA Figure 11 displays the percentage of students receiving special education services under the label of a Specific Learning Disability (SLD), Speech or Language Impairment (S/L), Intellectual Disability (IND), Emotional/Behavioral Disorder (EBD), Other Health Impairment (OHI), and other disabilities out of all students eligible for exceptional student education services. Of students aged 6 through 21 who received special education services, 8.0% recei ved services under the label of EBD. As Figure 1

PAGE 17

17 reveals, EBD was the fourth largest category of students receiving services. The percentage of students with an identified emotional disturbance has remained unchanged since 1993 (U. S. Department of Educati on, 2005). Although the Department of Education reports the percentage of all students who receive services under some specific categories (e.g., in 2003, 4.1% of all students received services for SLD) these data are not available for EBD. From the data provided, it can be estimated that 0.73% of all students aged 6 through 21 received services for an emotional disturbance in 2003 (U. S. Department of Education, 2005). The Department of Education also provides statistics related to race/ethnicity of stud ents receiving special education services. Figure 1-2 displays the percentage of special education students in each race/ethnic category that received services under the label of EBD. The Department of Education collects data on the following racial/ethnic categories: African American (AA), American Indian/Alaskan Native (AI/AN), White (W), Hispanic (H), and Asian/Pacific Islander (A/PI). Overall, 11.2% of all African American students who were eligible for special education received services for EBD (compa red to 44.9% of African American students who received services for SLD). The statistics for other race/ethnic categories are as follows: 8.0% of American Indian/Alaskan Native students, 7.9% of White students, 4.9% of Hispanic students, and 4.6% of Asian/ Pacific Islander students. Outcomes for Students with Emotional/Behavioral Disorders Receiving special education services is generally regarded as positive for students who experience educational difficulties. For example, Wagner and Cameto (2004) documen t that students with EBD receive several types of services and supports to assist students in managing their emotional and behavioral issues at school including

PAGE 18

18 behavior management plans, behavioral interventions, mental health services, anger management t raining, substance abuse counseling, and social work services. However, researchers have identified several unintended consequences associated with being identified as EBD. Students with EBD as a group, compared to their peers, experience less school succe ss and poorer outcomes than any other group of children, with or without disabilities (Blackorby & Wagner, 1996; Bradley, Henderson, & Monroe, 2004; Landrum, Tankersley, & Kaufman, 2001). Students with EBD have lower levels of academic achievement than students in general education and special education students with SLD (Lane, Barton-Arwood, Nelson, & Wehby, 2008; Wagner & Cameto, 2004). In addition, the academic deficits of students with EBD persist and, in some cases, worsen over time (Bradley, Doolittle, & Bartolotta, 2008; Nelson, Benner, Lane, & Smith, 2004). Bradley et al. (2008) found that 75% of students with EBD were below expected grade level in reading and 97% were below expected grade level in math. Further, students with EBD are more likely th an general education and other special education students to drop out of high school. In 2003, only 35.4% of students with EBD aged 14 or older graduated from high school with a regular diploma, compared to 36.9% of students with mental retardation, 45.3% of children with multiple disabilities, and 57.4% of children with specific learning disabilities (U. S. Department of Education, 2005). Finally, students with EBD have high rates of suspension and expulsion (Bradley et al., 2008) In addition to these sequela, students with EBD also experience negative social and emotional outcomes. Students with EBD exhibit lower social and communication skills than their peers (Wagner, 1995; Wagner, Kutash, Duchnowski, Epstein, & Slum,

PAGE 19

19 2005). Parents of students with EBD report that their children have difficulty getting along with other students and teachers (Wagner & Cameto, 2004). In addition, students with EBD are more likely than other students with disabilities to be victims of bullying, bully others, get into fight s, and be suspended (Wagner & Cameto, 2004). Furthermore, students with EBD tend to display poor post -school outcomes, including high rates of unemployment and arrests and low rates of post -secondary education (Bullis & Cheney, 1999; Hayling, Cook, Gresham State, & Kern, 2008; Wagner, 1995). Wagner (1995) noted that, three to five years after exiting high school, 58% of students with EBD had been arrested at some point. Sacks and Kern (2008) examined the overall quality of life for students with EBD, including how they felt about themselves, their relationships, and their environment. They found that students with EBD experience significantly lower levels of quality of life indicators than their non-disabled peers (Sacks & Kern, 2008). It is important to note that the label of EBD does not cause these negative outcomes. However, if a student is identified with EBD and placed into a secluded classroom with peers who are more likely to experience these negative outcomes, the chance of that particular student t o experience similar outcomes is increased. Given these concerns, identification of EBD should only occur for students who are truly in need, and inappropriate identification should be avoided at all costs. Disproportionate Representation of Racial/Ethnic Minority Students Disproportionate representation is defined as the extent to which membership in a given ethnic group affects the probability of being placed in a specific special education disability category (Oswald et al., 1999, p.198). Cullinan an d Kauffman (2005) describe the disparity of some racial/ethnic groups who are identified with EBD. By comparing the percentage of students identified with EBD with the percentage of

PAGE 20

20 students in the population, they show that some racial/ethnic groups are d isproportionately represented. Using national data on the prevalence of students with EBD and the number of students in public schools in 2002 and 2003, the authors show that African American students represent 27% of students with EBD but only make up 17% of public school students, reflecting an overrepresentation of African American students identified with EBD. The trend is also noticed for Caucasian students (63% of students with EBD and 61% of public school students; Cullinan & Kaufman, 2005). The auth ors also report that some racial/ethnic groups are underrepresented, such as Asian or Pacific Islander students (1% of students with EBD, 4% of public school students) and Hispanic students (8% of students with EBD and 16% of public school students). The authors point out that the disproportionality of African American students is approximately 160% of what we would expect based on their representation in the general population (Cullinan & Kaufman, 2005). Likewise, Oswald et al. (1999) found that African Am erican students were 1.5 times as likely to be identified as EBD than their Caucasian peers. Hypotheses associated with Disproportionate Representation The idea of overrepresentation of minority students in special education was first expressed by Lloyd D unn in 1968 when he asserted that a vast majority of students identified as educable mentally retarded were from some racial/ethnic minority groups and/or families with a low socioeconomic status; groups who have been previously considered by some to be of a lower social status (MacMillan & Reschly, 1998). A widespread assumption is that the proportion of a particular ethnic group identified with special education needs should equal the proportion of that ethnic group in the school aged population, given th e assumption that all groups are equal. If the proportion of a

PAGE 21

21 particular ethnic group identified for a specific category of special education is higher than the proportion of that group in the school aged population, a common explanation is that the dispr oportion is due to discrimination and that rates of disproportionate representation are evidence of bias in the system (MacMillan & Reschly, 1998). Given these concerns, the Committee on Minority Representation in Special Education of the National Resear ch Council (NRC) examined the rates of racial/ethnic minority students identified as LD, MR EBD, and gifted and found that the rates of some racial/ethnic minority students (in particular African American students) with LD, MR, and EBD are higher than exp ected based on rates of racial/ethnic minority students in the population and that rates for gifted are lower than expected. They show that the disproportionate rates of identification are evidence of an overrepresentation for LD, MR, and EBD, and an under representation for gifted (Donovan & Cross, 2002). The authors noted that disproportion of racial/ethnic minorities identified with MR, LD, and EBD has been evident since the passage of IDEA in 1975. Gelb and Mizokawa (1986) mentioned that overrepresentati on of racial/ethnic minority students is mostly apparent in the cases of LD, MR, and EBD. Further, they refer to these categories as judgmental because, from the authors perspective, they are not clearly tied to biological underpinnings and thus require a great deal of subjectivity when determining eligibility. This concern is especially apparent in the identification of EBD as inappropriate behavior is a main variable that initiates a referral for EBD and individuals have different opinions about what beh aviors are inappropriate or intolerable. In their review, Donovan and Cross (2002) noted that African American students were at higher risk for being

PAGE 22

22 identified with EBD than Caucasian, American Indian/Alaskan Native, Asian/Pacific Islander, and Hispanic s tudents. Although several researchers (e.g., Abidin & Robinson, 2002; Bahr & Fuchs, 1991; Rong, 1996) have examined the topic of disproportionate representation of racial/ethnic minority students identified with special education needs, it is difficult t o apply one specific conceptual framework to understand why or how disproportionate representation occurs (Coutinho & Osw ald, 1998; Oswald et al., 1999). Serwatka, De ering, and Grant (1995) found that the causes most commonly cited in the literature includ e cultural differences that may lead to a predisposition of a diagnosis of an emotional handicap, a lack of uniform identification procedures, bias in the assessment instruments used in diagnosis, the attendant problems of poverty, and a general pattern of racial discrimination in society reflected in school s ystems. ( p. 493) Artile s and Trent (1994) argued that a number of variables should be considered when discussing issues related to overrepresentation. In their review of the history of overrepresentati on, they identified litigation cases, debate about systemic issues relating to definitions of special education categories, debate about biased referral and assessment procedures, the effects of socioeconomic status on educational performance, theories addressing the school failure of minori ties and others as factors that are linked with overrepresentation of minority students in special education. Additional researchers have identified similar theories and hypotheses relating to the problem of overrepresentation. Donovan and Cross (2002) hypothesized that the standards for referral and assessment of students for special education may be biased, or that they may not be applied consistently across racial/et hnic groups, although no data are presented to support this claim. Other researchers have asserted that

PAGE 23

23 Caucasian teachers may be biased against racial/ethnic minority students, or may not understand their behavior (e.g., Downey & Pribesh, 2004). Given that the proportion of minority students has been in creasing while the proportion of minority teachers has been decreasing (Donovan & Cross, 2002; Irvine, 1990), understanding teachers views of students behaviors is imperative. The focus in this literature review will be narrowed to theories and hypothes es of disproportionate representation that are most common in the literature and those that are specifically related to this study. Hypotheses relating to cultural mismatches in the classroom, Ogbus (1991) oppositional culture theory, and biases in teachers referrals are commonly cited and are discussed briefly in the section that follows. Hypotheses related to teacher bias, including early research on teacher expectancy and research using behavior rating scales, is subsequently discussed more thoroughly as the current study builds upon that literature. The limitations of current research are discussed, and additional hypotheses of mean group differences in behavior are proposed (i.e., teacher self -efficacy, school level variables, factorial equivalence). Commonly Cited Hypotheses Cultural m ismatch h ypothesis An idea that is commonly represented in the literature is that difficulties associated with academic failure of some racial/ethnic minority students and overrepresentation of some racial/ethnic minor ity students in special education are due to a cultural mismatch. The cultural mismatch may be between racial/ethnic minority students and Caucasian teachers (Hilliard, 1992; Ladson -Billings, 2005; Sleeter, 2001; Valles, 1998) or racial/ethnic minority students and the institution of education (Boykin, Tyler, & Miller, 2005). According to this hypothesis, cultural differences between

PAGE 24

24 teachers and students may lead to referral and later identification of EBD (Serwatka et al., 1995) perhaps because Caucasian teachers are biased against racial/ethnic minority students or lack an understanding of their culture. Ladson -Billings (2005) proposes that difficulties may begin before the teacher enters the classroom, because the overwhelmingly majority of Caucasian f aculty who teach preservice educators are likely very far removed from the needs of the urban, minority classroom. Similarly, Research has shown that Caucasian preservice teachers have minimal cross -cultural experience and may lack confidence regarding their ability to teach African American students (see Sleeter, 2001, for a review). Hilliard (1992) argued that there is a clear difference between the school system, which is in line with the cultural style of most Caucasian Americans, and the cultural beha vioral style of African American students. Furthermore, he stated that a misunderstanding of students cultural behavioral style may lead to inaccurate estimation of intellectual, academic, and language abilities. Hilliard additionally stated that different behavioral styles exist for African American students and, if ignored during lesson planning and instruction delivery, negative outcomes emerge (e.g., academic underachievement, perception of behaviors as being problematic). Boykin et al. (2005) found a misalignment between the culture of African American students and families in America (e.g., movement, communalism) and the mainstream culture in the classroom (e.g., individualism, competitiveness). The authors examined mostly African American students i n classrooms with African American teachers, so the proposed cultural misalignment was between the African American culture and the Caucasian American classroom or school More specifically, they proposed that the classroom promotes

PAGE 25

25 certain cultural themes (e.g., individualism, competitiveness) that are in contrast to cultural themes promoted by African American students and families. However, the difficulties seen in the classroom, including the underachievement and behavioral problems exhibited by some A frican American students, cannot be fixed just by adding more racial/ethnic minority teachers to the teaching force. Ladson-Billings (2005) stated that, if this was the case, then school districts in Detroit and Washington D. C. would be exemplary for educ ating African American students. Therefore, researchers often call for preservice teachers to be trained in methods that are useful and appropriate for majority and racial/ethnic minority students (Valles, 1998). Monroe and Obidah (2004) conducted observations and interviews with an African American teacher and concluded that, when there is cultural synchronization of teachers and students (e.g., African American teachers with African American students), there are effective styles of classroom and behavior management. Furthermore, the authors noted that more culturally responsive teaching may lead to fewer discipline referrals and less negative outcomes. A major limitation of this study was that only one classroom was examined. Although several researchers (Boykin et al., 2005; Hilliard, 1992; LadsonBillings, 2005; Monroe & Obidah, 2004; Sleeter, 2001; Valles, 1998) have discuss ed hypotheses related to a cultural mismatch, it has been difficult to draw a direct line between cultural differences and disproportionate representation of minority students identified with EBD because there are several potential confounding variables that may play a role as well. Therefore, several researchers (e.g., DeMeis & Turner, 1978; Downey & Pribesh, 2004; Pigott & Cowen, 20 00; Rong, 1996) have examined hypotheses related to a possibility of teacher bias as a main variable of interest.

PAGE 26

26 Hypotheses and research related to teacher bias are discussed later, and two remaining commonly cited hypotheses are discussed below (i.e., Ob gus Oppositional Culture Theory, Biases in Teacher Referrals). Ogbus (1991) oppositional culture theory A second hypothesis is that African American students may actually behave worse when placed with Caucasian teachers than with African American teache rs. Downey and Pribesh (2004) linked this concept to Ogbus (1991) oppositional culture theory. Ogbu (1991) distinguishes between voluntary minorities, or those minority groups who immigrated to the United States (U.S.) by choice and for hope of opportunit y, and involuntary minorities or caste like minorities who were brought to the U.S. against their will. Obgu (1991) identifies American Indians, African Americans, southwestern Mexican Americans, and Native Hawaiians as involuntary minorities or groups of people who were immigrated into the U.S. through slavery, conquest, and colonization. American Indians and Native Hawaiians clearly did not immigrate to the United States, although it can be argued that they have experienced oppression similar to what has been experienced by African Americans and Mexican Americans. When describing why some African American students tend to fail in school, Ogbu (1991) said that a history of oppression and lack of job opportunity can lead young African American students to feel hopeless about their future, and thus, to exhibit less effort than their Caucasian peers. He goes on to explain how African American students may have animosity against the school system and, in an effort to maintain their culture, will engage in behaviors thought to be consistent with African Americans and avoid acting Caucasian. This theory may help explain why African American students exhibit more problem behaviors in the classroom than Caucasian students, if

PAGE 27

27 that is the case. In addition, Ogbu (1991) explains that African American students and their families may have a lack of trust for Caucasian teachers and administrators, and so it may be that African American students do engage in more problem behaviors when placed with Caucasian teachers. Alt hough Ogbus (1991) oppositional culture theory has been frequently cited to explain behavioral differences (e.g., Ainsworth-Darnell and Downey, 1998; Downey & Pribesh, 2004), when it has been used to explain differences in the way teachers rate students behaviors, the theory has not been supported by data in the literature. Downey and Pribesh (2004) tested the theory by examining teacher ratings of kindergartners and 8th graders behaviors by compiling data from the Early Childhood Longitudinal Study (Na tional Center for Education Statistics, 2000). The authors hypothesized that if Ogbus theory was correct, teachers ratings of students behaviors should be similar for Caucasian and African American students in kindergarten but that African American stud ents behavior should be rated worse by teachers in 8th grade after African American students have had substantial amount of time in an educational setting and time to develop a learned helplessness regarding their educational potential. Downey and Pribesh (2004) created an Externalizing Problem Behaviors Scale from teachers evaluations of students externalizing behaviors (e.g., anger, aggression), and an Approaches to Lear ning Scale from teachers ratings of students prosocial behaviors. The authors concluded that, since African American students were rated as exhibiting more problematic behavior than Caucasian students as early as kindergarten, there was little support for the oppositional culture theory because kindergartners are likely too young to un derstand biases against them and thus exhibit problem behaviors

PAGE 28

28 for those reasons. The authors concluded that the mean group differences in teachers ratings of students were better explained by teacher bias since mean group differences in behavior ratings were evident as early as kindergarten (Downey & Pribesh, 2004). Additional variables that may have predicted mean group differences in behavior were not examined. Ainsworth-Darnell and Downey (1998) also examined the oppositional culture theory by testing each of 4 hypotheses associated with the theory: (a ) that involuntary minority students (e.g., African Americans) value education less because they are lack optimism about their future opportunities, (b) involuntary minority students display more resist ance to school than their immigrant minority ( e.g., Asian Americans) peers, (c ) involuntary minority students who are high achieving are looked down upon by their peers, and (d) the achievement gap noticed between involuntary minorities and their immigrant minority and Caucasian peers is due to the resistance to school experienced by the involuntary minorities. They found little support for three out of four hypotheses of the theory, but found mixed results on the fourth hypothesis. Their results found that teachers viewed African American students as putting forth significantly less effort and being significantly more disruptive than Caucasian students. However, African American students themselves reported more positive attitudes towards school than Caucas ian students, were significantly less likely to agree that it is OK to break rules, and were significantly more likely to report satisfaction from doing what is expected of them in class (Ainsworth -Darnell & Downey, 1998). Because the results of these studies do not support Ogbus (1991) oppositional culture theory, his theory is not appropriate to describe the overrepresentation of minority students identified for special education.

PAGE 29

29 Bias in teacher referral hypothesis Gerber and Semmel (1984) proposed that, instead of using psychometric criteria for identifying specific learning disabilities, mild mental retardation, and mild emotional disturbance (i.e., EBD), that teachers can be used as the tests. They argued that teachers judgments are reliable to use as indicators of which students should be identified as needing special education services and that we do not need psychometric tests. Gresham, Reschly, and Carey (1987) showed that teachers judgments may be just as accurate in identifying students with a learning disability as psychometric tests. Similarly, Shinn, Tindal, and Spira (1987) found support for the accuracy of teacher judgments for identifying students with mental handicaps. However, they also found support for gender and racial/ethnic bias in the teachers referrals. For example, the authors found that males and females who were referred had the same reading performance, but compared to the population base rates of poor readers, significantly more males were referred for reading problems by th eir teachers. Additionally, African American students were referred by their teachers at a rate that was higher than expected (Shinn et al., 1987). However, the referred African American students also performed lower than referred Caucasian students, so this particular example may not be indicative of teacher bias. Bahr and Fuchs (1991) examined whether teachers judgments about difficult -to teach (DTT) students (i.e., students who were considered DTT by their teachers) were ethnically and racially biased by examining teachers descriptions of students, teachers ratings of academics and behavior, student reading achievement, and observed classroom behavior. They found that teachers rated African American DTT students as more appropriate for referral even though the African American DT T students were

PAGE 30

30 observed to behave just as badly as their Caucasian DT T peers. However, the African American DDT students also displayed significantly lower reading skills, so it is likely that teachers referral decisions were made with the students academic achievement in mind. The authors noted that there were no significant differences between African American and Caucasian teachers rates of referral of African American and Caucasian students for special education ser vices. MacMillan, Gresham, Lopez, and Bocian (1996) found that, among students who were referred for special education evaluation, Caucasian students displayed significantly higher verbal IQ and reading performance than African American and Hispanic stud ents when assessed with standardized tests. However, teachers ratings of students academic competence were not significantly different across racial/ethnic groups. The authors concluded that since mean group differences were not found, teachers judgment s and referrals were not influenced by biases. Summary of Research on Cultural Mismatch, Oppositional Culture, and Teacher Bias in Referrals Hypotheses Researchers have discussed several hypotheses to explain the disproportionate representation of minori ty students identified with EBD or with special education needs in general. Cultural mismatch theories, Ogbus oppositional culture theory, and theories related to biases of teachers referrals have all been proposed to help explain why some African Americ an students exhibit poorer educational outcomes and are seen as exhibiting more problem behaviors than their Caucasian peers. However, these theories are not suitable for explaining why African American and other minority students appear to be overrepresented in the EBD category of special education. The cultural mismatch theories (Boykin, Tyler, & Miller, 2005; Hilliard, 1992; Ladson-Billings, 2005; Serwatka et

PAGE 31

31 al., 1995; Sleeter, 2001; Valles, 1998) are too broad in that they claim that difficulties exper ienced by African American students are due to having teachers of a different ethnicity. As Ladson-Billings (2005) pointed out, we would then expect to see little difficulties experienced by African American students with African American teachers, and this is not the case. Ogbus oppositional culture theory (Obgu, 1991) attempts explain some difficulties experienced by African American students by stating that African American students essentially rebel against the Caucasian mainstream classroom. However his theory has not been empirically supported (Ainsworth-Darnell & Downey, 1998; Downey & Pribesh, 2004). Finally, researchers have examined potential bias in teachers referrals for special education (Bahr & Fuchs, 1991; MacMillan, Gresham, Lopez, & Boc ian, 1996). While this may be a good starting point for understanding overrepresentation of racial/ethnic minority students in the EBD category, there are several steps between referral and identification (e.g., the evaluation process) that may play a role as well. More narrow hypotheses and more recently proposed h ypotheses that may be better suited for understanding disproportionate representation of some racial/ethnic minority students are discussed below. These hypotheses include the possibility of tea cher bias examined via teacher expectancy and behavior rating scales, the role of teacher self efficacy, school level variables, and psychometric properties of assessments across cultural groups. It is important to note that many researchers have also examined the role of poverty, or the interaction of race /ethnicity and poverty, in disproportionate representation of minority students in special education. However, it has been shown

PAGE 32

32 that poverty is a weak predictor of disproportionate representation in the EBD category (Skiba, Poloni -Staudinger, Simmons, Feggins -Azziz, & Chung, 2005) and the discussion of poverty contributing to disproportionate representation is beyond the scope of this study However, it has also been shown that poverty magnifies already existing racial differences in the identification of EBD (Skiba et al., 2005) and so the effects of socioeconomic status of families and children cannot be ignored. Teacher Bias Hypothese s In order to examine teacher bias as a hypothesis contributing t o overrepresentation of some racial/ethnic minority students identified with EBD, we need to understand the identification process and teachers role in that process. In many school systems the identification of EBD begins primarily with teacher referral and it is typically the students who annoy or bother their teachers the most who get referred (Kauffman, 2001). Kauffman (2001) notes that when we examine individuals from a different cultural background, our own cultural background may lead to biases in our perception of their behavior. Further, he states that almost all standards and expectations regarding behavior, and thus, judgments about them, are bound by culture. Therefore, when we ask Caucasian teachers to rate racial/ethnic minority students behav iors, their own culture may influence their ratings of those students. However, it is important to note that if teachers do engage in cultural serotype bias (e.g., biases against students from a different racial/ethnic background) it is most likely of a su btle nature as opposed to blatant discrimination (Donovan & Cross, 2002). In Irvines (1990) explanation of why some African American students are failing in school, he asserts that a cultural misunderstanding between teachers and students leads to confli ct, distrust, hostility, and potential school failure. In addition, he claims

PAGE 33

33 that, at times, some teachers may have African American and low -income students whom they do not want to teach and ma y not like. Irvine (1990) stated that the concept of cultural aversion, the tendency for teachers and administrators to avoid discussing race/ethnicity related issues, leads to a greater lack of cultural synchronization and lack of teachers understanding of African American students. He further asserted that teachers disregard components of African American students culture (e.g., beliefs, behaviors, perceptions) when they fail to acknowledge their students race/ethnicity and that they should be incorporating aspects of students culture in their approaches to tea ching. This lack of cultural understanding among teachers, or what some view as teacher bias, may lead to the overrepresentation of African American students identified with EBD. Researchers began thinking about how teachers might view racial/ethnic minor ity students differently from their Caucasian peers as early as the 1960s (e.g., Gottlieb, 1964). The early research on teachers potentially different views of students focused on teacher expectations and how those differed based on students race/ethnic ity. More recently, research has examined how teachers views of students differ more objectively through the use of behavior rating scales. Early research on teachers perceptions of Caucasian and minority students Although the impact of a prejudicial sc hool environment has been a topic of interest for researchers, it is difficult to study the nature of bias and prejudice in the schools because of its subtle nature. In earlier studies, researchers used hypothetical situations and vignettes in order to eli cit information about teachers expectations (Donovan & Cross, 2002). Pioneering the field, Gottlieb (1964) investigated how both Caucasian and African American elementary school teachers viewed their work and

PAGE 34

34 their students. He administered surveys and in terviewed 89 elementary school teachers in the Midwest who taught in schools that he classified as inner -city. He found that African American teachers were more likely than Caucasian teachers to come from similar backgrounds of inner city children and that Caucasian teachers in these schools were more critical of students and parents than African American teachers. In order to identify teachers perceptions of inner -city students, Gottlieb (1964) asked teachers to choose adjectives to describe the outstan ding characteristics of the children (p. 352) of the students they were working with from a list of 33 adjectives. Caucasian teachers selected Talkative, Lazy, Fun Loving, High Strung, and Rebellious with the highest frequency. In contrast, Afr ican American teachers selected Fun Loving, Happy, Cooperative, Energetic, and Ambitious with t he highest frequency. Gottlieb (1964) concluded that African American teachers were less critical of their students than Caucasian teachers probably be cause many of them had come from similar backgrounds and had overcome many of the same social barriers that faced their students. With the publication of Rosenthal and Jacobsons (1968) Pygmalion in the Classroom which asserted that teachers expectations about student performance contributed to a self -fulfilling prophecy and actual student performance, researchers began to study how teachers expectations of students academic achievement and behavior differed by gender and ethnicity. Cooper, Baron, and Lowe (1975) asked 128 Caucasian, female, mostly middle class undergraduate psychology and education students to provide their expectations for students given student vignettes that varied by ethnicity and socioeconomic status. Teachers indicated lower exp ectations for the

PAGE 35

35 vignette for the African American lower -class student in terms of report card grades than the vignette for Caucasian, middle-class students, which the authors claimed supported their hypothesis that expectations about a students academi c achievement are influenced by race/ethnicity and socioeconomic status (Cooper et al., 1975). However, it is difficult to determine whether the different expectations were due to bias or the fact that actual academic mean group differences exist between students of different backgrounds because limited variables were studied and the models proposed were vignettes as opposed to real student data. DeMeis and Turner (1978) evaluated the effects of students ethnicity, dialect, and attractiveness on teacher s evaluations by asking 68 Caucasian elementary school teachers to listen to a cassette tape recording of 12 different males responses to a standardized question: What happened on your favorite TV show the last time you watched it? (p. 79). The taped r esponses were paired with pictures that varied in ethnicity and attractiveness and the teachers were asked to rate the speakers personalities, quality of responses, current academic abilities, and future academic abilities. The results indicated that stud ents race/ethnicity, dialect, and level of attractiveness all influenced the teachers expectations for the students and the Caucasian students were rated higher on all 4 outcome variables than African American students (DeMeis & Turner, 1978). In a met a analysis of 77 studies on teacher expectancies, Duesk and Joseph (1983) determined that student attractiveness, conduct, cumulative folder information, ethnicity, and socioeconomic status were significantly related to teachers expectations about student s performance. Specifically regarding race/ethnicity, Duesk and Joseph

PAGE 36

36 (1983) found that 11 out of 24 studies that examined teacher expectancies as a function of students race/ethnicity (Caucasian vs. African American) revealed teacher expectancies that favored Caucasian students over African American students. Clifton, Perry, Parsonson, and Hryniuk (1986) examined teacher expectancies of students representing six ethnic groups in Canada. Students identified themselves as British, French, German, Canadian -Indian, Filipino, or Portuguese and their homeroom teachers completed a questionnaire that measured normative (i.e., social behavior) and cognitive (i.e., academic performance) expectations for their students. The researchers controlled for students soc ioeconomic status, intellectual ability (measured via Ravens Progressive Matrices ; Raven, 1947), academic performance, and students expectations of themselves. They found that the students race/ ethnicity continued to influence the teachers expectations for students as teachers indicated different expectations for the students from different racial/ethnic groups in terms of their likelihood of completing 12th grade English and Math courses and their level of cooperation and reliability (Clifton et al. 1 986). However, in this study, some of the students from different racial/ethnic groups displayed actual differences in terms of academic achievement as well (e.g., German students outperformed British students). Early researchers found powerful information related to teacher expectancy and teachers perceptions of students behaviors. While the early research laid the foundation for current work, several limitations are apparent in the early studies. Specifically, much of the research (e.g., Cooper, Baron, & Lowe, 1975; DeMeis & Turner, 1978; Gottlieb, 1964) relied on teacher interviews and assessing teachers perceptions of students based on vignettes rather than actual student data. Research

PAGE 37

37 studies examining teachers perceptions of students behaviors ut ilizing more objective methods for data collection are discussed below. Teacher ethnicity and behavior rating scales More recent research has examined how teachers rate students behaviors with the use of standardized, nationally -normed behavior rating sc ales. As previously mentioned, teachers perceptions of students behavior as expressed via behavior rating scales are important to the topic of disproportionate representation of racial/ethnic minority students identified for EBD because behavior rating scales are often used when assessing for eligibility criteria for special education. Sattler and Hoge (2006) specifically recommend behavior rating and checklist measures, in addition to other assessment techniques, when assessing for social, emotional, or behavioral difficulties because they are easy to administer and compare a students performance to a national normative sample in an attempt to provide objective data. Cohen, DuRant, and Cook (1988) examined the relationships between students age, sex, and race/ethnicity and teachers ratings of students externalizing behaviors. The authors identified 626 general and special education students and asked their teachers to complete the Conners Teacher Rating Scale (Conners, 1973) for study participants. The findings revealed that African American students who had been identified with a behavioral disorder or as mildly mentally handicapped were rated significantly worse than Caucasian students in the same category on the conduct disorders subscale. However these differences were not found for students identified with an emotional disturbance or learning disability (Cohen et al., 1988). More recent research examining students in Grades 5 through 9 found that teachers rated African American students higher t han Caucasian students on factors relating to externalizing

PAGE 38

38 behaviors (e.g., conduct problems) using the Conners Teacher Rating Scale (Epstein, March, Conners, & Jackson, 1998). Sbarra and Pianta (2001) examined teachers ratings of 540 children during kindergarten and first grade while controlling for mothers education (as a measure of socioeconomic status) and students gender. Their results revealed that African American students were rated as having worse behavior problems and less competence than thei r Caucasian peers (Sbarra & Pianta, 2001). Finally, using the Achenbach Behavior Rating Scales, Lau, Garland, Yeh, McCabe, Wood, and Hough (2004) found that teachers rated Caucasian students as having more internalizing problems and African American studen ts as having more externalizing problems, even though the students ratings of themselves indicated little variability by race/ethnicity. Downey and Pribesh (2004) mentioned that African American students may be rated as displaying worse behavior than their Caucasian peers because their actual behavior is worse. This phenomenon leads researchers to examine variables that may explain why African American students are rated less favorably. Some researchers have examined the interaction effects of both teach er and student race/ ethnicity on teachers ratings of students behaviors (Cullinan & Kauffman, 2005; Pigott & Cowen, 2000; Rayfield, 1997; Rong, 1996). The findings of th ese research studies have been mixed, with some research supporting a claim of teacher bias towards racial/ethnic minority students (because teacher ratings vary by teacher race/ethnicity) and other research that shows no differences in teacher ratings between Caucasian and racial/ethnic minority teachers Utilizing the norm sample from t he Behavioral Assessment System for Children (BASC ; Reynolds & Kamphaus, 1992 ), Rong (1996) examined the interaction of

PAGE 39

39 students and teachers gender and race/ethnicity on social behavior ratings. Participants included 984 teachers who completed the BASC on 984 African American and Caucasian students between 6and 11years of age. Using an analysis of covariance design, controlling for parental education as a measure of socioeconomic status, Rong (1996) found that both gender and race/ethnicity influenced teachers perceptions of students behaviors and that the magnitude and direction of the effects were influenced by teachers race/ethnicity. Specifically, African American teachers ratings were not significantly different for African American and Cauca sian students on all four subscales of the BASC (adaptability, social desirability, leadership, and social skills). However, a significant and large effect of students ethnicity was found for Caucasian teachers. Caucasian teachers rated African American s tudents significantly lower on the social desirability, social skills, and leadership subscales (Rong, 1996). When controlling for teacher and student gender, the differences in Caucasian teachers ratings of female students disappeared, but remained signi ficant for males. More specifically, Caucasian female teachers rated Caucasian male students as more favorably than African American male students. Significant differences in behavior ratings were not evident for females, regardless of teacher and student race/ethnicity (Rong, 1996). The sample in this study was limited, however, in that no Caucasian male teachers rated African American students so we cannot tell whether Caucasian male and female teachers rate differently. Additionally, in this study, each teacher only rated one student so an examination of how the same teacher would rate multiple students was not possible.

PAGE 40

40 In an unpublished dissertation, Rayfield (1997) examined teachers ratings of elementary school students externalizing behaviors via the Sutter -Eyeberg Student Behavior Inventory (SESBI Sutter & Eyeberg, 1984). The teacher sample consisted of 52 teachers: 44 were Caucasian, 8 were African American, 49 were female, and 3 were male. Each teacher was asked to complete the SESBI for 2 Cauca sian females, 2 Caucasian males, 2 African American females, and 2 African American males, resulting in data for 415 students (206 Caucasian students, 205 African American students, 2 Hispanic students, and 2 multi -racial students). The results indicated t hat teachers rated boys of their own race/ethnicity as exhibiting fewer behavior problems. In addition, Caucasian female students were rated as having less behavior problems when rated by Caucasian teachers. African American female students were rated simi larly by both Caucasian and African American teachers. Rayfield (1997) also examined observational data and found that African American students were observed to exhibit more behavior problems in the classroom than their Caucasian peers. Using data from t he Early Childhood Longitudinal Study Kindergarten class of 1998 1999, Downey and Pribesh (2004) examined teachers perceptions of externalizing problems and approaches to learning (a prosocial variable) for 2,707 African American and 10,282 Caucasian students who had either an African American or Caucasian teac her in Kindergarten. They found that African American students were rated as having more externalizing problems and fewer approaches to learning skills than their Caucasian peers. These differenc es held constant when teachers race/ethnicity was added into the models. The authors examined teachers ratings when the students were in 8th grade, and African American students were rated as having

PAGE 41

41 more externalizing behavior and fewer approaches to learning skills than their Caucasian peers. However, when they added teachers race/ethnicity into the model, they found that African American students with African American teachers were rated better than Caucasian students with Caucasian teachers (Downey & Pribesh, 2004). Therefore, the authors concluded that the differences in students behavior problems may not persist when students are matched with a teacher of their same race/ethnicity indicated by the finding that African American students placed with A frican American teachers were rated more favorably than Caucasian students. Other researchers have claimed that teachers perceptions of students behavior are not biased. In a study evaluating teachers perceptions of students behaviors, Pigott and Cowen (2000) asked teachers in 70 classrooms to rate two Caucasian female, two Caucasian male, two African American female, and two African American male students, in kindergarten through 5th grade, using the Teacher -Child Rating Scale (T CRS ; Hightower et al. 1986). The T -CRS measures school problem behaviors and competencies. Their findings revealed that African American students were rated as exhibiting more school adjustment problems, fewer competencies, and poorer educational prognosis than Caucasian stud ents. However, the authors noted that teachers ethnicity did not differentiate ratings of African American and Caucasian students or predict differences in ratings. In addition, Cullinan and Kauffman (2005) examined Caucasian and African American teacher s ratings of 769 students with EBD using the Scale for Assessing Emotional Disturbance (SAED ; Epstein & Cullinan, 1998 ). Their results did not support the hypothesis of teacher bias as differences between Caucasian and African American

PAGE 42

42 students on the subscales were not predicted by an interaction of student and teacher race/ethnicity. Finally, Abidin and Robinson (2002) assessed teachers perceptions of students behaviors and referral judgments of students with behavioral difficulties. The authors asked teachers to complete the Achenbach Teacher Report Form (Achenbach, 1991) and the Social Skills Rating System (Gresham & Elliot, 1990) and the results indicated that there were no statistically significant differences among students gender, race/ethnicity, SES, and age on either of the behavioral rating scales. Most studies examining the interaction effects of teacher race/ethnicity and student race/ethnicity analyzed data using multivariate analysis of variance (MANOVA) or analysis of covariance (ANCOVA). In these studies, teachers only rated one student or a few students in their classes. Little research has examined the effects when teachers rate all students in their classes. When teachers rate all students in their classes, it is possible to examine wh ether variance in meangroup differences (e.g., African American students being rated as behaving worse than Caucasian students) can be attributed to the teacher or the classroom. Mashburn, Hamre, Downer, and Pianta (2006) set out to explore within -class a nd between -class sources of variance in pre kindergarten teachers ratings of students social competence. In addition, they explored the effects of teacher variables related to professional background (e.g., years of experience), psychological characteris tics (e.g., depression, self efficacy), and characteristics of pre -K classrooms (e.g., child -teacher ratio). In order to explore these topics, the authors utilized hierarchical linear modeling (HLM, Raudenbush & Bryk, 2002). In HLM, an intraclass correlati on coefficient is a measure of how much variance is accounted for by the inclusion of a particular variable. In addition, analyses can be

PAGE 43

43 conducted in order to determine how much variance in an outcome variable lies at the student level and the classroom/t eacher level. Mashburn et al. (2006)s results indicated that 15% to 33% of the differences in teachers ratings of students behaviors can be attributed to differences between the teachers or classrooms. Given that a large portion of vari ance in behavior ratings lies at the teacher level, the next step is to examine teacher -level variables that may predict the teacher -specific differences. The authors found that teachers race/ethnicity was significantly related to three of the four student -level variable s examined in the study. Specifically, Caucasian teachers rated their students as having more behavior problems and less competence than African American teachers (Mashburn et al., 2006). Limitations of Current Research Research on teachers perceptions o f students behavior have been examined for at least 40 years and research has made it clear that African American students are viewed less favorably than Caucasian students. However, several gaps in the literature exist and, thus, many questions are still unanswered. Regarding the teacher bias hypothesis, some research has supported hypotheses of teacher bias by demonstrating differential ratings of students by Caucasian and African American teachers (Mashburn et al., 2005; Rayfield, 1997; Rong, 1996) whil e conflicting research concludes the opposite (Abidin & Robinson, 2002; Cullinan & Kauffman, 2005; Pigott & Cowen, 2000). Most of these studies are limited in that the teachers only rated one or two students, and so a complete evaluation of teacher variabl es that may affect teachers ratings is not possible. When teachers rate multiple students in their classes, it is possible to examine the proportion of mean group differences in behavior ratings that can be attributed to the teachers. Findings by Mashburn et al. (2005) indicating that 15% to

PAGE 44

44 33% of variance in teacher -rated behaviors are due to differences between teachers, not students, indicates a need for further examination of variance at the teacher/classroom level. An evaluation of teachers ratings of multiple students and use of HLM techniques would allow for a partition of variances across the student and teacher level and the ability to include predictors of teacher -specific differences in behavior ratings. Additional limitations of the previous ly discussed studies include lack of enough male and minority teachers to conduct complete analyses (Rayfield, 1997; Rong, 1996) and limited dependent variables (e.g., only measuring externalizing behaviors or social competency variables). In addition, a l arge majority of the studies focus exclusively on Caucasian and African American students. As other minority groups may be disproportionately represented in the category of EBD, it is important to study teachers perceptions of students from additional min ority groups as well. Finally, typical of research in this area, the research questions were limited to whether teachers race/ethnicity and/or gender may contribute to meangroup differences in behavior ratings (e.g., Abidin & Robinson, 2002; Cullinan & Kauffman, 2005; Mashburn et al., 2005; Pigott & Cowen, 2000 ; Rayfield, 1997; Rong, 1996). Several additional variables are of interest as potential predictors of meangroup and/or teacher -specific differences in behavior ratings. Variables at the teacher level include teacher self efficacy and years of experience teaching. Variables at the school level include school size, percentage of minority students, and the percentage of students on free or reduced price lunch (e.g, socioeconomic status of the school ). Finally, meangroup differences in behavior ratings may be more related to the internal indicators of

PAGE 45

45 test bias (i.e., the factor structure of instruments) than the teachers who complete the instruments. Additional Predictors of Mean-Group Differences in Behavior Ratings Teacher Self Efficacy A potential variable that may predict whether teachers are likely to rate students differently based on their race/ethnicity is associated with the idea of teacher self efficacy, or a belief that teachers can infl uence a students learning and behavior despite the presence of risk -factors. Theories of teacher efficacy have been based on Banduras (1977) theory of self efficacy. According to Ashton and Webb (1986), teacher self -efficacy is an important concept to fa cilitate understanding of teachers self perceived role, attitudes towards their job, and interactions with students. Two dimensions of teacher efficacy have been identified: sense of teaching efficacy and sense of personal teaching efficacy (Ashton & Webb, 1986; Dembo & Gibson, 1985; Gibson & Dembo, 1984). Teachers sense of teaching efficacy refers to general expectations that teaching can influence student learning. For example, teachers with low sense of efficacy may believe that some students are unabl e to learn and that no teacher will be able to help them. Teachers sense of personal teaching efficacy refers to a teachers evaluation of their own teaching competence. Teachers perceptions of their own teaching ability have been proposed to influence c hoices related to classroom management and instructional strategies (Ashton & Webb, 1986). Ideas related to teachers sense of efficacy can be related to biases in referral decisions and ratings of students behaviors. For example, a teacher with high self efficacy may be less likely to refer a student for special education services if they believe they can work effectively with a student exhibiting learning or behavior problems. Conversely, a teacher with low

PAGE 46

46 self -efficacy may be more likely to refer the s ame student for special education if he or she believes the students needs are beyond what the teacher can provide. This may be the case especially in situations where the student comes from a different ethnicity than the teacher and the teacher does not understand the students behaviors from a cultural perspective. Podell and Soodak (1993) asked teachers to read a case study about a student experiencing academic difficulties and to indicate the appropriateness of the students class and whether they would refer the student for special education. Six case studies were designed to vary in the students socioeconomic status (SES) and learning problems. The findings revealed that teachers with low self efficacy indicated they would refer students from a low er -SES background and teachers with high self efficacy did not differentiate referrals by SES (Podell & Soodak, 1993). Similarly, Soodak and Podell (1993) examined teacher efficacy and student problem type on teachers referral and placement decisions. The authors also used case studies describing a student with learning or behavioral difficulties. The results indicated that both special and regular educators were comfortable with the general education setting for students with difficulties when they had hi gh teaching and personal efficacy. However, those regular educators with lower personal efficacy were more likely to say they would refer students to special education (Soodak & Podell, 1993). Frey (2002) found that teachers with higher self efficacy with regards to classroom management and discipline recommended less restrictive special education placements for students with EBD than teachers with lower self efficacy. Additionally, teachers with high self efficacy have been shown to be less likely to refer difficult students to special

PAGE 47

47 education (see TschannenMoran, Hoy, & Hoy, 1998, for a review). It is possible that teachers with higher self efficacy related to classroom management and discipline may see fewer behavioral differences between Caucasian an d racial/ethnic minority students. School and District Level Variables Another hypothesis proposes that school level variables may contribute to the disproportionate representation of racial/ethnic minority students (Donovan & Cross, 2002). The authors noted that schools may have an independent influence on the academic success and behavioral problems of students that varies with the racial/ethnic composition of students in the school, or with the race or ethnicity of the individual student (Donovan & Cr oss, 2002, pp. 91). It does not appear as if anyone has examined whether variables at the school level may influence teachers ratings of students behavior. Research using HLM has shown that variance in some outcomes can be attributable to school -level fa ctors. For example, school size has been shown to be a significant predictor of students perceptions of school climate (Koth, Bradshaw, & Leaf, 2008) and post graduation employment for special education students (Schalock, Holl, Elliott, & Ross, 1992). In addition, average school SES has been shown to be a predictor of academic achievement among Hispanic children (Ryabov & Van Hook, 2007) and school SES, in addition to individual SES and student race/ethnicity, has been shown to predict academic achievemen t (Caldas & Bankston, 1998). Some researchers have examined district level data as well. For example, Serwatka et al. (1995) found that districts in which more African American students were enrolled had fewer difficulties with the overrepresentation of A frican American students identified with EBD. The authors also found that the districts that had more

PAGE 48

48 African American teachers had fewer difficulties with overrepresentation (Serwatka et al., 1995). Similarly, Oswald et al. (1999) found that districts wit h fewer African American students displayed higher probability for African American students to be identified as EBD, especially in low poverty communities. However, districts with more African American students were less likely to identify disproportionat e rates of African American students with EBD. The results of these studies suggest that school and district level factors may contribute to teachers perceptions of students behaviors. Cross -Cultural Examination of Psychometric Properties An additional hypothesis is that the assessment techniques that psychologists use may not be valid and reliable across members from different racial/ethnic groups. As previously mentioned, Donovan and Cross (2002) state that the standards for referral and assessment may be biased or may be applied differently across racial/ethnic groups. While it is difficult to conceptualize and measure their claims, we can measure variables related to the reliability and validity of the instruments we use in the evaluation process. St ated differently then, the hypothesis proposes that the tests we use may not be valid for use across different racial/ethnic groups. Jensen (1980) and others differentiate between internal criteria and external criteria when examining tests for bias across groups. Internal criteria of test bias include the psychometric qualities of the test, including the factor structure and differential item analyses. External sources of bias include those variables that are not related to the test per se, but to factors that may influence performance on the test including whether the tests are timed and the race/ethnicity and gender of the examiner (Jenson, 1980). The assessment process for EBD typically includes the use of behavior rating scales (Sattler & Hoge, 2006). When using behavior rating scales with students from

PAGE 49

49 different racial/ethnic groups, a main question that emerges is whether there is construct equivalence across groups or whether the scales assess the same constructs across members from different racial /ethnic groups (Reid, 1995; Reid, Casat, Norton, Anastopoulos, & Temple, 2001). Reid et al. (2001) identify construct equivalence as a critical factor in cross-cultural assessment. The authors assert that if a behavior rating scale consists of different fa ctor structures when used across groups, then the scores across the groups will not indicate the same construct and, therefore, cannot be compared. In addition to tests exhibiting the same factor structure across groups, factor loadings should also be very similar across groups (Jenson, 1980). Epstein et al. (2005) note that if support is found for factorial invariance across groups that differences in ratings of behavior may be more indicative of actual differences than biased measurement. Much of the res earch that has been conducted on factorial equivalence has been limited to an evaluation of rating scales that assess ADHD (See Reid, 1995, for a review). In addition, the limited research findings have been inconsistent. For example, Epstein, March, Conners, and Jackson (1996) examined the factor structure of the Conners Teacher Rating Scal e across student gender and race/ethnicity. The authors found that the main differences in factor structure related to the presence of an antisocial factor in African Am erican males and an inattention factor in Caucasian females. However, the authors also stated that when conducting factor analyses separately by gender similar factors were revealed across racial/ethnic groups. They also found that African American student s were rated higher on externalizing problems than Caucasian students although they could not conclude whether differences were a

PAGE 50

50 result of actual differences or teacher bias (Epstein et al., 1996). Conversely, Reid et al. (2001) examined factors of the IO WA Conners Rating Scale across African American and Caucasian students. Their results revealed that the same 2-factor structure was appropriate for both African American and Caucasian students. Similar to findings by Epstein et al. (1996), Reid et al. (2001) found that African American students received significantly higher ratings than Caucasian students and they found a significant teacher race/ethnicity by student race/ethnicity interaction. Finally, Walthall, Konold, and Pianta (2005) identified the fac tor structure of the social skills rating system across gender and race/ethnicity. They found factorial invariance between Caucasian and non -Caucasian groups and near factorial invariance between males and females. Purpose The purpose of the current study was to build upon the current literature by evaluating student, teacher, and school variables in order to determine whether there are differences in teacher -rated behaviors across racial/ethnic groups, and whether those differences are related to a variet y of teacher and school level variables. The use of HLM techniques were employed in order to partition the variances that may be associated with teacher ratings of students behaviors at the student, teacher, and school level. Four outcome variables from a standardized, nationally normed, representative behavior rating scale were examined including students internalizing behaviors, externalizing behaviors, competence, and social skills among students of multiple racial/ ethnic groups. Specifically, I exami ned whether meangroup differences exist in teacher -rated behaviors for internalizing, externalizing, social skills, and school competence variables across Caucasian, African American, and Hispanic students. I explored the partition of variance components across the student -, teacher and school -

PAGE 51

51 levels in order to determine whether predictors of differences should be included. Finally, I included teacher race/ethnicity and age as predictor variables to remain consistent with previous literature, and added teacher age, years of experience, and self -efficacy as predictor variables to build upon the current literature. A second purpose of this study was to examine the construct equivalence of an omnibus behavior rating scale that is likely to be used in the a ssessment of EBD. I hypothesized that mean group differences in teacher -rated behaviors may be due to a lack of factorial equivalence across racial/ethnic groups. I utilized Confirmatory Factor Analytic techniques to explore this hypothesis. Figure 11. Percentage of ESE students in racial/ethnic categories 47.4 18.7 9.6 8 7.5 8.8 0 10 20 30 40 50 SLD S/L IND EBD OHI Other Percentage of ESE StudentsCategoryPercentage of ESE Students in Each Category Percent

PAGE 52

52 Figure 12 Percentage of students eligible for ESE who received services for EBD 11.2 8 7.9 4.9 4.6 0 2 4 6 8 10 12 AA AI/AN W H A/PIPercentage of StudentsRace/EthnicityPercentage of Students Eligible for ESE who Received Services for EBD Percent

PAGE 53

53 CHAPTER 2 METHODS Data for this study were compiled from a large data set collected by the Prevention Research Te am (PRT). The PRT examined the effectiveness of a cognitivebehavioral intervention, Tools for Getting Along (see Daunic, Smith, Brank, & Penfield, 2006), and collected preand post test data on students and teachers from 2006 2009 in North Central Flor ida. Pre -test data were used in the current study to avoid any effects that TFGA may have had on the students behavior or teachers ratings. Participants The PRT recruited schools from a pool of approximately 70 elementary schools in North Central Florid a (suburban and rural districts) that contained a high percentage of students who were eligible for free or reducedprice lunch (from 6095%). Overall, 18 schools and 140 4th and 5th grade classrooms participated in the study. Data on all students in parti cipating classrooms were collected, although only data for students whose parents returned a signed informed consent form were analyzed. The rate for returned informed consent forms was approximately 70% of participating students. Each teacher completed pr e -test measures on their students over a two to three week period prior to the beginning of the intervention. Teachers knew their students for approximately 3 months prior to completing the pre test measures about their students. Teachers completed their s elf -report measures at the same time. Measures were delivered to teachers in random order to avoid order effects of teachers completing multiple surveys in the same order. For one variable of interest in this study, teacher efficacy, a smaller set of data are available. The Teacher Efficacy in Classroom Management and Discipline Scale (TECMD; Emmer & Hickman, 1991) was added

PAGE 54

54 during years two and three of the study. Therefore, two sets of data were analyzed in the Hierarchical Linear Models (described below) one set with teacher efficacy data and one set without, to ensure a larger sample size when examining all of the other student, teacher, and school variables. Participants in the complete data set without the teacher efficacy variable consisted of 1818 4th and 5th grade students, 126 t eachers, and 17 schools. When I compared the HLM models for the data set without self efficacy data and with complete data, findings regarding statistical significance across all student, teacher, and school variables remain ed significant with the smaller, more complete data set. Therefore, only results for the smaller data set are discussed. Participants in the data set with the teacher efficacy variable consisted of 982 4th and 5th grade students, 65 teachers, and 11 schools. The majority of students were Caucasian ( n = 628, 64%), followed by African Americans ( n = 209, 21.3%), Hispanic American ( n = 101, 10.3%), other ( n = 40, 4.1%) and Asian American ( n = 4, 0.4%) Given the limited sample size for students identified as other and Asian American, they were not included in the analyses. Regarding gender, 5 1.4% of the students were girls (n = 505) and 48.6% of the students were male ( n = 477). Regarding a measure commonly used as an indicator of socioeconomic status, 68.8% of students ( n = 676) were eligible for free or reduced-price lunch. The age of teachers ranged from 23 to 69 years ( M = 41.18, SD = 11.33). The majority of teachers were Caucasian ( n = 56, 86.2%) followed by African American ( n = 7, 10.8%), and Hispanic A merican (n = 2, 3.1%). Teachers were 76.9% female ( n = 50) and 23.1% male ( N = 15). Teachers had a range of years experience from 0 (new teachers) to 38 ( M = 9.28, SD = 8.15). Regarding schools, they had a range of students (K -5) from 411 to 1108 ( M = 628. 55,

PAGE 55

55 SD = 240.70). The schools had a range of 5.60% to 86.50% for percentage minority students enrolled ( M = 40.75, SD = 25.78). Finally, the schools had a range of 59.30% to 88.20% for percentage of students eligible for free or reducedpriced lunch ( M = 7 3.08, SD = 9.61). The larger data set was utilized for the confirmatory factor analyses as I w anted to include all possible data for the CAB -T. One factor analysis was conducted on a Caucasian sample that included data for 1,059 4th and 5th grade students A second factor analysis was conducted on an African American sample that included data for 569 4th and 5th grade students. Measures Clinical Assessment of Behavior Teacher Form (Bracken & Keith, 2004) The Clinical Assessment of Behavior Teacher Form (C AB-T) is an omnibus behavior rating scale that is designed to aid in the assessment of children who may need behavioral, educational, or psychiatric treatment or intervention. The CAB -T is aligned with current diagnostic criteria and includes additional cl usters that assess areas of public concern (e.g., bullying, gifted). The CAB -T contains 70 items and is nationally normed on a representative sample and can be used with students who are 5 through 18years old. The CAB -T produces scores for two Clinical s cales including the Internalizing Behaviors Scale (INT) and the Externalizing Behaviors Scale (EXT). The CAB -T further produces scores for Clinical Clusters including an Anxiety Cluster, Depression Cluster, Anger Cluster, Aggression Cluster, Bullying Clust er, and Conduct Problems Cluster. In addition, the CAB -T produces scores for two Adaptive scales including the Social Skills Scale (SOC) and the Competence Scale (COM). The CAB -T further produces scores for Adaptive Clusters including a Gifted and Talented Cluster

PAGE 56

56 and an Executive Function Cluster. Finally, the CAB -T produces scores for the following additional clusters: Learning Disability Cluster, Mental Retardation Cluster, Autistic Spectrum Behaviors Cluster, and an Attention-Deficit/Hyperactivity Clust er. Both the scale scores and the cluster scores are highly reliable and interpretable (Bracken & Keith, 2004). For the purposes of this study, the Internalizing, Externalizing, Social Skills, and Competence (a measure of adjustment and adaptive strength i n an area that is closely related to cognitive and academic functioning scales) were used as outcome variables. The CAB -T utilizes standard T scores with a mean of 50 and a standard deviation of 10. For the Internalizing and Externalizing scales, higher T scores indicate higher risk. Scores are considered to be mild clinical risk, scores between 70 and 79 are considered significant clinical risk, and scores nificant clinical risk. The Social Skills and Competence scales are adaptive scales and, as such, higher T scores indicate better adjustment. T Scores weakness, scores between 20 29 are a significant adapti ve weakness, scores between 30 and 39 are a mild adaptive weakness, scores between 40 59 are considered to be in the normal range, scores between 60 and 69 are considered a mild adaptive strength, scores between 70 and 79 are considered a significant adaptive strength, and scores Bracken and Keith (2004) suggest ed that scores on the CAB -T are highly reliable and cluster interpretations are valid for students across racial/ethnic groups. They show ed that minim al variance in both clinical and adaptive behaviors is due to

PAGE 57

57 demographic factors (e.g., race/ethnicity, gender). Further, they show ed that the CAB -T produces minimal mean score differences as a result of demographic variables. The authors examined the rel iabilities of the cluster scores across different ethnic groups and found similar reliabilities across groups on all clusters and scales and found few .97 .98 for Hispanic students). The CAB -T was selected for use in this study based on the assertion that the scores are highly reliable and valid for students across racial/ethnic groups. Teacher Efficacy in Classroom Management and Discipline Scale ( Emmer & Hickman, 1991) The Teacher Efficacy in Classroom Management and Discipline Sca le (TECMD) was developed by Emmer and Hickman in 1991. They modified a teacher efficacy scale created by Gibson and Dembo (1984) that assessed two factors: Personal Teaching Efficacy and Teaching Efficacy. Emmer and Hickman developed additional items to a ssess teacher efficacy regarding classroom management and discipline focusing on skills and capabilities. Additionally, they created items to further assess teachers belief in the strength of external factors (e.g., home life, peer influences) on students behaviors as opposed to teacher influences. They retained items from the Gibson and Dembo scale that assessed personal teaching efficacy. The TECMD therefore contains 3 subscales. The classroom management and discipline factor assesses teachers self per ception of their abilities to manage and discipline their students (e.g., I am confident of my ability to begin the year so that students will learn to behave well). The external influences factor assesses teachers belief regarding the strength of exter nal influences that are beyond the teachers control (e.g., If students arent disciplined at home, then

PAGE 58

58 they arent likely to accept it at school). The personal efficacy factor assesses teachers perception that they can make a difference in students li ves (e.g., If a student masters a new concept quickly this might be because I knew the necessary steps in teaching the concept). Emmer and Hickman (1991) show that their scale has adequate test retest reliability, internal consistency, and construct vali dity. Student -Level Data Gender Information on students gender was obtained from each students perspective school district. Gender was represented as a dummy coded variable with girls assigned a value of 0 and boys assigned a value of 1. Gender is inc luded as a student level variable in order to examine differences in teachers perceptions of students behaviors due to students gender (e.g., Clifton et al., 1986) Race/Ethnicity Information on students ethnicity was obtained from each students persp ective school district. This student -level variable is the main variable of interest. R ace/ethnicity was represented by dummy coded variable s representing Caucasian, African American, and Hispanic American students. Table 2.1 presents the dummy codes. Fam ily Income Information on whether students are eligible for free or reducedprice lunch was obtained from each students perspec tive school district and was used as a measure of family income. Family income was represented as a dummy coded variable with t hose not eligible for free or reduced-price lunch coded 0 and those students who are eligible for free or reducedprice lunch coded as 1. Family income is included as a student -level variable in order to examine differences in teachers perceptions of stud ents behaviors

PAGE 59

59 based on students family income and socioeconomic status (e.g., Dusek & Joseph, 1983). Teacher -Level Data Gender Information on teachers gender was obtained from each teachers perspective school district. Gender was represented as a du mmy coded variable with females assigned a value of 0 and males assigned a value of 1. Teacher gender is included as a variable of interest in order to examine interaction effects of teacher gender and classroom -level mean differences in behavior ratings due to the student level variables (i.e., teacher -specific differences). Race/ethnicity Information on teachers race/ethnicity was obtained from each teachers perspective school district. Ethnicity was represented as a dummy coded variable with Caucas ian teachers coded 0 and non-Caucasian teachers coded 1. Teacher race/ethnicity is included as a variable of interest in order to examine interaction effects between teacher race/ethnicity and teacher -specific differences in behavior ratings. Age Informat ion on teachers age was obtained from each teachers perspective school district. Teacher age is included as a variable of interest in order to examine interaction effects between teachers age and teacher -specific differences in behavior ratings. Years of Experience Information on teachers years of teaching experience was obtained from each teachers perspective school district. Teacher s years of experience is included as a

PAGE 60

60 variable of interest in order to examine interaction effects between teacher s years of teaching experience and teacher -specific differences in behavior ratings. Self -Efficacy Teachers self efficacy was measured via the Teacher Efficacy in Classroom Management and Discipline Scale (TECMD; Emmer & Hickman, 1991). Three components o f self efficacy including classroom management and discipline, external influences, and personal teaching efficacy were included to examine the interaction effects between teachers self efficacy and teacher -specific differences in behavior ratings. Schoo l Level Data School -level data were included to assess variance components at the school level and to determine whether differences in ratings are evident across school variables. School -level variables were not included as predictors of differences in ra tings as three-level interactions among student, teacher, and school variables become very difficult to interpret. School Size The size of each school was obtained from each school district. School size is included as a school level variable as research has indicated school size to predict other important outcomes (Koth, Bradshaw, & Leaf, 2008; Schalock, Holl, Elliott, & Ross, 1992). I was interested in whether differences in ratings varied across school size. Average Family Income The percentage of st udents eligible for free or reducedprice lunch in the school was included as a school -level variable. Average family income is included as a school -

PAGE 61

61 level variable as research has indicated various indicators of school socioeconomic status to predict other important outcomes (Caldas & Bankston, 1998; Ryabov & Van Hook, 2007). I was interested in whether differences in ratings varied across average family income. Racial/Ethnic Minority Students The percentage of racial/ethnic minority students in each school was obtained from each district and was included as a school level variable. Schools that enroll few minority students may have a different climate than schools that enroll a large percentage of minority students. Percentage of minority students in each school is included as a school level variable to examine whether differences in teacher -rated behaviors across racial/ethnic groups is related to the percentage of minority students in the school. Analyses Hierarchical Linear Modeling The purpose of thi s study was to explore the effects of students gender, ethnicity, and family income and teachers gender, ethnicity, age, years of experience, and perception of self -efficacy on teachers ratings of students clinical and adaptive behavior. An ad ditional purpose of this study wa s to examine whether variance in teachers ratings of students behaviors are related to school level factors including school size, average of family income, and percentage of minority students in each school. In this particular research design, students are nested within classrooms which are nested within schools. Therefore, hierarchical linear modeling techniques (HLM; Raudenbush & Bryk, 2002) were used to analyze the data. The use of HLM procedures allows for the estimation of variance and covariance components with nested data (i.e.,

PAGE 62

62 determining what proportion of variance in teachers ratings of students behaviors is attributable to the teachers and the schools) and for an analysis of interactions between variables at the diffe rent levels. The preliminary analysis consisted of a random effects ANOVA model, or a completely unconditional threelevel model, to determine whether additional variance exists at the teacher level (Level 2) and the school level (Level 3), and whether added teacher and school -level variables are justified. If a significant amount of variance exists at the teacher and school levels, the results indicate that variables at the teacher or school level also contribute to teacher -rated behaviors. Therefore, th e model was expanded to include student -, teacher -, and school -level variables in an intercepts and slopes as outcomes (ISAO) model. From the ISAO models, I examined the main effects of the Level 1 variables (i.e., student gender, family income, and race/e thnicity), the main effects of the Level 2 variables (i.e., teachers race/ethnicity, gender, age, years of experience, and self efficacy), and the main effects of the Level 3 variables (i.e., school size, average family income, and perc entage of minority students). I also examined the interactions between the Level 1 and Level 2 variables. In an ISAO model, significant interactions can be interpreted as a Level 2 variable predicting differences among Level 1 variables. Interactions with students gender an d race/ethnicity with teachers race/ethnicity, gender, age, years of experience, and self efficacy were examined, controlling for students family income, school size, average family income, and percentage of minority students. The procedure of examining the unconditional model and intercepts and slopes as outcomes model was carried out four times in order to evaluate the four outcome

PAGE 63

63 variables of the CAB -T -T (INT, EXT, SOC, and COM). In the models, Level 1 represents child i in class j, Level 2 represent s class j, and Level 3 represents school k. The following questions were answered. 1 What proportion of variance in teachers ratings of students behaviors is associated with the teacher/classroom and the school? 2 What are the mean differences in teachers ratings of students behaviors across students gender, race/ethnicity, and family income? 3 What are the interaction effects of teachers gender, race/ethnicity, age, years of experience, and perce ived self efficacy on teacher -specific differences for race/ ethnicity and gender in teacher -rated behaviors? Structural Equation Modeling Factor analysis techniques were employed in order to determine whether construct equivalence exists on the CAB -T for students of various racial/ethnic groups. The four -fa ctor m odel displayed in figure 31 was compared across Caucasian and African American samples. Data for Hispanic students were insufficient to perform a confirmatory factor analysis at the item level. The latent factor structure of the 70 CAB-T item scores was e xamined via CFA using the Mplus version 5.1 statistical software progr am (Muthen & Muthen, 1998 2004). The estimation method was diagonally weighted least squares with robust estimation of standard errors. This method uses data from students who have score s on all items and from students who have incomplete data and is appropriate for ordinal item scores Mean and variance adjusted goodness of fit chi -square statistics were calculated. I used the chi -square and degrees of freedom goodness of -fit test to determine goodness of -fit for t he two models. In addition, I used the following goodness of -fit indices to compare models: Bentlers comparative fit index (CFI), the Tucker -Lewis

PAGE 64

64 Index (TLI), and the Root Mean Square Error of Approximation (RMSEA) Criter ia for good fit include CFI and TLI and RMSEA Table 2 1. Dummy coded variables for student race/ethnicity Group Z Z Caucasian 0 0 African American 1 0 Hispanic American 0 1

PAGE 65

65 CHAPTER 3 RESULTS The results of this s tudy are presented in four sections. The first section presents descriptive statistics for the variables of interest in this study (i.e., CAB -T and TECMD). The second section includes Random ANOVA models for the four CAB -T outcome variables (INT, EXT, SOC, and COM). The third section includes Intercepts and Slopes as Outcomes models for the four CAB -T outcome variables. Finally, results for the confirmatory factor analyses by ethnicity are presented. A discussion of these results is presented in Chapter 4. Descriptive Statistics Clinical Assessment of Behavior Table 3 1 displays descriptive statistics for the four outcome variables measured in this study. The mean T scores for internalizing and externalizing were lower than the norm group mean of 50. The mean T scores for social skills and competence were higher than the norm group mean of 50. Teacher Efficacy in Classroom Management and Discipline Scale Table 3 2 displays descriptive statistics for the three self efficacy variables measured via the Teacher Efficacy in Classroom Management and Discipline Scale (TECMD). Teachers displayed the highest self efficacy scores for self -efficacy in behavior management. The next highest score was self efficacy for personal teaching efficacy. Finally, teachers demon strated the lowest self efficacy for external influences. Random ANOVA Hierarchical Linear Models Table 3 3 displays the results from the threelevel random ANOVA models (i.e., no predictors) for the four outcome variables in the study. The random ANOVA mo dels

PAGE 66

66 were included in order to examine the proportion of variance that lies at the student, teacher, and school levels. For teacher ratings of students internalizing behaviors, 56.71% of the variance in ratings was attributed to the student level, 34.56% to the teacher/classroom level, and 8.73% to the school level. All variance components were significant at the p behaviors, 70.73% of the variance was attributable to the student level, 22.63% to the teacher/classroom level, and 6.64% to the school level. All variance components were significant at t he p the variance was attributable to the student level, 22.24% to the teacher/classroom level, and 4.23% to the school level. The teacher/classroom level variance component was signif icant at the p was significant at p = 0.021. Finally, for teacher ratings of students competence, 79.27% of the variance was attributable to the student level, 19.97% to the teacher/classroom level, and 0.76% to the school level. The variance component for the teacher/classroom level was significant at the p for the school level was not significant ( p = 0.280). Intercepts and Slopes as Outcomes Models The IASO models were included in order to examine the main effects of the Level 1 variables (i.e., student gender, race/ethnicity, and family income), the main effects of the Level 2 variables (i.e., teachers age, years experience, self -efficacy, gender, and race/ethnicity), and the main effects of the Level 3 variables (i.e., school size, percentage minority students per school, and average family income per school). Additionally, interactions between some Level 1 and all Level 2 variables were examined for significance. Significant interactions can be interpreted as a Level 2

PAGE 67

67 variable (e.g., teacher race/ethnicity) predicting a teacher -specific difference for a Level 1 variable (e.g., student race/ethnicity). For each outcome variable, the main effects for L evel 1 variables are presented first, followed by the main effects of Level 2 variables. Interactions among all teacher variables are then presented for African American and Caucasian students, Hispanic American and Caucasian students, and male and female students. Finally, the main effects of Level 3 variables are presented. In all of the ISAO models, if the residual variance for a variable was non-significant, the variance was set to zero and the model was re-run. This method was chosen in an attempt to k eep the models efficient given the extraordinary size of the three level models with multiple variables at each level. Internalizing Table 3 4 displays the coefficients for the Level 1 variables in the hierarchical linear model with the Internalizing sca le as the outcome variable. The Internalizing intercept coefficient was 45.048. Students who were eligible for free or reducedpriced lunch were rated as more internalizing than their peers who were not eligible for free or reduced -priced lunch, although t he coefficient was not significant. African Americans were rated more internalizing than their Caucasian peers, although this was not a significant difference. Hispanic Americans were rated as significantly less internalizing than their Caucasian peers. Fi nally, males were rated as significantly less internalizing than their female peers. Table 3 5 displays the coefficients for the Level 2 variables of the hierarchical linear model with the Internalizing scale as the outcome variable. Teachers with more ye ars experience rated students as significantly less internalizing. Teachers with higher self -efficacy for behavior management also rated students as significantly less

PAGE 68

68 internalizing. Additionally, teachers with a higher belief that external influences beyo nd teachers control affect student outcomes rated students significantly less internalizing. Conversely, teachers with a higher sense of personal self efficacy rated students as significantly more internalizing. The coefficients for teacher age, gender, and race/ethnicity were not significant. Table 3 6 displays the coefficients for the cross -level interactions between teacher level variables and the contrast of African American and Caucasian students for the internalizing variable. None of the cross -leve l interactions were significant implying that the relationships between scores on the internalizing variable and teacher variables are similar for African Am erican and Caucasian students. Table 3 7 displays the coefficients for the cross -level interactions between teacher lev el variables and the contrast of Hispanic American and Caucasian students on the internalizing variable. None of the cross -level interactions were significant implying that the relationships between scores on the internalizing variabl e and teacher variables are similar for Hispanic American and Caucasian students. Stated differently, the size of the mean difference between Hispanic American and Caucasian students on the internalizing scale observed for specific teachers was not predict ed by the teacher level variables. Table 3 8 displays the coefficients for the cross -level interactions between teacher level variables and the contrast of male and female students on the internalizing variable. None of the cross -level interactions were s ignificant implying that the relationships between scores on the internalizing variable and teacher variables are similar for male and female students. Stated differently, the size of the mean difference

PAGE 69

69 between male and female students on the internalizin g scale observed for specific teachers was not predicted by the teacher -level variables. Table 3 9 displays the coefficients for the Level 3 variables in the hierarchical linear model with the Internalizing scale as the outcome variable. Larger schools had significantly higher ratings for students internalizing behaviors. The coefficients for school percentage of minority students and school percentage of students eligible for free or reducedpriced lunch were not significant. Externalizing Table 3 10 di splays the coefficients for the Level 1 variables in the hierarchical linear model with the Externalizing scale as the outcome variable. The Externalizing intercept coefficient was 46.032. Students who were eligible for free or reducedpriced lunch were ra ted as more externalizing than their peers who were not eligible for free or reduced -priced lunch, although the coefficient was not significant African Americans were rated as significantly more externalizing than their Caucasian peers. Hispanic Americans were rated as less externalizing than their Caucasian peers, although this was not a significant difference. Finally, males were rated as significantly less externalizing than their female peers. Table 3 11 displays the coefficients for the Level 2 vari ables in the hierarchical linear model with the Externalizing scale as the outcome variable. Teachers with higher self -efficacy for behavior management rated students as significantly less externalizing. Additionally, teachers with a higher belief that ext ernal influences beyond teachers control affect student outcomes rated students significantly less externalizing. Conversely, teachers with a higher sense of personal self efficacy rated students as

PAGE 70

70 significantly more externalizing. The coefficients for t eacher age, teacher years experience, gender, and race/ethnicity were not significant. Table 3 12 displays the coefficients for the cross -level interactions of the teacher lev el variables and the difference between African American and Caucasian students. There is a significant interaction for teacher self efficacy for behavior management. From T able 3.10, the mean externalizing difference between African American and Caucasian students is 2.996 points for teachers whose management self -efficacy is at the mean for their school For teachers with management self -efficacy above the mean for the ir school, the difference is smaller than 2.996. The difference can be exemplified by examining the simple slope. For teachers one standard deviation above the mean fo r management self efficacy, the simple slope can be calculated as follows: 2.996 + ( 4.419)(0.66) = 0.08. The equation shows that, for teachers who are one standard deviation above the mean for management self -efficacy, the difference between ratings for A frican American and Caucasian students on the externalizing scale is 0.08. Conversely, applying the same formula to teachers who are one standard deviation below the mean for management self efficacy [2.996 + ( -4.419)( 0.66) = 5.912], the difference between ratings for African American and Caucasian students grows to 5.912, with African American students being rated higher on externalizing behaviors than Caucasian students. Stated differently, teacher self efficacy for behavior management significantly predicted the teacher -specific difference between African American and Caucasian students. The size of the mean difference between African American and Caucasian students on the externalizing scale observed for specific teachers was not predicted by any additi onal teacher level variables.

PAGE 71

71 Table 3 13 displays the coefficients for the cross -level interactions between teacher level variables and the contrast of Hispanic American and Caucasian students on the Externalizing scale. None of the cross -level interactions were significant implying that the relationships between scores on the externalizing variable and teacher variables are similar for Hispanic American and Caucasian students. Table 3 14 displays the coefficients for the cross -level interactions of the teacher level variables and the significant difference between male and female students. There is a significant int eraction for teacher age. From T able 3.10, the mean externalizing difference between male and female students is 3.253 (females rated more ext ernalizing) for teachers who are at the mean for teacher age for their school For teachers who are above the mean for teacher age (e.g., older teachers), the difference is smaller than 3.253. Conversely, for teachers below the mean for teacher age (e.g., younger teac hers), the difference is larger than 3.253. Stated differently, age significantly predicted the teacher -specific difference between male and female students. The size of the mean difference between male and female students on the externalizing scale observed for specific teachers was not predicted by any additional teacher level variables. Table 3 15 displays the coefficients for the Level 3 variables in the hierarchical linear model with the Externalizing scale as the outcome variable. Larger schools had significantly higher ratings for students externalizing behaviors. The coefficients for school percentage of minority students and school percentage of students eligible for free or reducedpriced lunch were not significant.

PAGE 72

72 Social Skills Tab le 316 displays the coefficients for the Level 1 variables in the hierarchical linear model with the Social Skills scale as the outcome variable. The Social Skills intercept coefficient was 52.502. Students who were eligible for free or reducedpriced lunch received lower social skills ratings than their peers who were not eligible for free or reducedpriced lunch, although the coefficient was not significant African Americans received significantly lower social skills ratings than their Caucasian peers Hispanic Americans received higher social skills ratings than their Caucasian peers, although this difference was not significant. Finally, males received significantly higher social skills ratings than their female peers. Table 3 17 displays the coeffi cients for the Level 2 variables in the hierarchical linear model with the Social Skills scale as the outcome variable. Teachers with more years experience produced significantly lower social skills ratings. Teachers with a higher belief that external influences beyond teachers control affect student outcomes produced significantly higher social skills ratings. Conversely, teachers with a higher sense of personal self efficacy produced significantly lower social skills ratings. The coefficients for teacher age, self efficacy for behavior management, gender, and race/ethnicity were not significant. Table 3 18 displays the coefficients for the cross -level interactions of the teacher leve l variables and the difference between African American and Caucasian st udents on the Social Skills scale. There is a significant interaction for teacher self -efficacy for behavior management. From T able 3.16, the mean social skills difference between African American and Caucasian students is 2.886 points (African Americans rated as exhibiting fewer social skills) for teachers whose management self efficacy is at the

PAGE 73

73 mean for their school For teachers with management self efficacy above the mean for the ir school the difference is smaller than 2.886 points. For teachers wit h management self -efficacy below the mean for their school the difference is larger than 2.886 points. Stated differently, teacher self efficacy for behavior management significantly predicted the teacher -specific difference between African American and Caucasian students. The size of the mean difference between African American and Caucasian students on the social skills scale observed for specific teachers was not predicted by any additional teacher level variables. Table 3 19 displays the coefficients for the cross -level interactions between teacher level variables and the contrast of Hispanic American and Caucasian students on the Social Skills scale. None of the cross -level interactions were significant implying that the relationships between scores on the social skills variable and teacher variables are similar for Hispanic American and Caucasian students. Table 3 20 displays the coefficients for the cross -level interactions between teacher level variables and the contrast of male and female students on the social skills variable. None of the cross -level interactions were significant implying that the relationships between scores on the social skills variable and teacher variables are similar for male and female students. Table 3 21 displays the coef ficients for the Level 3 variables in the hierarchical linear model with the Social Skills scale as the outcome variable. The coefficients for school size, school percentage of minority students, and school percentage of students eligible for free or reduc edpriced lunch were not significant.

PAGE 74

74 Competence Table 3 22 displays the coefficients for the Level 1 variables in the hierarchical linear model with the Competence scale as the outcome variable. The Competence intercept coefficient was 52.914. Students w ho were eligible for free or reducedpriced lunch received significantly lower competence ratings than their peers who were not eligible for free or reducedpriced lunch. African American students received significantly lower competence ratings than their Caucasian peers. Hispanic Americans received higher social competence ratings than their Caucasian peers, although this difference was not significant. Finally, males received significantly higher competence ratings than their female peers. Table 3 23 dis plays the coefficients for the Level 2 variables in the hierarchical linear model with the Competence scale as the outcome variable. Teachers with higher self -efficacy for behavior management rated students as significantly more competent. Teachers with a higher belief that external influences beyond teachers control affect student outcomes also produced significantly higher competence ratings. Conversely, teachers with a higher sense of personal self efficacy produced significantly lower competence ratings. The coefficients for teacher age, years experience, gender, and race/ethnicity were not significant. Table 3 24 displays the coefficients for the cross -level interactions between teacher level variables and the contrast of African American and Caucasian students on the competence scale. None of the cross -level interactions were significant implying that the relationships between scores on the competence variable and teacher variables are similar for African American and Caucasian students. Stated differ ently, the size of the

PAGE 75

75 mean difference between African American and Caucasian students on the competence scale observed for specific teachers was not predicted by any teacher -level variables. Table 3 25 displays the coefficients for the cross -level interac tions between teacher level variables and the contrast of Hispanic American and Caucasian students on the competence scale. None of the cross -level interactions were significant implying that the relationships between scores on the competence variable and teacher variables are similar for Hispanic American and Caucasian students. Table 3 26 displays the coefficients for the cross -level interactions between teacher level variables and the contrast of male and female students on the competence scale. None of the cross-level interactions were significant implying that the relationships between scores on the competence variable and teacher variables are similar for male and female students. Stated differently, the size of the mean difference between male and female students on the externalizing scale observed for specific teachers was not predicted by any teacher level variables. Table 3 27 displays the coefficients for the Level 3 variables in the hierarchical linear model with the Competence scale as the outcome variable. Larger schools produced significantly lower ratings on the Competence scale. The coefficients for school percentage of minority students and school percentage of students eligible for free or reducedpriced lunch were not significant. Confirm atory Factor Analysis Examining Factorial Equivalence To examine the latent structure of the CAB by ethnicity, two sets of data were analyzed using latent variable structural equation modeling with the Mplus version 5.1 program. The CAB manual suggests that each question contributes to one of four factors: Internalizing, Externalizing, Social Skills, or Competence. Each item proposed

PAGE 76

76 in the manual to lie on one of the four factors was entered into the model. The Internalizing scale consists of 16 questions while the three remaining scales each consist of 18 questions. Figure 3-1 shows the model that was proposed for samples of Caucasian and African American students. Table 3.28 displays a summary of fit statistics from the CFA on the models run on samples o f Caucasian students and African American Students. Model 1, the model run addition, CFI = 0.651 and RMSEA = 0.215, also indicating poor fit. However, TLI = 0.949 indic ated good fit. Results for Model 2, the model run with an African American student sample, also indicated poor fit across most goodness of -fit indices. For this run with a Caucasian American sample, TLI = 0.966, indicating good fit. Table 3 1. Descriptive statistics for the clinical assessment of behavior subscales Variable N M SD Range Internalizing 974 44.80 10.57 22.0 87.0 Externalizing 973 45.47 9.86 28.0 87.0 So cial Skills 973 52.79 9.61 13.0 81.0 Competence 973 53.14 9.68 26.0 80.0 Table 3 2. Descriptive statistics for teacher efficacy in classroom management and discipline scale Variable N M SD Range Management 65 4.73 0.66 3.00 6.00 External 65 3. 04 0.64 1.71 4.50 Personal 65 4.06 0.68 2.57 5.57 Table 3 3. Percentage of variance at student, teacher/classroom, and school levels Variable Student Level Teacher/Classroom Level School Level Internalizing 56.71% 34.56% 8.73% Externalizing 70.73% 22.63% 6.64% Social Skills 73.53% 22.24% 4.23% Competence 79.27% 19.97% 0.76%

PAGE 77

77 Table 3 4. Hierarchical linear model with internalizing as outcome: coefficients for student level independent variables Fixed Effect Coefficient Standard Error T Ratio Appro x df p Value Intercept, P0 b 45.068 0.981 45.947 7 0.000 FRL Eligible Slope, P1a 1.448 0.733 1.975 64 0.052 African American Slope, P2a 0.229 0.899 0.255 57 0.800 Hispanic American Slope, P3a 3.026 1.194 2.535 57 0.014* Male Slope, P4 a 2.121 0.637 3.327 57 0.002* aThis variable has been centered around its group mean. bThe residual parameter variance for the parameter has been set to zero. Table 3 5. Hierarchical linear model with internalizing as outcome: coefficients for teacher level independent variables Fixed Effect Coefficient Standard Error T Ratio Approx df p Value Teacher Age, B01 ab 0.057 0.40 1.412 932 0.158 Teacher Year, B02 ab 0.123 0.063 1.963 932 0.050* Management, B03 ab 3.987 0.845 4.717 932 0.000* External, B04 ab 2.720 0.62 7 4.336 932 0.000* Personal, B05 ab 3.312 0.743 4.460 932 0.000* Male Teacher, B06 ab 0.442 0.872 0.507 932 0.612 Non White Teacher, B07 ab 0.034 1.049 0.032 932 0.974 aThis variable has been centered around its group mean. bThe residual parameter variance for the parameter has been set to zero.

PAGE 78

78 Table 3 6. Hierarchical linear model with internalizing as outcome: coefficients for cross level interaction of teacher level var iables and differences between African American and C aucasian students Fixed Effe ct Coefficient Standard Error T Ratio Approx df p Value Teacher Age, B21 ab 0.093 0.108 0.869 57 0.389 Years Experience, B22 ab 0.037 0.162 0.227 57 0.821 Management, B23 ab 0.712 2.326 0.306 57 0.760 External, B24 ab 2.265 1.739 1.303 57 0.198 Personal B25 ab 1.099 1.931 0.569 57 0.571 Teacher Gender, B26 ab 0.744 2.303 0.323 57 0.748 Teacher Race/Ethnicity, B27 ab 1.480 2.520 0.587 57 0.559 aThis variable has been centered around its group mean. bThe residual parameter variance for the parameter h as been set to zero. Table 3 7. Hierarchical linear model with internalizing as outcome: Coefficients for cross-level interaction of teacher -level var iables and differences between H ispanic A merican and C aucasian students Fixed Effect Coefficient Standar d Error T Ratio Approx df p Value Teacher Age, B31 ab 0.018 0.126 0.145 57 0.885 Years Experience, B32 ab 0.086 0.217 0.397 57 0.693 Management B33 ab 1.593 3.156 0.505 57 0.615 External, B34 ab 1.910 2.068 0.924 57 0.360 Personal, B35 ab 2.022 2. 961 0.683 57 0.497 Teacher Gender, B36 ab 0.293 2.439 0.120 57 0.905 Teacher Race/Ethnicity, B37 ab 5.189 3.568 1.454 57 0.151 aThis variable has been centered around its group mean. bThe residual parameter variance for the parameter has been set to zero. Table 3 8. Hierarchical linear model with internalizing as outcome: Coefficients for cross-level interaction of teacher -level variables and differences between male and female students Fixed Effect Coefficient Standard Error T Ratio Approx df p Value Teacher Age, B41 ab 0.074 0.075 0.987 57 0.328 Years Experience, B42 ab 0.102 0.111 0.921 57 0.361 Management, B4 ab 1.595 1.601 0.996 57 0.324 External, B44 ab 0.779 1.178 0.662 57 0.511 Personal, B45 ab 0.107 1.436 0.074 57 0.942 Teacher Gender, B4 6 ab 0.156 1.639 0.096 57 0.925 Teacher Race/Ethnicity, B47 ab 1.184 1.989 0.595 57 0.554 aThis variable has been centered around its group mean. bThe residual parameter variance for the parameter has been set to zero.

PAGE 79

79 Table 3 9. Hierarchical linear model with internalizing as outcome: Coefficients for school level independent variables Fixed Effect Coefficient Standard Error T Ratio Approx df P Value School Size, G001 a 0.014 0.006 2.529 7 0.039* Percentage Minority Students in School, G002a 0.002 0. 048 0.044 7 0.967 Percentage Students Eligible for FRL, G003 a 0.165 0.153 1.074 7 0.319 aThis variable has been centered around its group mean. Table 3 10. Hierarchical linear model with externalizing as outcom e: Coefficients for student level independent variables Fixed Effect Coefficient Standard Error T Ratio Approx df p Value Intercept, P0 b 46.032 0.716 64.276 7 0.000* FRL Eligible Slope, P1a 1.326 0.684 1.938 64 0.057 African American Slope, P2a 2.996 0.843 3.554 57 0.001* Hispanic American Sl ope, P3a 2.132 1.120 1.904 57 0.062 Male Slope, P4 a 3.253 0.599 5.430 57 0.000* aThis variable has been centered around its group mean. bThe residual parameter variance for the parameter has been set to zero.

PAGE 80

80 Table 3 11. Hierarchical linear model with externalizing as outcome : Coefficients for teacher level independent variables Fixed Effect Coefficient Standard Error T Ratio Approx df p Value Teacher Age, B01 ab 0.043 0.038 1.131 931 0.259 Years Experience, B02 ab 0.102 0.058 1.741 931 0.082 M anagement, B03 ab 3.601 0.789 4.560 931 0.000* External, B04 ab 2.244 0.587 3.825 931 0.000* Personal, B05 ab 2.798 0.694 4.034 931 0.000* Male Teacher, B06 ab 0.048 0.817 0.059 931 0.953 Non White Teacher, B07 ab 0.281 0.981 0.287 931 0.774 aThis vari able has been centered around its group mean. bThe residual parameter variance for the parameter has been set to zero. Table 3 12. Hierarchical linear model with externalizing as outcome : Coefficients for cross-level interaction of teacher -level variable s and differences between African A merican and C aucasian students Fixed Effect Coefficient Standard Error T Ratio Approx df p Value Teacher Age, B21 ab 0.099 0.101 0.986 57 0.329 Years Experience, B22 ab -0.007 0.152 -0.048 57 0.963 Management, B23 ab -4.4 19 2.175 -2.032 57 0.046* External, B24 ab 1.906 1.624 1.174 57 0.246 Personal, B25 ab 2.587 1.807 1.431 57 0.158 Teacher Gender, B26 ab 2.711 2.159 1.256 57 0.215 Teacher Race/Ethnicity, B27 ab 0.914 2.353 0.388 57 0.699 aThis variable has been cent ered around its group mean. bThe residual parameter variance for the parameter has been set to zero.

PAGE 81

81 Table 3 13. Hierarchical linear model with externalizing as outcome : Coefficients for cross-level interaction of teacher -level variables and difference s between H ispanic A merican and C aucasian students Fixed Effect Coefficient Standard Error T Ratio Approx df p Value Teacher Age, B31 ab 0.016 0.118 0.132 57 0.896 Years Experience, B32 ab 0.091 0.204 0.446 57 0.657 Management, B33 ab 0.838 2.964 0. 283 57 0.778 External, B34 ab 2.987 1.941 1.539 57 0.129 Personal, B35 ab 1.726 2.779 0.621 57 0.537 Teacher Gender, B36 ab 0.164 2.287 0.071 57 0.944 Teacher Race/Ethnicity, B37 ab 3.343 3.350 0.998 57 0.323 aThis variable has been centered around its group mean. bThe residual parameter variance for the parameter has been set to zero. Table 3 14. Hierarchical linear model with externalizing as outcome : Coefficients for cross-level interaction of teacher -level variables and differences between mal e and female students Fixed Effect Coefficient Standard Error T Ratio Approx df p Value Teacher Age, B41 ab 0.151 0.070 2.152 57 0.035* Years Experience, B42 ab 0.156 0.104 1.500 57 0.139 Management, B43 ab 1.723 1.505 1.145 57 0.258 External, B44 ab 0.103 1.107 0.093 57 0.927 Personal, B45 ab 2.174 1.350 1.611 57 0.112 Teacher Gender, B46 ab 0.378 1.540 0.246 57 0.807 Teacher Race/Ethnicity, B47 ab 2.510 1.870 1.343 57 0.185 aThis variable has been centered around its group mean. bThe resi dual parameter variance for the parameter has been set to zero.

PAGE 82

82 Table 3 15. Hierarchical linear model with externalizing as outcome : Coefficients for school level independent variables Fixed Effect Coefficient Standard Error T Ratio Approx df p Value S chool Size, G001 a 0.012 0.004 2.850 7 0.025* Percentage Minority Students in School, G002a 0.017 0.036 0.468 7 0.653 Percentage Students Eligible for FRL, G003 a 0.218 0.111 1.967 7 0.089 aThis variable has been centered around its group mean. Table 3 16. Hierarchical linear model with social skills as outcome: Coefficients for student level independent variables Fixed Effect Coefficient Standard Error T Ratio Approx df p Value Intercept, P0 b 52.502 0.726 72.330 7 0.000 FRL Eligible Slope, P1a 0.981 0.686 1.431 64 0.157 African American Slope, P2a 2.886 0.860 3.357 57 0.002* Hispanic American Slope, P3a 1.380 1.134 1.217 57 0.229 Male Slope, P4 a 2.679 0.602 4.449 57 0.000* aThis variable has been centered around its group m ean. bThe residual parameter variance for the parameter has been set to zero. Table 3 17. Hierarchical linear model with social skills as outcome: Coefficients for teacher level independent variables Fixed Effect Coefficient Standard Error T Ra tio Approx df p Value Teacher Age, B01 ab 0.072 0.038 1.895 931 0.058 Years Experience, B02 ab 0.121 0.058 2.077 931 0.038* Management, B03 ab 0.867 0.790 1.097 931 0.274 External, B04 ab 1.936 0.587 3.297 931 0.001* Personal, B05 ab 1.648 0.694 2.373 931 0.018* Male Teacher, B06 ab 1.324 0.818 1.618 931 0.106 Non White Teacher, B07 ab 0.522 0.982 0.531 931 0.595 aThis variable has been centered around its group mean. bThe residual parameter variance for the parameter has been set to zero.

PAGE 83

83 Table 3 18. Hierarchical linear model with social skills as outcome: Coefficients for cross-level interaction of teacher -level variables and differences between African American and Caucasian students Fixed Effect Coefficient Standard Error T -Ratio Approx df p Value Teacher Age, B21 ab -0.146 0.103 -1.424 57 0.160 Years Experience, B22 ab 0.059 0.155 0.382 57 0.704 Management, B23 ab 4.610 2.215 2.081 57 0.042* External, B24 ab 1.884 1.661 1.135 57 0.262 Personal, B25 ab -3.476 1.840 1.889 57 0. 064 Teacher Gender, B26 ab 2.189 2.199 0.995 57 0.324 Teacher Race/Ethnicity, B27 ab -1.803 2.405 -0.750 57 0.456 aThis variable has been centered around its group mean. bThe residual parameter variance for the parameter has been set to zero. Table 3 1 9. Hierarchical linear model with social skills as outcome: Coefficients for cross-level interaction of teacher -level variables and differences between Hispanic American and Caucasian students Fixed Effect Coefficient Standard Error T Ratio Approx df p Val ue Teacher Age, B31 ab 0.117 0.120 0.976 57 0.333 Years Experience, B32 ab 0.152 0.206 0.737 57 0.464 Management, B33 ab 0.272 2.995 0.091 57 0.928 External, B34 ab 1.930 1.964 0.983 57 0.330 Personal, B35 ab 0.911 2.809 0.324 57 0.747 Te acher Gender, B36 ab 0.663 2.320 0.286 57 0.776 Teacher Race/Ethnicity, B37 ab -4.869 3.403 -1.431 57 0.158 aThis variable has been centered around its group mean. bThe residual parameter variance for the parameter has been set to zero.

PAGE 84

84 Table 3 20. H ierarchical linear model with social skills as outcome: Coefficients of cross-level interaction of teacher -level variables and differences between male and female students Fixed Effect Coefficient Standard Error T -Ratio Approx df p Value Teacher Age, B41 a b -0.065 0.070 0.927 57 0.358 Years Experience, B42 ab 0.072 0.104 0.690 57 0.493 Management, B43 ab 0.419 1.513 0.277 57 0.783 External, B44 ab 0.690 1.112 0.620 57 0.537 Personal, B45 ab 0.682 1.357 0.503 57 0.616 Teacher Gender, B46 ab 1.222 1. 548 0.789 57 0.433 Teacher Race/Ethnicity, B47 ab 1.730 1.880 0.920 57 0.362 aThis variable has been centered around its group mean. bThe residual parameter variance for the parameter has been set to zero. Table 3 21. Hierarchical linear model with soci al skills as outcome: Level 3 coefficients Fixed Effect Coefficient Standard Error T Ratio Approx df p Value School Size, G001 a 0.008 0.004 1.982 7 0.087 Percentage Minority Students in School, G002a 0.003 0.036 0.093 7 0.929 Percentage Students Eligible for FRL, G003 a -0.181 0.112 -1.606 7 0.152 aThis variable has been centered around its group mean.

PAGE 85

85 Table 3 22. Hierarchical linear model with competence as outcome: Coefficients for student level independent variables Fixed Effec t Coefficient Standard Error T Ratio Approx df p Value Intercept, P0 b 52.914 0.465 113.736 7 0.000 FRL Eligible Slope, P1a 1.553 0.731 2.124 64 0.037* African American Slope, P2a 2.235 0.886 2.524 57 0.015* Hispanic American Slope, P3a 0.630 1. 196 0.526 57 0.600 Male Slope, P4 a 2.475 0.631 3.921 57 0.000* aThis variable has been centered around its group mean. bThe residual parameter variance for the parameter has been set to zero. Table 3 23. Hierarchical linear model with compet ence as outcome: Coefficients for teacher level independent variables Fixed Effect Coefficient Standard Error T Ratio Approx df p Value Teacher Age, B01 ab 0.007 0.039 0.169 931 0.866 Teacher Year, B02 ab 0.052 0.059 0.881 931 0.379 Management, B03 ab 2 .027 0.806 2.514 931 0.012* External, B04 ab 1.689 0.601 2.808 931 0.006* Personal, B05 ab 1.540 0.709 2.171 931 0.030* Male Teacher, B06 ab 0.028 0.840 0.034 931 0.973 Non White Teacher, B07 ab 0.479 1.005 0.477 931 0.633 aThis variable has been centered around its group mean. bThe residual parameter variance for the parameter has been set to zero.

PAGE 86

86 Table 3 24. Hierarchical linear model with competence as outcome: Coefficients for cross-level interaction of teacher -level variables and differences between African American and Caucasian students Fixed Effect Coefficient Standard Error T Ratio Approx df p Value Teacher Age, B21 ab -0.096 0.106 -0.908 57 0.368 Years Experience, B22 ab 0.035 0.159 0.222 57 0.825 Management, B23 ab 1.021 2.285 0.447 57 0.656 External, B24 ab 1.677 1.706 0.982 57 0.330 Personal, B25 ab 0.142 1.898 0.075 57 0.941 Teacher Gender, B26 ab 1.630 2.269 0.718 57 0.475 Teacher Race/Ethnicity, B27 ab 0.079 2.473 0.032 57 0.975 aThis variable has been centered around its group mean. bThe residual parameter variance for the parameter has been set to zero. Table 3 25. Hierarchical linear model with competence as outcome: Coefficients for cross-level interaction of teacher -level variables and differences between Hispanic American and Caucasian students Fixed Effect Coefficient Standard Error T Ratio Approx df p Value Teacher Age, B31 ab 0.156 0.127 1.225 57 0.226 Years Experience, B32 ab 0.178 0.219 0.811 57 0.421 Management, B33 ab 0.568 3.175 0.179 57 0.859 External, B34 ab 2.420 2.080 1.164 57 0.250 Personal, B35 ab 1.697 2.967 0.572 57 0.569 Teacher Gender, B36 ab 0.629 2.459 0.256 57 0.799 Teacher Race/Ethnicity, B37 ab 6.596 3.635 1.814 57 0.074 aThis variable has been centered around it s group mean. bThe residual parameter variance for the parameter has been set to zero.

PAGE 87

87 Table 3 26. Hierarchical linear model with competence as outcome: Coefficients for cross-level interaction of teacher -level variables and differences between male and female students Fixed Effect Coefficient Standard Error T Ratio Approx df p Value Teacher Age, B41 ab 0.024 0.074 0.324 57 0.747 Years Experience, B42 ab 0.021 0.109 0.191 57 0.849 Management, B43 ab 0.510 1.580 0.323 57 0.748 External, B44 ab 0.874 1.163 0.751 57 0.455 Personal, B45 ab 0.365 1.412 0.258 57 0.797 Teacher Gender, B46 ab 2.022 1.622 1.247 57 0.218 Teacher Race/Ethnicity, B47 ab 0.193 1.965 0.098 57 0.923 aThis variable has been centered around its group mean. bThe residual paramet er variance for the parameter has been set to zero. Table 3 27. Hierarchical linear model with internalizing as outcome: Coefficients for school level independent variables Fixed Effect Coefficient Standard Error T -Ratio Approx df p Value School Size, G001 a -0.008 0.003 -2.995 7 0.021* Percentage Minority Students in School, G002a 0.015 0.024 0.616 7 0.557 Percentage Students Eligible for FRL, G003 a 0.132 0.069 1.900 7 0.098 aThis variable has been centered around its group mean. Table 3 28. Summary of fit statistics for the confirmatory factor analyses by race/ethnicity Race/Ethnicity df CFI TLI RMSEA Caucasian American Studentsa 6750.983* 135 0.671 0.949 0.215 African American Studentsb 13258.348* 113 0.784 0.966 0.211 aN = 1059. bN = 569

PAGE 88

88 Figure 31. CAB factor structure. Competence Questions 2, 4, 5, 12, 15, 18, 19 24, 36, 43, 45, 50, 53, 55, 62, 67 Questions 7, 9, 13, 22, 25, 26, 28, 29, 31, 34, 38, 44, 57, 60, 61, 65, 66, 68 Social Skills Internalizing Externalizing Questions 1, 6, 10, 11, 17, 20, 30, 32, 35, 39, 47, 51, 52, 59, 63, 64, 69, 70 Questions 3 8, 14, 16, 21, 23, 27, 33, 37, 40, 41, 42, 46, 48, 49, 54, 56, 58

PAGE 89

89 CHAPTER 4 DISCUSSION The purpose of this study was to build upon the current literature by examining possible influences on the disproportionate representation of some rac ial/ethnic minority groups in the EBD category of special education. The prim ary instrument I used was the CAB-T and I examined four outcome variables: internalizing, externalizing, social s kills, and competence. First, I was interested in examining the variance components in teacher -rated behaviors at the student, teacher/clas sroom, and school levels. I was also interested in examining whether meangroup differences in teacher -rated behaviors existed across Caucasian, African American, and His panic student s. Additionally, I was interested in whether teacher -specific differences in ratings could be predicted by a host of teacher variables. Finally, I was interested in whether the CAB -T demonstrates factorial validity across race/ethnic groups. A discussion of the most relevant findings of each research question and their implications for research and practice is discussed below. Research Question 1: Proportion of Variance One goal of this study was to examine variance components in teacher -rated behaviors. When teachers are asked to complete behavior rating scales, they are asked with the intent to capture information regarding how the student behaves compared to sameage peers. While information related to student behavior is obtained, it is important to k eep in mind that the data present the teachers perception of how the student behaves. By examining variance components, researchers can attem pt to understand how much of behavior rating scale data are related to actual student

PAGE 90

90 behaviors and how much may b e attributed to other variables, such as teachers and schools. Using threelevel r andom ANOVA models, I examined the proportion of variance in teacher rated behaviors at the student -, teacher/classroom and school -level s for four outcome variables. I fo und that 56.71% of variance in internalizing behavior ratings, 70.73% of variance in externalizing behavior ratings, 73.53% of variance in social skills ratings, and 79.27% of variance in competence ratings was attributable to the student level. Additional ly, I found that 34.56% of variance in internalizing behavior ratings, 22.63% of variance in externalizing behavior ratings, 22.24% of variance in social skills ratings, and 19.97% of variance in competence ratings was attributable to the teacher/classroom level. These results are consistent with Mashburn et al. (2006)s findings that 15% to 30% of the variance in teacher rated behavior is attributable to differences in teachers or classrooms The differences between teachers or classrooms could be a vari ety of variables, five of which were examined in this study. Regardless, that 20% to 35% of variance lies at the teacher/classroom level is highly important to consider when decisions are made about students based on behavior rating scales. For example, some teachers may not be as perceptive as others on identifying behaviors associated with internalizing disorders (e.g., depression, anxiety). Conversely, some teachers may tend to rate a student globally negatively if they do not like that particular studen t. Some behavior rating scales (e.g., the BASC 2 ) have attempted to resolve this issue by including a validity scale in their results. These findings support the best practices model to obtain

PAGE 91

91 information regarding students behaviors across multiple locat ions and multiple raters (Sattler & Hodge, 2006). Finally, I also found that a significant proportion of variance in teacher -rated behaviors is attributable to the school level. Specifically, 8.73% of variance in internalizing behavior ratings, 6.64% of variance in externalizing behavior ratings, and 4.23% of var iance in social skills ratings wa s attributable to the school level. The variance component for competence ratings was not significant at the school level. These findings suggest that variables at the school level contribute to either behavior ratings or actual behavior of students in the school. Research Question 2: Mean Group Differences Another aim of this study was to examine whether mean group differences in behavior ratings exist across gender, race/ethnicity, and family income. Several studies have shown that African American students are rated as exhibiting more problematic behavior than their Caucasian peers (e.g., Cohen, DuRant, & Cook, 1988; Downy & Pribesh, 2004; Epstein, March, Conne rs, & Jackson, 1998; Sbarra & Pianta, 2001; Rayfield, 1997; Rong, 1996). Additionally, research typically shows that males receive higher ratings indicating more problem behavior than females even when controlling for race/ethnicity (e.g., Rayfield, 1997; Rong, 1997). I examined mean group differences while holding other variables constant. I found that Hispanic American students were rated as significantly less intern alizing than their Caucasian peers while controlling for gender and family income, or fr ee/reduced price lunch status (FRL status). I also found that males were rated significantly less internalizing than their female peers while controlling for race/ethnicity and FRL status. Significant differences were not found for African American student s or

PAGE 92

92 students eligible for free/reduced price lunch on internalizing behaviors. I was surprised to find that Hispanic American students were rated as significantly less internalizing than their Caucasian peers. In a study examining ratings for Hispanic and Caucasian students related to Attention Deficit/Hyperactivity Disorder, Dominguez and Shapiro (1998) found a consistent, yet non-significant trend across several measures and subscales indicating that Hispanic students received lower ratings than their Caucasian peers. They hypothesized that Hispanic families socialize children by instilling values related to obedience and rule following. Conversely, Glover, Pumariega, Holzer, Wise, and Rodriguez (1999) conducted a study in which adolescents completed se lf -report measures. Two groups in Texas were compared: one group consisting of 94% Mexican Americans and another group consisting of Mexican Americans, Caucasian Americans, and African Americans. They found that the predominately Mexican American group indicated significantly higher levels of anxiety and that students born outside of the United States also indicated significantly higher levels of anxiety than those born in the States. They hypothesized that Hispanic Americans, particularly females, who are experiencing acculturation express more feelings of internalized distress. It may be a possibility that our teachers are not picking up on some indicators of anxiety and depression in Hispanic American students and that may contribute to underrepresentation of Hispanic American youth identified with EBD (Cullinan & Kauffman, 2005). Regarding externalizing behaviors, I found that African American students were rated as significantly more externalizing than their Caucasian peers while controlling for gender and FRL status. This is consistent with previous research (e.g., Epstein, March,

PAGE 93

93 Conners, & Jackson, 1998; Sbarra & Pianta, 2001) and an important finding as I found differences even after controlling for gender and SES. I was surprised to find that mal es were rated significantly less externalizing than their female peers while controlling for race/ethnicity and FRL status. Cell sizes were relatively similar: 109 African American females, 100 African American males, 316 Caucasian females, and 316 Caucasi an males. In a recent study, Moller -Leimkuhler and Yucel (2009) examined self -ratings of depression and other variables in a college sample in Germany. They found that females displayed significantly higher ratings of depression in addition to many externalizing variables including aggression and irritability. Indeed, Moretti, Catchpole, and Odgers (2005) discuss ed trends on females involvement in aggressive behavior and call ed for additional research studying female aggression as rates have been increasin g for over 20 years. Regarding social skills, I found that African American students were rated as displaying significantly less adaptive social skills than their Caucasian peers while controlling for gender and FRL status. I also found that males were r ated as displaying significantly more adapt ive social skills than females while controlling for race/ethnicity and FRL status. This finding related to race/ethnicity is similar to Rong (1996)s findings that African American students received lower social skills ratings than their Caucasian peers. The finding related to gender opposes previous research in that Mashburn et al. (2006) found that teachers rated females as displaying significantly better social skills. Regarding competence, I found that students eligible for FRL received significantly lower competence ratings than their peers who were not eligible for FRL while controlling for gender and race/ethnicity. I also found that African American

PAGE 94

94 students received significantly lower competence ratings than their Caucasian peers while controlling for gender and FRL status. Finally, I found that males received significantly higher competence r atings than their female peers controlling for their race/ethnicity and FRL status. The link between lower compet ence ratings and FRL status is no surprise. Barbarin et al. (2006) found that students from poor families displayed significantly lower receptive language and math scores than their peers. Additionally, they found that students who were not from poor famil ies scored higher on all preacademic tasks when assessed at the beginning of pre kindergarten. Finally, larger schools were associated with higher rati ngs of internalizing behaviors, higher rati ngs of externalizing behaviors and lower ratings of competence. These differences were small but significant. School size was not significantly related to ratings of social skills, and percentage of minority students and percentage of students eligible for FRL in the school were not significantly related to any of the four outcome variables. Research Question 3: Predictors of Teacher -Specific Differences in Ratings Although Hispanic American students and males were rated as significantly less internalizing than their peers, none of the teacher variables (age, yea rs experience, self efficacy, race/ethnicity, or gender) significantly predicted the size of the teacher -specific differences for these groups Considering that approximately 35% of variance in teacher -rated internalizing problems lies at the teacher/class room level and approximately 9% at the school level, additional research is needed to explore additional potential predictors of teacher -specific mean differences in internalizing behavior ratings. Some t eachers may not be aware of some of the less -noticeable

PAGE 95

95 behaviors exhibited by students with internalizing disorders (e.g., withdrawal, irritability, nervousness). Regarding the externalizing variable, African American students were rated as exhibiting significantly more externalizing behaviors. Teache r age, years experience, gender, race/ethnicity, and feelings of self efficacy related to external influences and personal efficacy did not significantly predict the teacher -specific differences for the race/ethnicity variables in this study. However, I fo und that teacher self efficacy related to classroom management and discipline significantly predicts the teacher -specific differences in ratings for Caucasian and African American students. As teachers sense of self efficacy regarding their behavior manag ement and classroom disci pline skills increased, the teacher -specific difference between African American students and Caucasian students decreased. However, as teachers self efficacy decreases, teacher specific differences in ratings increase. There are two possible reasons for this finding. One hypothesis is that teachers with higher self efficacy in behavior management skills actually employ better skills in the classroom, and therefore, their students may exhibit less behavior problems. A second hypoth esis is that teachers with higher self -efficacy skills relating to behavior management do not perceive behaviors exhibited by African American students as being as problematic as teachers with a lower sense of self efficacy. In either case, the implications for practice imply a greater need for preservice and inservice training for teachers regarding classroom management and discipline. Further research in this area is warranted as well. I also found that males were rated as less externalizing than female s in this study. Teacher age, but no other variables, significantly predicted the teacher -specific

PAGE 96

96 gender differences observed in this study As teachers age increased, the teacher specific gender difference in ratings grew smaller. This may be explained by recent research discussing an increasing trend in externalizing behaviors for females (Moller Leimkuhler & Yucel, 2009; Moretti, Catchpole, & Odgers, 2005). It may be that older teachers still perceive females as less externalizing than males while youn ger teachers are more perceptive to changing gender differences. This is another area where additional research is needed. Finally, I found that African Americans received significantly lower social skills ratings than their Caucasian peers. The only tea ch er variable that predicted a teacher specific difference in social skills ratings across these groups was self -efficacy in clas sroom management and discipline. This finding suggests that, as teacher self efficacy for classroom management and discipline i ncreases, the size of the teacher specific difference in social skills ratings decreases. However, as teacher self efficacy for classroom management and discipline decreases, the size of the teacher -specific difference in social skills ratings increases. M ashburn et al. (2006) found that teachers self -efficacy ratings were positively associated with their reports of students social competence as well. Additional research on the relationship between teacher self efficacy in classroom management and discipl ine and perceptions of students social skills is needed. Research Question 4: Factorial Equivalence A final aim of this study was to examine the factorial equivalence of the CAB -T across racial/ethnic groups. My hypothesis was that meangroup differenc es in ratings on behavior rating scales may be related to the fact that different factors are measured across di fferent racial/ethnic groups. I conducted a confirmatory factor analysis on a

PAGE 97

97 sample of Caucasian students data and a sample of Af rican America n students data. My findings suggested that models for both groups of students did not fit well according to three indicators of goodness of -, CFI, and RMSEA). Findings for the TLI indicated adequate fit for both groups. Findings for each indicator of goodness of -fit were very similar, suggesting a possibility that we are measuring the same factors across Caucasian and African American students. However, as my results indicated poor fit of the data to the model proposed by the CAB -T, I cannot conclude factorial invariance until strong factors emerge. E xploratory factor analyses are recommended to identify revisions to the model proposed by the authors. Limitations First, this sample was limited to 4th and 5th grade students. The restricti on of range of this sample limits the ability to generalize my findings to younger and older s tudents. Additionally, all of these data were collected from suburban and rural districts from a large state in the southeast, further limiti ng the ability to gen eralize these findings to other regions of the United States. Similar to many other studies, the teachers in my sample were predominately Caucasian females. Of my 65 teachers, 43 were Caucasian females, 13 were Caucasian males, 6 were African American females, one was an African American male, one was an Asian American female, and one was an Asian American male. I did not have any Hispanic American teachers in my study. Due to the limited number of racial/ethnic minority teachers, I analyzed the data comp aring Caucasian teachers ratings to non-Cauc asian teachers ratings. That I only had 9 racial /ethnic minority teachers in my sample may have limited the power to de tect significant findings in this study. Howe ver, as previously mentioned, I ran the same H LM models with a larger

PAGE 98

98 data set not including data on teacher self efficacy. The larger data set included 127 teachers: 104 Caucasian teachers and 23 non-Ca ucasian teachers. I found the same results regarding significance in all of the models, suggesting that even with a larger sample size teacher gender and race/ethnicity did not predict meangroup differences in behavior ratings. A major limitation of this study is that I did not have observational data of students in their classrooms. The discussion and findings in this study are all related to teachers completion of a behavior rating scale. Even with the finding that teacher self efficacy predicts teacher -specific difference s in externalizing ratings for African American and Caucasian students, quest ions still remain as to whether African American students exhibit fewer problem behaviors when placed with teachers with higher self -efficacy or if the teachers perceive problem behaviors as less problematic. Research comparing teachers ratings of student s behaviors in an HLM model with observational data of students behaviors would provide much needed information to help answer the question of whether differences in behavior ratings are evidence of real group differences in behavior or a matter of teach ers perceptions. Finally, this study was limited in regards to racial/ethnic minority groups included. I did not have enough data available for Hispanic students to conduct a confirmatory factor analysis at the i tem level on the CAB -T to explore factor ial equivalence. However, the CAB -T manual provides internal consistency ratings across racial/ethnic groups and results suggest good internal consistency for Hispanic American students. Furthermore, this entire sample was limited to Caucasian, African Ame rican, and Hispanic American students. While I did have some students who were of Asian or Pacific Islander decent

PAGE 99

99 or wh o indicated mixed ethnicity, the numbers were not large enough to include them as a sub-group. Additional research examining ratings of additional racial/ethnic groups is needed. Summary and Implications of Findings This study produced several interesting findings. First and foremost, I found that a significant amount of variance in teacher rated behaviors lies at the teacher/classroom l evel (20% to 35%) and at the school level (5% to 9%). It is imperative to keep this in mind when evaluating student data and making decisions about students educational placement and provision of services. It is always recommended to gather data from mult iple sources and compare behavior ratings across raters and environments (Sattler & Hodge, 2006). I found that Hispanic American students were rated significantly less internalizing than their Caucasian peers and that no teacher variables in this study predicted the teacher -specific difference s regarding internalizing behaviors Further research could examine whether Hispanic students really do experience less internalizing problems or whether teachers do not pick up on symptoms exhibited by Hispanic students. I found that African American students were rated as significantly more externalizing than their Caucasian peers and that teacher self efficacy related to classroom management and discipline predicted the teacher -specific difference s This is perhaps t he most important finding of this study, as I did not find that teacher race/ethnicity or gender predicted teacher -specific differences related to externalizing behaviors as others have found related to mean group differences (e.g., Mashburn et al., 2006; Rayfield, 1997; Rong, 2006). Teacher race/ethnicity is a variable that cannot be controlled, while teacher self -efficacy related to classroom management and discipline is something that can be enhanced through preservice and inservice training.

PAGE 100

100 I also found that females were rated as significantly more externalizing than males an d that teacher age predicts the teacher -specific difference s related to externalizing behaviors across gender Research focusing on increasing rates of externalizing behavior amon g females is needed before suggestions can be made regarding inservice training for teachers regarding these concerns. Additionally, I found that African American students received significantly lower social skills ratings than their Caucasian peers and th at teacher self efficacy in classroom management and discipline predicted the teacher -specific difference s This finding is similar to the finding that teacher self efficacy predicted the teacher -specific differences for African American and Caucasian students on externalizing behavior. I found several mean group differences for school competence: students eligible for FRL, African American students, and females received significantly lower competence ratings. No teacher variables included in this study significantly predicted any teacher -specific differences related to these variables This finding may indicate that there are real group differences in school competence. However, as 19% of the var iance in competence ratings was attributable to the teacher/cl assroom level, additional research examining possible predictors is warranted. Finally, I found that goodness of -fit indices for Caucasian and African American student s rated via the CAB -T were similar, suggesting that we may be measuring the same factors across both groups However, the results from this factor analyses indicated a poor fit of the model to the data, indicating that additional research is needed on psychometric properties of the CAB -T.

PAGE 101

101 LIST OF REFERENCES Abidin, R. R. & Robinson, L. L. (2002). Stress, biases, or professionalism: What drives teachers referral judgments of students with challenging behaviors? Journal of Emotional and Behavioral Disorders, 10, 204 212. Achenbach, T. M. (1991). Manual for the Teachers Report Form and 1991 Profile. Burlington: University of Vermont, Department of Psychiatry. Ainsworth-Darnell, J. W. & Downey, D. B. (1998). Assessing the oppositional culture explanation for racial/ethnic differences in school performance. American Sociological Review, 63, 5 36 553. Algozzine, B., Christenson, S., & Yesseldyke, J. E. (1982). Probabilities associated with the referral to placement process. Teacher Education and Special Education, 5, 19 23. Artiles, A. J. & Trent, S. C. (1994). Overrepresentation of minority students in special education: A continuing debate. The Journal of Special Education, 27, 410 437. Ashton, P. T. & Webb, R. B. (1986). Making a difference: Teachers sense of efficacy and student achievement. White Plains, NY: Longman, Inc. Bahr, M. W & Fuchs, D. (1991). Are teachers perceptions of difficult -to -teach students racially biased? School Psychology Review, 20, 599 610. Bandura, A. (1977). Self -Efficacy: Toward a unifying theory of behavioral change. Psychology Review, 84, 191 215. Bar barin, O., Bryant, D., McCandies, T., Burchinal, M., Early, D., Clifford, R., Pianta, R., & Howes, C. (2006). Children enrolled in public pre k: The relation of family life, neighborhood quality, and socioeconomic resources to early competence. American Jo urnal of Orthopsychiatry, 76, 265-276. Blackorby, J., & Wagner, M. (1996). Longitudinal post -school outcomes of youth with disabilities: Findings from the National Longitudinal Transition Study. Exceptional Children, 62, 399 413. Bracken, B. A. & Keith L. K. (2004). Clinical Assessment of Behavior. Lutz, FL: Psychological Assessment Resources. Bradley, R., Doolittle, J., & Bartolotta, R. (2008). Building on the data and adding to the discussion: The experiences and outcomes of students with emotional disturbance. Journal of Behavioral Education, 17, 4 23. Bradley, R., Henderson, K., & Monfore, D. A. (2004). A national perspective on children with emotional disorders. Behavioral Disorders, 29, 211 223.

PAGE 102

102 Bower, E. M. (1969). Early identification of e motionally handicapped children in school (2nd ed.). Springfield, IL: Charles C. Thomas. Boykin, A. W., Tyler, K. M., & Miller, O. (2005). In search of cultural themes and their expressions in the dynamics of classroom life. Urban Education, 40, 521 549. Bullis, M., & Cheney, D. (1999). Vocational and transition interventions for adolescents and young adults with emotional or behavioral disorders. Focus on Exceptional Children, 31, 1 24. Caldas, S. J. & Bankston, C. (1998). The inequality of separatio n: Racial composition of schools and academic achievement. Educational Administration Quarterly, 34, 533 557. Clinfton, R. A., Perry, R. P., Parsonson, K., & Hryniuk, A. (1986). Effects of ethnicity and sex on teachers expectations of junior high school students. Sociology of Education, 59, 58 67. Cohen, M., DuRant, R. H., Cook, C. (1988). The Conners teacher rating scale: Effects of age, sex, and race with special education children. Psychology in the Schools, 25, 195 202. Conners, C. K. (1973). Rat ing scales for use in drug studies with children. Psychopharmacology Bulletin (Special Issue-Pharmacotherapy with Children), 24 84. Cooper, H. M., Baron, R. M., & Lowe, C. A. (1975). The importance of race and social class information in the formation o f expectancies about academic performance. Journal of Educational Psychology, 67, 312 319. Coutinho, M. J. & Oswald, D. P. (1998). Understanding identification, placement, and school completion rates for children with disabilities: The influence of economic, demographic, and educational variables. In T. E. Scruggs & M. A. Mastropieri (Eds.), Advances in learning and behavioral disabilities: Vol 12 (pp. 43 78). Greenwich, CT: JAI Press. Cullinan, D. & Kauffman, J. M. (2005). Do race of student and race o f teacher influence ratings of emotional and behavioral problem characteristics of students with emotional disturbance? Behavioral Disorders, 30, 393 402. Daunic, A. P., Smith. S. W., Brank, E. M., & Penfield, R. D. (2006). Classroom based cognitive-beha vioral intervention to prevent aggression: Efficacy and social validity. Journal of School Psychology. 44 123139.

PAGE 103

103 Dembo, M. H. & Gibson, S. (1985). Teachers sense of efficacy: An important factor in school improvement. The Elementary School Journal, 86 173 184. DeMeis, D. K. & Turner, R. R. (1978). Effects of students race physical attractiveness, and dialect on teachers evaluations. Contemporary Educational Psychology, 3, 77 86. Donovan, M. S., & Cross, C. T. (Eds.). Minority students in specia l and gifted education. Washington, DC: National Academy Press. Downey, D. B. & Pribesh, S. (2004). When race matters: Teachers evaluations of students classroom behavior. Sociology of Education, 77, 267 282. Dusek, J. B. & Joseph, G. (1983). The bas es of teacher expectancies: A meta analysis. Journal of Educational Psychology, 75, 327 346. Emmer, E. T. & Hickman, J. (1991). Teacher efficacy in classroom management and disicipline. Educational and Psychological Measurement, 51, 755 765. Epstein, J N., March, J. S., Conners, C. K., Jackson, D. L. (1998). Racial differences on the conners teacher rating scale. Journal of Abnormal Child Psychology, 26, 109 118. Epstein, J. N., Willoughby, M., Valencia, E. Y., Tonev, S. T., Abikoff, H. B., Arnold, L. E., & Hinshaw, S. P. (2005). The role of childrens ethnicity in the relationship between teacher ratings of attentiondeficit/hyperactivity disorder and observed classroom behavior. Journal of Consulting and Clinical Psychology, 73, 424 434. Epstein, M H. & Cullinan ,D. (1998). Scale for assessing emotional disturbance. Austin, TX: Pro ed. Florian, L., Hollenweger, J., Simeonsson, R. J., Wedell, K., Riddell, S., Terzi, L., & Holland, A. (2006). Cross -cultural perspectives on the classification of c hildren with disabilities: Part I. Issues in the classification of children with disabilities. The Journal of Special Education, 40, 36 45. Florida Department of Education, Bureau of Exceptional Education and Student Services. (200 9 ). Florida Statutes and State Board of Education Rules: Excerpts for special programs. Tallahassee, FL: Author. Frey, A. (2002). Predictors of placement recommendations for children with behavioral or emotional disorders. Behavioral Disorders, 27, 126 126. Gelb, S. A. & Mi zokawa, D. T. (1986). Special education and social structure: The commonality of exceptionality American Educational Research Journal, 23, 543 557.

PAGE 104

104 Gerber, M. M. & Semmel, M. I. (1984). Teacher as imperfect test: Reconceptualizing the referral process Educational Psychologist, 19, 137 148. Gibson, S., & Dembo, M. (1984). Teacher efficacy: A construct validation. Journal of Educational Psychology, 76, 569 582. Gottlieb, D. (1964). Teaching and students: The views of negro and Caucasian teachers. Sociology of Education, 37, 345 353. Gresham, F. M. & Elliot, S. N. (1990) Social Skills Rating System. Circle Pines: MN: American Guidance Service. Gresham, F. M., Reschly, D. L., & Carey, M. P. (1987). Teachers as tests: Classification accuracy and concurrent validation in the identification of learning disabled children. School Psychology Review, 1987, 543 553. Hayling, C. C., Cook, C., Gresham, F. M., State, T., & Kern, L. (2008). An analysis of the status and stability of the behaviors of studen ts with emotional and behavioral difficulties. Journal of Behavioral Education, 17, 24 42. Hightower, A. D., Work, W. C., Cowen, E. L., Lotyczewski, B.S., Spinell, A. P., Guare, J. C., & Rohrbeck, C. A. (1986). The teacher -child rating scale: A brief ele mentary measure of elementary childrens school problem behaviors and competencies. School Psychology Review, 15, 393 409. Hilliard, A. G. (1992). Behavioral style, culture, and teaching and learning. The Journal of Negro Education, 61, 370 377. Jensen A. R. (1980). Bias in mental testing. New York: The Free Press. Kauffman, J. M. (2001). Characteristics of emotional and behavioral disorders of children and youth (7th ed.). Upper Saddle River, NJ: Prentice -Hall, Inc. Koff, C. W., Bradshaw, C. P., & Leaf, P. J. (2008). A multilevel studt of predictors of student perceptions of school climate: The effect of classroom level factors. Journal of Educational Psychology, 100, 96 104. Kranzler, J. H., Miller, M. D., & Jordan, L. (1999). An examination of racial/ethnic and gender bias on curriculum -based measurement of reading. School Psychology Quarterly, 14, 327 342. LadsonBillings, G. J. (2005). Is the team all right? Diversity and teacher education. Journal of Teacher Education, 56, 229 234.

PAGE 105

105 Landr um, T. J., Tankersly, M., & Kauffman, J. M. (2003). What is special about special education for students with emotional or behavioral disorders? The Journal of Special Education, 37, 148 156. Lau, A. S., Garland, A. F., Yeh, M., McCabe, K. M., Wood, P. A ., & Hough, R. L. (2004). Race/ethnicity and inter informant agreement in assessing adolescent pathology. Journal of Emotional and Behavioral Disorders, 12, 145 156. MacMillan, D. L. & Reschly, D. J. (1998). Overrepresentation of minority students: The c ase for greater specificity or reconsideration of the variables examined. The Journal of Special Education, 32, 15 24. MacMillan, D. L., Gresham, F. M., Lopez, M. F., & Bocian, K. M. (1996). Comparison of students nominated for prereferral interventions by ethnicity and gender. The Journal of Special Education, 30, 133 151. Mashburn, A. J., Hamre, B. K., Downer, J. T., & Pianta, R. C. (2006). Teacher and classroom characteristics associated with teachers ratings of prekindergartners relationships and behavior. Journal of Psychoeducational Assessment, 24, 367 380. Moller -Leimkuhler, A. M. & Yucel, M. (2009). Male depression in females? Journal of Affective Disorders, 121, 2229. Monroe, C. R. & Obidah, J. E. (2004). The influence of cultural synchr onization on a teachers perceptions of disruption: A case study of an African American middle school classroom. Journal of Teacher Education, 55, 256 268. Moretti, M. M., Catchpole, R. E. H., & Odgers, C. (2005). The dark side of girlhood: Risk factors and trajectories to aggression and violence. Canadian Child and Adolescent Psychiatry and Review, 14, 21 -25. Muthen, B. O. (1991). Multilevel factor analysis of class and student achievement components. Journal of Educational Measurement, 28, 338 354. Muthen, B. O. (1994). Multilevel covariance structure analysis. Sociological Methods and Research, 22, 376 398. National Center for Education Statistics (2000). ECLS K Base Year Data Files and Electronic Codebook. Rockville, MD: Westat. Nelson, J. R., Benner, G. J., Lane, K., Smith, B. W. (2004). Academic achievement of k 12 students with emotional and behavioral disorders. Exceptional Children, 71, 59 73.

PAGE 106

106 Ogbu, J. U. (1991). Minority coping responses and school experience. Journal of Psychohistory, 1 8, 433 456. Oswald, D. P., Coutinho, M. J., Best, A. M., & Singh, N. N. (1999). Ethnic representation in special education: The influence of school -related economic and demographic variables. The Journal of Special Education, 32, 194 206. Patton, J. M. (1998). The disproportionate representation of African Americans in special education. The Journal of Special Education, 32, 25 31. Pigott, R. L. & Cowen, E. L. (2000). Teacher race, child race, racial congruence, and teacher ratings of childrens sc hool adjustment. Journal of School Psychology, 38, 177 196. Podell, D. M. & Soodak, L. C. (1993). Teacher efficacy and bias in special education referrals. Journal of Educational Research, 86, 247 253. Raudenbush, S. W. & Bryk, A. S. (2002). Hierarchic al linear models: Applications and data analysis methods (2nd ed.). Thousand Oaks, CA: Sage Publications. Raven, J. C. (1947). Progressive Matrices. London: Lewis. Rayfield, A. (1997). Concurrent validity of the sutter -eyeberg student behavior inventor y with grade school children. Unpublished doctoral dissertation, University of Florida. Reid, R. (1995). Assessment of ADHD with culturally different groups: The use of behavioral rating scales. School Psychology Review, 24, 537 560. Reid, R., Casat, C D., Norton, H. J., Anastopulos, A. D. & Temple, P. (2001). Using behavior rating scales for ADHD scross ethnic groups: The IOWA Conners. Journal of Emotional and Behavioral Disorders, 9, 210 218. Reynolds, C. R. & Kamphaus, R. W. (1992). Behavior Asses sment Scale for Children: Manual. Circle Pines, MN: American Guidance Service, Inc. Rong, X. L. (1996). Effects of race and gender on teachers perception of the social behavior of elementary students. Urban Education, 31, 261 290. Rosenthal, R. & Jaco bson, L. (1968). Pygmalion in the classroom. New York: Holt, Rinehart, & Winston. Ryabov, I. & Van Hook, J. (2007). School segregation and academic achievement among Hispanic children. Social Science Research, 36, 767 788.

PAGE 107

107 Sacks, G. & Kern, L. (2008). A comparison of quality of life variables for students with emotional and behavioral disorders and students without disabilities. Journal of Behavioral Education, 17, 111 127. Sattler, J. M. & Hoge, R. D. (2006). Assessment of children: Behavioral, social and clinical foundations (5th ed.). La Mesa, CA: Jerome Sattler, Publisher, Inc. Sbarra, D. A. & Pianta, R. C. (2001). Teacher ratings of behavior among African American and Caucasian children during the first two years of school. Psychology in the Schools, 38, 229 238. Schalock, R. L., Holl, C., Elliott, B., & Ross, I. (1992). A longitudinal follow up of graduates from a rural special education program. Learning Disability Quarterly, 15, 29 38. Serwatka, T. S., Deering, S., & Grant, P. (1995). Disproportionate representation of African Americans in emotionally handicapped classes. Journal of Black Studies, 25, 492 506. Shinn, M. R., Tindal, G. A., & Spira, D. A. (1987). Special education referrals as an index of teacher tolerance: Are teachers imp erfect tests? Exceptional Children, 54, 32 40. Skiba, R. J., Poloni -Staudinger, L., Simmons, A. B., Feggins -Azziz, L. R., & Chung, C. (2005). Unproven links: Can poverty explain ethnic disproportionality in special education? The Journal of Special Educa tion, 39, 130 144. Sleeter, C. E. (2001). Preparing teachers for culturally diverse schools: Research and the overwhelming presence of Caucasianness. Journal of Teacher Education, 52, 94 106. Soodak, L. C. & Podell, D. M. (1993). Teacher efficacy and student problem as factors in special education referral. The Journal of Special Education, 27, 66 81. Sutter, J. & Eyeberg, S. M. (1984). Sutter -Eyeberg Student Behavior Inventory. (Available fr om Shelia Eyeberg, Department of Clinical and Health Psychology, University of Florida, Gainesville, FL, 32610. TschannenMoran, M., Hoy, A. W., & Hoy, W. K. (1998). Teacher efficacy: Its meaning and measure. Review of Educational Research, 68, 202 248. U. S. Department of Education (2006). Federal register: Assistance to states for the education of children with disabilities and preschool grants for children with disabilities; final rule. Washington, DC: Author.

PAGE 108

108 U.S. Department of Education. (2005). Twenty -seventh annual report to Congress on implementation of the Individuals with Disabilities Education Act. Washington, DC: Author. U.S. Department of Health and Human Services (1999). Mental Health: A Report of the Surgeon General Rockville, MD: U.S Department of Health and Human Services, Substance Abuse and Mental Health Services Administration, Center for Mental Health Services, National Institutes of Health, National Institute of Mental Health. Valles, E. C. (1998). The disproportionate representation of minority students in special education: Responding to the problem. The Journal of Special Education, 32, 52 54. Wagner, M. M. (1995). Outcomes for youths with serious emotional disturbance in secondary school and early adulthood. The Futures o f Children, 5, 90 112. Wagner, M., Kutash, K., Duchnowski, A. J., Epstein, M. H., & Sumi, W. C. (2005). The children and youth we serve: A national picture of the characteristics of students with emotional disturbances receiving special education. Journa l of Emotional and Behavioral Disorders, 13, 79 96. Wagner, M. & Cameto, R. (2004). The characteristics, experiences, and outcomes of youth with emotional disturbances. NLTS2 Data Brief, 3 (2). Walthall, J. C., Konold, T. R., & Pianta, R. C. (2005). Fa ctor structure of the social skills rating system across child gender and ethnicity. Journal of Psychological Assessment, 23, 201 215.

PAGE 109

109 BIOGRAPHICAL SKETCH Christina Peters was born in Winter Park, Florida, in 1983. She spent many wonderful years in t he area and graduated from Winter Park High School in 2001. She earned her B.S. in Psychology and her B.A. in Criminology from the University of Florida (UF) in 2005. She earned her M.Ed. in school psychology from UF in 2007. Specializing in Emotional and Behavioral Disorders (EBD), Christina had the opportunity to complete many unique practicum placements to enhance professional development including the public school system, a center day school for students with EBD, a Department of Juvenile Justice Resid ential Facility, Shands Teaching Hospital, P. K. Yonge, and PACE Center for Girls. Christina completed a year -long internship with Hillsborough County Public Schools as a culminating experience, throughout which she learned how to become an independent pra ctitioner. Christina has taken the opportunity to travel to several international locations in order to enhance cultural knowledge and awareness and to explore the outdoors. Additionally, she has led commercial expedition wilderness trips for college students in Alaska and several other domestic trips with friends. Christinas future plans include obtaining licensure as a psychologist, working collaboratively with school system professionals to enhanc e the lives of children, exploring the world one country at a time, and living a fulfilling life with family and friends.