Hierarchical linear modeling

Abstract

Teachers of students with emotional and/or behavioral disorders (EBD) are less experienced and more likely to have emergency certification than teachers of students with other disabilities. Yet, to date, research has not examined the relationship between the academic achievement of students with EBD and characteristics associated with highly qualified teachers (teachers’ education level, certification status, and years of experience). Using a nationally representative longitudinal data set of students with disabilities, this study examined the relationship between teacher characteristics and the academic achievement of students with EBD. Using hierarchical linear modeling, the study found low academic achievement for students with EBD, null effects for change in achievement across time, and null effects for the relation between (a) teachers’ educational level, certification status, and years of experience and (b) student academic achievement. Results indicate further research is needed to examine whether and how teacher characteristics may impact the academic achievement of students with EBD.

 

With the passage of No Child Left Behind (NCLB) and the 2004 reauthorization of the Individuals With Disabilities Education Act (IDEA), the term highly qualified teacher surfaced as a contentious idiom. Policymakers have highlighted the impact of high-quality teachers on student performance in general and for specific groups of learners, including minority and low-socioeconomic status (SES) students (U.S. Department of Education, 2004). However, specific characteristics of training and/or qualifications predicting highly effective and qualified teachers have been difficult to define (Darling-Hammond & Youngs, 2002). In this study, we examined whether certification type, education level, and years of experience–variables associated with highly qualified teachers in previous research (e.g., Carlson, Lee, & Schroll, 2004)–predicted academic growth for students with emotional and/or behavioral disorders (EBD).

Teachers of Students With EBDs

The expectation of all students having access to highly qualified teachers is problematic for students with EBD. Not only is there a national shortage of teachers of students with EBD (American Association for Employment in Education in the United States, 2008), but those that are in the field are more likely to have an emergency or alternative certification and fewer years of teaching experience than other special education teachers. Billingsley, Fall, and Williams (2006) examined the characteristics of teachers of students with EBD with the Study of Personnel Needs in Special Education (SPeNSE) database to identify differences between special education teachers generally and teachers of students with EBD. Results indicated that teachers of students with EBD were statistically significantly younger than other special education teachers (M= 42.3 vs. M = 43.4, p < .01), had fewer years of teaching experience (M= 10.0 vs. M= 12.7,p < .01), and were less likely to be fully certified (83.9% vs. 92.8%, p< .01). Similar results were identified by Henderson, Klein, Gonzalez, and Bradley (2005), who found that teachers of students with EBD were less likely to have a master’s degree (52.9% vs. 60.3%, p = .021), less likely to be fully certified for their main teaching assignment (89.3% vs. 92.8%, p = .002), and more likely to be emergency certified (10.4% vs. 4.0%, p = .002). Twice as many EBD teachers were credentialed through alternative certification programs than other special educators (11.7% vs. 6.7%, p = .006).

Academic Achievement and Students With EBDs

Research has consistently found that students with EBD experience poor academic outcomes (Bradley, Doolittle, & Bartolotta, 2008; Wagner, 1995; Wagner & Davis, 2006). In their meta-analysis of academic achievement and students with EBD, Reid, Gonzalez, Nordness, Trout, and Epstein (2004) found that most students with EBD performed at or below the 25th percentile in general academic functioning and found a moderate to large negative effect (-0.69) related to the academic achievement of students with EBD in comparison with nondisabled students. Research continues to corroborate these findings. Lane, Barton-Arwood, Nelson, and Wehby (2008) assessed 42 students with EBD educated in self-contained or segregated classroom settings, finding that the sample performed below the 25th percentile on all math, reading, and written expression measures of the Woodcock-Johnson III Test of Achievement (WJ-III: Woodcock, McGrew, & Mather, 2001). Findings from the Special Education Elementary Longitudinal Survey (SEELS), which included a nationally representative sample of students with EBD, indicated that for reading, 61% of student with EBD scored in the bottom quartile on the Passage Comprehension subtest of the WJ-III, and for math, 43% were in the bottom quartile on the Calculation subtest (Wagner, Kutash, Duchnowski, Epstein, & Sumi, 2005).

Although research findings identify low academic achievement for students with EBD, there is heterogeneity of findings (Trout, Nordness, Pierce, & Epstein, 2003), meaning not all students experience low achievement. Numerous factors have been examined to explain differences in academic achievement. For example, research has found that students served in more restrictive settings have larger academic deficits than students served in less restrictive settings (Carr-George, Vannest, Willson, & Davis, 2009; Lane, Wehby, Little, & Cooley, 2005). Wiley, Siperstein Bountress, Forness, and Brigham (2008) examined the relation between school context, particularly school-level SES, and the academic achievement of 140 elementary students with EBD and found that, based on the WJ-III, students with EBD in high-SES schools performed better than those in low-SES schools. In a follow-up study, Siperstein, Wiley, and Forness (2011) followed 86 students with EBD for 2 years from low-SES schools and high-SES schools and found that, although differences were evident at initial status, the students made no statistically significant improvement in academic achievement across time. Although research has examined the relationship of school-level and placement characteristics to the academic achievement of students with EBD, we identified no research that has specifically examined the relationship between teacher characteristics and academic achievement of students with EBD.

Relation Between Teacher Characteristics and Academic Achievement

Researchers have suggested there may be a relation between (a) teachers’ education, certification, and experience and (b) the academic achievement of students with EBD (Manning, Bullock, & Gable, 2009; Sutherland, Denny, & Gunter, 2005). Although researchers have examined the relation of variables associated with highly qualified teachers (e.g., certification type, education level, and years of experience) and academic outcomes for many groups of students, we are not aware of research that has empirically examined these relations specifically for students with EBD.

Multiple studies have reported positive correlations between teacher certification and student achievement (Betts, Zau, & Rice, 2003; Dee, 2004; Feng & Sass, 2013). However, in their systematic review of the relation between student achievement gains and teacher characteristics, Wayne and Youngs (2003) found mixed results for the relation between teachers’ certification (e.g., alternative, subject-specific, dual certification in general and special education) and their students’ achievement gains, findings that were corroborated by Harris and Sass (2011). In addition, Phillips (2010) found that students in the Early Childhood Longitudinal Study Kindergarten Cohort identified as at-risk–defined as (a) the student living in a single-parent home, (b) the student’s mother not having a high school diploma, (c) the student’s home language not being English, or (d) the student’s family living below the poverty line–did not benefit from having a teacher that was fully certified.

Studies have found a positive relation between teachers’ education level (e.g., holding a master’s degree) and student achievement (Feng & Sass, 2013; Nye, Konstantopoulos, & Hedges, 2004). Desimone and Long (2010) reported a positive and significant relation between teachers’ education level and mathematics growth rates for students in early elementary school, and Phillips (2010) reported similar findings for reading achievement of older elementary-aged students. Feng and Sass (2013) evaluated the relation between achievement and teacher education level for both students with and without disabilities. Overall, they found a significant positive relation between teacher education and student mathematics achievement for both groups of students.

Research examining the relation between teachers’ experience and student achievement has been mixed. Wayne and Youngs (2003) found positive correlations between teachers’ experience and student academic gains, but the studies from which the findings were culled were methodologically suspect and the authors suggested that results be interpreted with caution. Harris and Sass (2011) reviewed research published after 2003 and found variation by school and academic focus. Results for elementary math were evenly split between positive and no effects of teaching experience on student achievement, whereas positively correlated results were found for elementary reading. In middle school, the results were reversed, with teacher experience being positively correlated with math performance and no consistent relation evident for middle school reading. One study of high school teachers found strong positive effects (Clotfelter, Ladd, & Vigdor, 2007), whereas two studies found no significant correlation between teacher experience and student achievement (Aaronson, Barrow, & Sander, 2007; Betts et al., 2003). Harris and Sass’s (2011) study of teachers in Florida found generally positive associations between teaching experience and students’ academic achievement.

Much less research has specifically examined the relation between teacher characteristics and the academic achievement of students with disabilities. Using a state-level database, Feng and Sass (2010) examined academic achievement gains for students with disabilities across 5 years and whether teacher characteristics were associated with different patterns of performance. Their study found no relation between achievement gains and in-service professional development, but did identify an association between pre-service certification and academic gains. Students with disabilities educated in general education settings performed significantly better in reading and math if their teacher had special education certification, and students with disabilities educated in self-contained settings performed significantly better in reading if their teacher had special education certification. However, the majority of teachers in self-contained settings were special education certified, so there was little variability in the predictor. Of note is that these findings differ from those for general education students and their teachers’ certifications.

The relation between students’ academic outcomes and the quality of their teachers is well established (Darling-Hammond & Youngs, 2002; Glazerman, Protik, Teh, Bruch, & Max, 2013; Sutherland, Lewis-Palmer, Stichter, & Morgan, 2008). However, a clear, empirically derived definition of teacher quality eludes the education field generally, and the EBD field specifically. Simpson, Peterson, and Smith (2011) outlined critical components of education programs for successfully supporting the diverse needs of students with EBD and identified teachers’ qualities, specifically well-trained and certified teachers, as the most important ingredient. Yet, research to date has not empirically verified the relation between teacher qualifications (e.g., advanced degrees, full certification, and experience) and academic achievement for students with EBD, and it is unclear whether those teacher characteristics are positively correlated with academic gains with this population.

Purpose

For students with EBD, it has been posited that poor academic outcomes (Nelson, Benner, Lane, & Smith, 2004; Reid et al., 2004; Wagner et al., 2005) are correlated with teachers’ level of certification and experience (Manning et al., 2009; Sutherland et al., 2005). However, to date, we are not aware of research that has directly examined the relations between (a) academic achievement of students with EBD and (b) their teachers’ level of education, certification level, and years of teaching experience. To address this gap, this exploratory study examined the relation between the academic achievement of students with EBD over time and teachers’ level of education, credentials, and years teaching using a national longitudinal extant database. Specific research questions were as follows:

Research Question 1: Do students with EBD have heterogeneous or homogeneous academic abilities and growth in academic achievement over time?

Research Question 2: Does the level of education, certification level, or number of years teaching of teachers of students with EBD relate to rates of growth in academic achievement?

Method

This study was designed to explore the relation between the academic achievement of students with EBD and teacher characteristics, specifically level of education, certification, and years teaching, by conducting a secondary data analysis of the Special Education Elementary Longitudinal Study (SEELS). SEELS is a national policy study of elementary and middle school students with disabilities mandated by the U.S. Department of Education and developed and implemented by Stanford Research Institute (SRI) International. SEELS collected data across three waves over 5 years (2001-2006) on students, their families, their teachers, and their schools. The SEELS database has a nationally representative weighted sample of students with disabilities and provides a broad range of information at the student level for analysis. This study utilized the weighted data from all three waves of the teacher interview and the direct assessment databases (see www.seels.net for more information about the SEELS databases). The weights provide a population estimate based on the observed sample and were developed so that they summed to the number of students in each disability category as reported by states for the 1999-2000 school year.

Sample

Although the SEELS database includes a nationally representative sample of subjects with EBD, the presence of missing data between instruments (teacher interview and direct assessment) reduced the number of students included in this analysis. This study analyzed data from a weighted sample of 39,561 students with data from all three waves. The age range for the sample was 7 to 14 years (M= 10.5) during Wave 1,9 to 15 years (M= 11.9) during Wave 2, and 10 to 17 years (M = 13.8) during Wave 3. Sixty-eight percent were male and 32% were female, which is consistent with national estimate (Wagner et al., 2005). Eighty-three percent of the sample was Caucasian, 12.3% was African American, and 4.2% was Hispanic. SES of the sample was equally distributed across household earnings, with 29.1% from households earning US$25,000 or less and 31.4% from households earning more than US$50,000. Eleven percent of the sample was from rural school districts, 64.7% was from suburban school districts, and 24.6% was from urban school districts.

Measures

A unique characteristic of SEELS as a national policy study was the first-ever effort to directly assess students with disabilities (Wagner et al., 2005). The study used the research editions of specific subtests of the WJ-III (Woodcock et al., 2001). The WJ-III is a norm-referenced, comprehensive assessment of academic skills and knowledge with well-researched psychometric properties. This study utilized the standard scores of WJ-III subtests in (a) Applied Problems (mathematics); (b) Calculations (mathematics); (c) Letter Word Recognition; (d) Passage Comprehension; as well as (e) Oral Reading Fluency, which was not standardized. Although each of these subtests focuses on a specific skill domain, taken together, they represent an overall academic skill construct in the areas of reading and mathematics. Because an overall achievement score was not available in the database, the five WJ-III subtests were included in a factor analysis to aggregate as a single construct score for each wave of data collection. The factor analysis constrained the model to the identification of a single factor and used varimax rotation to increase the factor loadings with a small number of variables (Pett, Lackey, & Sullivan, 2003). The resulting coefficients had a mean of 0 and a variance equal to the squared multiple correlations between the estimated factors scores and the true factor scores. The factor coefficient scores were used as the outcome variable as they statistically represent the construct of academic achievement used in this study.

Three predictor variables were included to identify longitudinal impact of teacher characteristics. The SEELS Teacher Interview was conducted with each student’s primary literacy arts teacher each year. Although characteristics of all teachers would be ideal, research has substantiated the relation between reading and problem behavior, including longitudinal negative trajectories (Morgan, Farkas, Tufis, & Sperling, 2008), suggesting that high quality language arts/reading instruction may have a concomitant impact on academic and behavior performance. The teachers were asked, “What type of certification do you hold?” “What is your highest level of education?” and “How many years have you been teaching?” Although other variables were available (e.g., number of professional development hours completed), we focused exclusively on these as they were used in prior research (Feng & Sass, 2013) and have been noted as possible correlates for achievement for students with EBD (Sutherland et al., 2005). Two of the three variables were categorical; therefore, they were dummy coded as follows:

* Education–both bachelor’s degree and bachelor’s degree with a fifth year were collapsed into a single bachelor’s degree variable with a value of “0”; master’s degree, master’s degree with an additional year, and PhD were collapsed into a single variable with a value of “1”; all other values were coded as missing as they did not fit within the two categories.

* Certification–the regular certification value was retained and coded as “0”; all other certifications were collapsed into an “other” category with a value of “l.”

* Years Teaching–the numerical value of years taught was retained without recoding.

Table 1 provides a breakdown for the education level and certification level reported by the teachers across waves. The mean values for teachers’ years of teaching were 12.2 years for Wave 1, 11.8 years for Wave 2, and 14.6 years for Wave 3. A mean value across waves for each variable was then calculated to represent the mean level of education (Mean_JED), certification (Mean_Cert), and years teaching (MeanYear) for each student.

Analysis Procedures

In addition to the factor analysis performed on the WJ-III subtests for each student for each wave, we utilized descriptive statistics and hierachical linear modeling (HLM) for analysis of the research questions. Descriptive statistics were utilzed to assess mean and variation of academic achiement of students with EBD. HLM was applied because it allows for variance explanation in individual growth trajectories and the variance in individual intercept (initial status) and slope (growth rate) explained by between-subject factors (Raudenbush & Bryk, 2002; Singer & Willett, 2003). In this study, we examined the predictive relation between teacher characteristics and students’ achievement growth curves. The analysis was conducted using HLM 7.0 Student Version (Raudenbush, Bryk, & Congdon, 2010), which allows for the inclusion of design weights. SRI recommended that inferential analyses using the SEELS database include adjusted standard errors because the sampling procedure was not fully random (Wagner, Marder, & Blackorby, 2002). Although HLM 7.0 estimates robust standard errors and adjusts for design weights, we took a conservative approach to reduce Type 1 error and adjusted the standard errors using the following formula recommended by SRI:

[mathematical expression not reproducible],

where SE is the standard error, M is the mean of the weight, and V is the variance of the weight. In this study, the M of the weight was 488.23, the V was 72,407.84, resulting in an adjustment of 1.427.

First, an Unconditional Growth Model (Model 1) was analyzed to test for differences in academic achievement at baseline (intercept) and students’ rate of growth (slope). In addition, the unconditional model’s residual variances were used to calculate an intraclass correlation coefficient (ICC) to quantify the population covariance between initial status (baseline) and change (growth; Singer & Willett, 2003). Next, we tested a Slopes as Outcome Model, which identified the effect of teacher characteristics on participants’ baseline academic achievement and rate of academic achievement growth. The predictor variables were not modeled on the initial status (intercept) because the teacher characteristics reported began the same year as the assessment of academic achievement; influence of predictor variables on initial status would be uninformative and assume previous performance was contingent on current and future teacher characteristics.

Results

Descriptive Statistics and Factor Analysis

The first substantive question in this study was the variability in initial status and rate of change (slope) of academic achievement of students with EBD. The descriptive statistics of the WJ-III subtest scores in Table 2 indicate, at a mean level, there was very little change in relative academic achievement across SEELS waves and, generally, the standard score change was negative, relative to the norms. These results suggest the standard scores of students with EBD on the WJ-III subtests in math and reading are decreasing across time (note that Oral Reading Fluency measures words read correctly per minute and is not standardized by grade).

The factor analysis, as described in the previous section, was conducted to have a single construct score as the outcome variable. The factor analysis constrained the number of identified factors to one, which, as indicated by Table 3, worked out well as none of the factor loadings were below .741 across the three waves. The cumulative percentage of variance accounted for by the single factor was 62.3% for Wave 1, 63.7% for Wave 2, and 61.9% for Wave 3.

Hierarchical Linear Model

Unconditional Growth Model. Before testing for the influence of teacher education, certification, and years teaching, the Unconditional Growth Model was tested to identify the presence or absence of intra- and inter-individual differences across waves. Table 4 displays the original standard error estimates and the calculated adjusted standard errors, and Table 5 provides the model estimates with the adjusted standard errors. The t ratio, which is used for identifying p values, is calculated from the model coefficients and standard errors. As can be seen in Table 4, the adjustment increased the standard errors, thereby decreasing the t ratio for a more conservative estimate of statistical significance. Results of the unconditional model indicates that the baseline academic achievement is not significant (p = .524), meaning that the estimate is zero, which is expected as zero is the mean value of the z score. The insignificant slope coefficient (p = .602) suggests that the rate of academic achievement growth is also zero. These results suggest that, on average, there was no significant achievement growth for the sample of students with EBD. These results are congruent with the individual subtest results reported in Table 2, meaning that, although a decreasing trend is present, the rate of decrease is not statistically significant.

The random effects were statistically significant, which suggests that there was variability among the students with regard to their achievement and there was variability in students’ growth. We used the variance estimates to calculate the ICC of the unconditional model. We found an ICC of .88, which suggests that a student’s baseline academic achievement accounts for 88% of the variability in rate of change.

Slopes as Outcome Model. In the Slopes as Outcome Model, the mean values across all three waves of teachers’ education, certification, and years teaching were modeled as predictors of slope variability. The results, as presented in Table 5, indicate that none of the teacher characteristics examined in this study accounted for a significant amount of the variance in the sample’s rate of academic achievement growth. The largest coefficient was for education level and it was negative. However, the coefficient was not statistically significant. Overall, these results suggest that growth of students with EBD in academic achievement was not systematically impacted by the overall mean level of their teachers’ education, certification, or years of experience.

Discussion

Prior research indicates that students with EBD experience poor short- and long-term academic and behavioral outcomes (Bradley et al., 2008). The results from this study replicate previous findings suggesting that the academic achievement of students with EBD was not only below the population mean, but did not increase over time relative to students without EBD (across 5 years of data collection). Furthermore, there was no significant impact on growth in academic achievement across time by teachers’ education, certification, and years of experience.

Academic Abilities and Growth in Academic Achievement Over Time

The findings support research indicating that students with EBD have poor academic achievement (Nelson et al., 2004). Based on the mean values across the four standardized WJ-III subtests, students with EBD in this study performed below average in both math and reading, and their achievement slightly decreased relative to test norms across time. Although the negative slope coefficient was not statistically significant, the lack of growth across time is troubling. The results indicate that students with EBD remained well below the mean in both reading and mathematics in elementary school. Greenbaum et al. (1996) reported similar results, finding 85% of students with EBD reading below grade level by high school. Similar trajectories have also been reported for mathematics achievement (Templeton, Neel, & Blood, 2008). Findings from this study in conjunction with those from prior research indicate that students with EBD are failing to reach academic success across time.

Characteristics of Teachers of Students With EBD and Rates of Growth in Academic Achievement

Policy and practice across states reflect the assumption that highly qualified teachers are integral to student success. Indicators used to judge quality often include observable and measurable teacher characteristics (i.e., teachers’ education, certification, and years of experience; Carlson et al., 2004). Findings from this study indicate that education level, certification, and years of experience are not associated with growth in academic achievement among elementary students with EBD. As such, results suggest that these three teacher characteristics may not be indicators of a quality teacher for students with EBD in respect to academic achievement. Although exploratory, findings suggest further study is needed to identify what measurable indicators, such as type or hours of professional development (Sutherland et al., 2005) or classroom management skills (Oliver & Reschly, 2010), might be associated with high quality teachers and academic achievement for students with EBD.

Limitations

Although this study utilized a nationally representative weighted sample, a number of methodological limitations necessitate highlighting. First, the sample analyzed was weighted, meaning the results are estimates from the population. However, the results should be interpreted with caution as not all students in the original sample had complete data for all three waves; therefore, the population estimates are reflective of the adjusted sample, not the full sample. In addition, the scale of the outcome variable makes interpretation of the growth model difficult because the scale is a composite z score for participants’ standard scores on each WJ-III subscale. However, despite these analysis limitations, the fit of study findings with previous research related to the low longitudinal achievement of students with EBD reduces concern surrounding the accuracy of results and interpretation. Also, due to the use of a previously collected sample, we were not able to investigate alternate measures of teacher characteristics and behavior, including verbal ability, course work, and observed instructional behaviors (e.g., opportunities to respond). The collection of additional characteristic information would have made for a more comprehensive study. Furthermore, the teacher characteristics were based on the students’ primary special education language arts teacher. Although the teachers that completed the SEELS Teacher Interviews had significant contact with each student, they were not the only teachers instructing the students. Finally, we essentially estimated a dose-response model with the cumulative dosage of teacher characteristics over 3 years by averaging each student’s teachers’ characteristics. This approach limits interpretation of a single teacher’s effect on achievement, but instead provides an estimate of student growth based on that student’s dosage of the characteristics, which may have varied year to year.

Implications

Although the findings are null, they suggest some preliminary implications for practice. First, because students with EBD are entering into special education services with profiles that typically indicate below average academic performance, findings suggesting a lack of improvement in academic achievement create significant concern and imply the need to pinpoint variables that directly impact the achievement of these students. Although the three teacher characteristics included in this study have been previously examined and promoted as indicators of highly qualified teachers, they were not indicative of improved achievement for the students with EBD in this study. Thus, placing students with EBD with teachers with considerable experience, full certification, and/or an advanced degree should not be assumed to result in improved achievement. It may be that specific training in effective instructional practices and behavior management approaches that have been validated for students with EBD (e.g., functional behavior assessment; Gage, Lewis, & Stichter, 2012) may be more important than teaching experience, licensure status, and advanced degrees when seeking to improve these students’ academic performance.

Moreover, results emphasize the importance of future research examining and identifying teacher-level variables and instructional practices that do predict relative growth in academic outcomes for students with EBD. For example, researchers have been examining special education teacher evaluation methods that accurately predict student growth using value-added modeling (Buzick & Jones, 2015). Although tools have not been fully developed for addressing the unique teacher-level variables for teaching students with EBD (Brownell & Jones, 2015), efforts are under way to develop and validate such tools (Semmelroth & Johnson, 2014).

Future Directions for Research

Due to the large number of factors affecting the performance of students with EBD, determining what works to increase the academic achievement of this population is a complex and difficult task. The current study lays a foundation for building an empirically supported understanding of longitudinal academic achievement for students with EBD by examining the links between common indicators of highly qualified teachers and student outcomes. Findings of this study focused exclusively on academic outcomes, but the interrelation between both academic and behavioral constructs is well documented in the literature (Kauffman, 2010). Future research is necessary to identify malleable factors mediating the interrelation between teacher characteristics, teacher behaviors, and students’ academic and behavioral outcomes. Specifically, research should seek to identify teacher characteristics that directly impact both academic and behavioral outcomes in observable and measurable ways. The identification of these essential characteristics should be used to generate an empirically supported description of highly qualified teachers for students with EBD that can be reflected in policy and practice.

Based on study findings, increased experience, attaining an advanced degree, and becoming fully certified in and of themselves do not appear to be effective approaches for improving academic achievement for students with EBD. Thus, other teacher characteristics and behaviors should be investigated to identify factors associated with highly qualified teachers for students with EBD. Although a number of teacher characteristics have been linked with student outcomes, including verbal ability and content knowledge (Wayne & Youngs, 2003), those links have not yet been corroborated with students with EBD. Research clearly documents that, on average, current policy and practice are not resulting in academic success for students with EBD (Bradley et al., 2008; Nelson et al., 2004), yet it seems apparent that many teachers are having a positive effect on the academic achievement of students with EBD. Researchers should seek to identify specific characteristics indicative of teacher effectiveness for students with EBD, which the field can then target in pre-service and in-service training efforts to improve the academic outcomes of students with EBD.