An Integrative Synthesis of Literature on Disproportionality in Special Education

Abstract

Research studies have historically indicated that students from racially and ethnically diverse backgrounds are overidentified for special education, suggesting bias in referral, assessment, and placement practices. Other studies, however, have suggested that students from racially and ethnically diverse backgrounds are not overrepresented in special education or may be underidentified for services. There is a perceptual interpretive element in defining the problem of disproportionality, as the use of different data sets and analyses impact how both the problem and results are interpreted. The purpose of this manuscript was to examine the ways in which current studies analyze disproportionality through statistical methods, and to compare those analyses based on the conceptualization of covariates. An integrative systematic review of the literature builds on previous works that examine the issue of disproportionality.

Keywords

overrepresentation underrepresentation disproportionality special education race/ethnicity racially and ethnically diverse students

Special education disproportionality is defined as the extent to which membership in a given group, such as gender, race/ethnicity, or socioeconomic strata, differentially affects the probability of being labeled as having a disability and placed in special education (Oswald, Coutinho, Best, & Singh, 1999). Since the issue of disproportionality by race and ethnicity was first identified (Dunn, 1968), it has been under constant legally and emotionally charged scrutiny (Artiles & Trent, 1994). The overrepresentation of students from racially and ethnically diverse backgrounds in special education has especially been a concern due to the potentially adverse effects on students, such as the risk of stigma and the risk of being placed in a segregated setting with less rigorous curriculum (Sleeter, 2010). In fact, it has been argued that categorizing and labeling students serves as a mechanism for stratification within schools in terms of differences in access to knowledge and attainment, and that placing students from racially and ethnically diverse backgrounds in special education at higher levels than White counterparts serves as justification for continued educational inequities (Rocha & Hawes, 2009; Sullivan & Artiles, 2011). This was evident in the cases of Larry P. v. Riles (1972), PASE v. Hannon (1980), and Diana v. State Board of Education (1970) when students from racially and ethnically diverse backgrounds were disproportionately placed in what the courts considered to be “dead-end classes” that focused on functional curriculum with little opportunity for academic advancement.

Concerns that, historically, schools have identified and placed students from racially and ethnically diverse backgrounds in special education at disproportionately high rates have resulted in policy briefs (e.g., National Education Association, 2007), position statements (e.g., Skiba, 2012), and federal mandates, including the Individuals With Disabilities Education Act (IDEA; 2004) stipulations that the special education assessment process include a nondiscriminatory evaluation by a multidisciplinary team, and that there is consideration of the maximum benefits and risks associated with a given academic placement. States and local education agencies (LEAs) are required to monitor and report on disproportionality (Zhang, Katsiyannis, Ju, & Roberts, 2014), and LEAs found to have significant disproportionality by race or ethnicity in the areas of identification, placement, or discipline must use IDEA funds to provide coordinated early intervening services (CEIS) that address root causes (IDEA Data Center, 2015).

Though the overrepresentation of students from racially and ethnically diverse backgrounds in special education has been identified as a significant problem, research is in nascent stages regarding understanding the interplay of factors that may cause and maintain disparities. This article focuses on disproportionality from a quantitative perspective, as the way in which the issue is framed and the choice of methodological design may impact the results. Federal and state agencies mandate reductions in overrepresentation of racially and ethnically diverse groups in special education, but research studies have produced inconsistent findings, as discussed subsequently, related to the extent of overrepresentation for various racial groups within specific eligibility categories and in special education generally. First, we discuss theoretical frameworks that shape how the issue is interpreted, and, second, we discuss the conceptualization of covariates within disproportionality research.

Design and Data

Our understanding of disproportionality in present day is, in part, attributable to the particular lens and theoretical framework from which studies derive. Waitoller, Artiles, and Cheney (2010) reviewed disproportionality literature and described three main ways in which overrepresentation research has been framed within the past 40 years; these include “[a] sociodemographic model in which characteristics of individuals and contexts are examined to understand the problem, analyses of the sociohistorical issues related to race and race relations, and examinations of overrepresentation through the study of professional practices” (p. 36). Out of 42 studies included in their review, 14 examined sociodemographic characteristics, most of which considered access to educational opportunities, economic indicators, health and environmental impacts, and community demographics; two examined sociohistorical factors, including the history of racial segregation and historical processes that situate minority groups as juxtaposed with dominant majority culture; and more than half (n = 26) closely considered the technical dimension of overrepresentation, including how practitioners enact and mediate the assessment process, individual beliefs and biases throughout the referral process, and decision-making in the eligibility process.

The theoretical framework should influence a study’s methodological design, but even with a well-articulated framework, research in the area of disproportionality “has long been restricted by the availability of data” (Sullivan & Bal, 2013, p. 478). Fine-grained analysis can be difficult as many districts do not collect or share data at the level needed to control for multiple covariates, and studies often cannot account for nesting within schools and throughout time. Studies are limited to variables in the form that they have been measured; however, the way in which one conceptualizes and applies covariates is likely to impact the statistical result. For example, within the United States, race and socioeconomic status (SES) are inextricably linked (Carter & Reardon, 2014); some studies have hypothesized that environmental risk factors associated with poverty are highly correlated with race or ethnicity (Donovan & Cross, 2002; Hibel, Farkas, & Morgan, 2010; Morgan et al., 2017) and may explain racial disparities in special education identification. When studying disproportionality, whether one controls for SES using free or reduced lunch or neighborhood mean income level will affect the interpretation of results. Parent education levels, community resource level, free or reduced lunch status, and composite SES indicators have all been used in recent studies (e.g., Coutinho, Oswald, & Best, 2002; Kincaid & Sullivan, 2017; Sullivan & Artiles, 2011) with varying outcomes.

There are also few large-scale representative studies that have statistically controlled for an extensive number of individual, family, and school-level covariates in attempting to determine which factors elevate a student’s likelihood of being placed in special education (Hibel et al., 2010), and at what point in a student’s school career these placements are most likely to occur. Recent studies have shown conflicting results; for example, Hibel et al. found that when controlling for SES and students’ initial levels of academic achievement, Black, Latino, and Asian students were significantly underrepresented in special education. Hosp and Reschly (2004) conducted a similar study and found that demographics such as race and English proficiency remained strong predictors of identification even after controlling for achievement and economic predictors, and although SES offset some of the difference, it did not eliminate race as a predictor.

Morgan et al. (2015) summarized several contradictory theories within the field of disproportionality and determined through their analyses that there was no evidence for overrepresentation of students from racially and ethnically diverse backgrounds in special education. Skiba, Artiles, Kozleski, Losen, and Harry (2016) countered Morgan et al. (2015) with the argument that although the authors applied a complex statistical procedure to their analysis, they failed to appreciate the complexity of the issue, and their results overreached considerably given methodological limitations (see Skiba et al. for the complete response). Morgan and Farkas (2016) rebuked the aforementioned claims in another set of technical comments published in Educational Researcher. The authors maintained that the best available evidence indicates that students from racially and ethnically diverse backgrounds are less likely than their White peers to receive special education services, even when displaying the same relative need (p. 227).

Purpose of the Study

Though federal and state agencies have identified overrepresentation of students from racially and ethnically diverse backgrounds in special education as a significant problem in the field, research in the area of disproportionality has presented inconsistent findings. In this article, we address the study of disproportionality from a quantitative perspective to better understand how methodological approaches impact outcomes, which, in turn, affect how we interpret and approach the issue. The purpose of this study was to synthesize available literature on quantitative studies of disproportionality by race or ethnicity in special education to better understand how the issue is framed, how researchers make use of available data sets, the statistical methods and models used, and the included covariates.

Method

A comprehensive search strategy was used in an effort to locate articles that were peer-reviewed, published, and relevant to the field of disproportionality in special education. Systematic searches were performed using the following key words: along with the connector and, the term special education was combined with disproportionality, disproportionate representation, overrepresentation, underrepresentation, and placement. The main search engines utilized included Google Scholar, EBSCO search complete, and JSTOR. As Waitoller et al.’s (2010) comprehensive literature review included a replicable search method and inclusion criteria similar to ours, we relied on its reference list for articles prior to 2006; this study searched for additional articles published in 2006 and more recent. Initial searches yielded 466 articles. Reference lists of identified articles were hand searched to identify additional studies as potentially eligible for this review, which produced five additional articles. We examined the 471 articles to determine if they met this study’s selection criteria, which included the following: (a) the study question, purpose, and hypothesis addressed the placement of students from racially and ethnically diverse backgrounds in special education generally and/or in any of the IDEA federal disability eligibility categories; (b) the study included patterns and/or predictors of students from racially and ethnically diverse backgrounds in special education generally and/or in any of the IDEA federal disability eligibility categories, and included factors that increased or decreased the likelihood of identification; (c) studies were published in peer-reviewed journals, which served to maintain the quality of research; this review excluded book chapters, technical reports, master’s theses, and dissertations; and (d) articles must have included quantitative research designs as the primary method of examination as this review was meant to examine statistical methods of understanding disproportionality patterns; articles that addressed disproportionality from a primarily qualitative or conceptual standpoint were excluded.

Studies must have met the aforementioned four criteria to be included in this review. Nine of the 35 studies included in Waitoller et al.’s (2010) review were considered, as they fit the above criteria and were pertinent to the investigation of statistical methods for examining disproportionality. In addition, this search yielded 17 studies conducted after 2006 that fit the criteria for inclusion. A total of 26 studies were included in this analysis.

Integrative Literature Review

The methodological approach to this literature search was an integrative one, which involves reviewing, critiquing, and synthesizing work on a topic such that new frameworks and perspectives emerge (Torraco, 2005). This literature review addressed a relatively mature topic, and the need for potential reconceptualization was warranted given disagreement in the field. After initial review of each study to determine inclusion, we coded each study on relevant features, discussed subsequently, and synthesized the work to provide a critique of strengths, key contributions, deficiencies, or other problematic aspects of the literature.

Each study was coded for data scope (i.e., national, state, district, community/municipality, or school), and disability type studied (i.e., intellectual disability [ID], speech or language impairment, specific learning disability [SLD], emotional disturbance [ED], Autism). Codes were limited to these disability types rather than all 13 disability categories listed under IDEA as low-incidence disabilities such as hearing impairment and visual impairment are typically not associated with disproportionality and, as such, are rarely studied. If the included studies examined disabilities other than those coded, such as low-incidence, that information is included in Table 2 for the reader’s reference. It should be noted that federal eligibility categories have changed over time, and there is some variability in labels across states. To maintain consistency in our reporting, we used the current federal labels described in IDEA as inclusive of all relevant or alternative labels.

Race categories were coded for whether the study examined differential representation between students who are Black and White, representation between students who are Latino and White, and representation between multiple race or ethnicity categories. To maintain consistency in our reporting, studies that included students who are African American or Black are heretofore referred to as Black. Studies that included students who are non-White Hispanic or Latino are heretofore referred to as Latino. The studies were also coded for if and how they controlled for SES and academic measures. If the study controlled for these features, they were coded for use of an aggregated measure or a measure that was disaggregated by the individual student and family. Studies could have used both disaggregated and aggregated measures in their analyses; as such, the codes were not mutually exclusive.

Studies were coded for decade of study, which allowed for an examination of whether the quantitative methods used in studies of disproportionality have changed over time, especially considering increased levels of diversity within schools. In addition, the analysis used in each study (i.e., risk ratio, regression, or multilevel regression) was coded to allow for an exploration of the most commonly used methods and the consistency of results within and across different analytical techniques. In this coding scheme, risk ratio included exposure odds ratios, odds ratios, relative risk ratios, and risk indices, but excluded analyses that used odds ratios as the outcome variable in a linear or logistic regression model. Studies that included a model to estimate the probably of placement or service in special education as a function of independent predictor variables were coded as regression. Studies that included multilevel linear or logistic regression (i.e., hierarchical modeling) that nested students within schools, districts, or communities to account for within- and between-cluster estimations were coded as multilevel regression. Because some studies used several analyses, these codes were not mutually exclusive; multiple codes could be applied to each study, except for the code “other,” which was applied to studies that did not use any of the aforementioned statistical analyses.

Interrater agreement

Both authors independently read and coded the 26 articles included in this review. The authors agreed on all but two items across the studies, which related to whether SES was controlled and how. For the studies where discrepancies occurred, both authors reread the articles and engaged in discussion until consensus was reached.

Results

Descriptive Characteristics

General features of the studies can be seen in Table 1. Most studies examined the issue of disproportionality with a nationally representative data set (n = 16). Five of the studies examined the issue with state-level data, three included district-level data, and two studies included data from a region or municipality. Though most of the studies used a national data set, the data source varied across studies. For example, Artiles, Aguirre-Muñoz, and Abedi (1998) used the National Education Longitudinal Study (NELS); Oswald et al. (1999) used the Elementary and Secondary School Civil Rights Compliance Report; and Hibel, Faircloth, and Farkas (2008) used the Early Childhood Longitudinal Study, Kindergarten (ECLS-K). Eight studies used the data sets provided by the U.S. Department of Education through the National Center for Education Statistics (NCES). Of those studies, two (Ong-Dean, 2006; Talbott, Fleming, Karabatsos, & Dobria, 2011) used data for one state only. Of those using the full national sample, five were published prior to 2005. The most recent study to use this data set (Travers, Krezmien, Mulcahy, & Tincani, 2014) looked solely at autism prevalence.

Table 1.

Descriptive Characteristics of Selected Studies.

Study feature	Description	Percent (n)
Data set level	National	61.54 (16)
	State	19.23 (5)
	District	11.54 (3)
	Municipality	7.69 (2)
	School	0.00 (0)
Disability type	Intellectual disability	46.15 (12)
	Speech/language impairment	26.92 (7)
	Specific learning disability	62.07 (15)
	Emotional disability	57.69 (16)
	Autism	3.85 (1)
	Multiple disabilities	57.69 (15)
Race/ethnicity	African American	100.00 (26)
	Latino	80.77 (21)
	Multiple race categories	69.23 (18)
Decade published	1990–1999	15.38 (4)
	2000–2009	34.62 (9)
	2010–Present	50.00 (13)
Controls for SES	Community aggregate	38.46 (10)
	Disaggregate	42.31 (11)
	Not controlled for	19.23 (5)
Controls for academics	Community aggregate	11.54 (3)
	Disaggregate	19.23 (5)
	Not controlled for	69.23 (18)
Analysis used	Risk ratio	57.69 (15)
	Linear or logistic regression	76.92 (20)
	Multilevel regression	30.77 (8)

Note. SES = socioeconomic status.

Studies most often examined high-incidence disabilities such as SLD and ED, as these categories typically rely more heavily on practitioner judgment and have historically represented a higher level of disproportionality (Bal, Sullivan, & Harper, 2014). Half (n = 13) of the studies examined disproportionality in the area of ID. Prior to 2000, studies examined disproportionality in only one or two disability categories, and those categories were either ED, ID, or SLD. After 2000, studies more often examined disproportionality across multiple disability categories. Ten of the 22 studies published after 2000 included all three high-incidence categories (i.e., SLD, ID, ED) in their analyses. Four of the studies in this review used receipt of special education services as a binary outcome variable and did not specify categories under the special education umbrella. It should be noted that studies that include receipt of or referral for special education generally may produce biased estimates as they include low-incidence categories, which are not typically associated with disproportionality.

When examining race and disability, all studies within the review used “White” as a reference category. Four studies exclusively examined the gap between White and Black students; no other racial categories were studied exclusively. All of the studies in this review included Black as one of the racial categories under study, 21 of the 26 studies included Latino as a racial category, and 18 of the studies included additional racial categories such as Asian, American Indian/Alaska Native, and Pacific Islander. It should be noted that the racial category labels were determined by the data source, and, consequently, varied from study to study. In Hibel et al. (2008), for example, the data source used for analysis was the ECLS-K, which included six racial/ethnic categories: American Indian/Alaska Native, non-Latino White, Black, Latino, Asian, and “other.” Yeh, Forness, Ho, McCabe, and Hough (2004) made use of the Patterns of Youth Mental Health Care in Public Service Systems data set, which included youths who were non-Hispanic White, African American, Asian/Pacific Islander American, or Latino. The inconsistent use of racial/ethnic labels across studies, as well as the limiting nature of the labels themselves, especially for students who may identify as multiple races/ethnicities, suggests a need for caution when interpreting patterns across groups.

Half (n = 13) of the studies included in this review were published within the last 7 years and half were published prior to 2010. There has been increased attention paid to this issue, but, interestingly, there is an intersection of decade of study, race categories studied, disability categories included, and type of analysis. As aforementioned, studies published prior to 2000 tended to focus exclusively on one disability category, and they focused almost exclusively on Black students. These studies also utilized less complex statistical analyses. Between 2002 and 2007, studies most often examined multiple race categories and did so using odds ratios and logistic regression. These studies also tended to control for SES using aggregated, community-level variables. Studies published since 2008 more often examined multiple racial categories across multiple disability categories, and did so using multilevel modeling. In terms of controls, it was more common to control for SES (n = 22) than to control for academic measures (n = 7), whether controls were aggregated or not, and more recent studies have controlled for SES using disaggregated individual/family measures.

Although many studies had access to data for multiple years, only three studies used a longitudinal model of examination, and only eight studies included a multilevel model. This research topic is especially suited to a multilevel model, as results of many studies highlighted school-, district-, and state-level confounding effects and students are nested within schools and communities (Raudenbush & Bryk, 2002). In addition, some results suggested that early childhood representation might be less prone to disproportionality issues, as lower incidence disabilities are often diagnosed in earlier years, while older students are often identified for SLD and ED (Morrier & Gallagher, 2012). Thus, longitudinal methods of examination may be more appropriate for future research in this area.

Key Themes in Reported Outcomes

Study characteristics and the key findings related to race are presented in Table 2. Bivariate results suggest that race is a strong predictor of disproportionality in identification for special education. Studies consistently found that Black students were overrepresented in categories such as ED and ID, and Asian students were underrepresented in all disability categories when compared with White counterparts. Bivariate results for Latino students were quite mixed, most likely because this demographic is relatively heterogeneous (see Mora, 2014 and Rumbaut, 2005 for information regarding Latino panethnicity and cultural identification within the category of “Latino”). Multivariate analyses that considered systematic differences between and within sociodemographic groups produced mixed results. SES, academic measures, and age of diagnosis seem to have a unique and often counterintuitive effect on results. In addition, considerations for academic measures, what they measure, and how they might relate to disability status is a particularly puzzling part of the equation. Each of these key features is discussed below.

Table 2.

Study Characteristics in Chronological Order.

Author(s), year	Type of disability	Race categories	Type of analysis	Data set, level	Key findings related to race
Serwatka, Deering, and Grant (1995)	ED	Black/White	Pearson product moment correlation	Florida Department of Education; state (Florida)	As population of Black students increased, there was a decrease in overrepresentation. Also, as Black teachers increased in schools, there was a decrease in overrepresentation.
Yeargin-Allsopp, Drews, Decoufle, and Murphy (1995)	ID	Black/White	Exposure odds ratios	Metropolitan Atlanta Developmental Disabilities Study; municipality (Atlanta)	Prevalence of ID was higher for Black children than for Whites, but odds reduced when controlling for SES, and even more when controlling for all other covariates (mom’s age and education level, sex of the child). No difference in ID identification between Black and White students prior to age 6; the disparity was the largest when diagnosed between 8 and 10 years old.
Artiles, Aguirre-Muñoz, and Abedi (1998)	SLD	Black/White, Latino/White	ANOVA	National Education Longitudinal Study (NELS); national	Results showed significant main effects for ethnicity. A main predictor for all of the three groups (Latino, Black, and White) included family structure and size. These variables were highly correlated with student perception of academic standing and locus of control.
Oswald, Coutinho, Best, and Singh (1999)	ID, ED	Black/White	Odds ratios, logistic regression	1992 Elementary and Secondary School Civil Rights Compliance Report Survey	Black students were nearly two and a half times as likely to be identified as having ID, and one and a half times as likely to be identified as having ED compared with non-African American peers. When including environmental factor variables into the model, race-related factors were more influential to identification rates of African American students than non-African American peers.
Coutinho, Oswald, and Best (2002)	SLD	Multiple race categories	Odds ratios, logistic regression	U.S. Department of Education Office of Civil Rights; NCES	As poverty increased, SLD identification rates among Black, Latino, and male Asian students increased, and SLD identification rates for White and American Indian students decreased. For all gender/racial groups, SLD identification rates decreased as the proportion of minority students in a district increased.
Coutinho, Oswald, Best, and Forness (2002)	ED	Multiple race categories	Odds ratios, logistic regression	U.S. Department of Education Office of Civil Rights and NCES	Poverty was positively associated with ED identification rates across racial groups. Disproportionality for minority students increased as poverty increased. Black students who lived in predominately White communities were more likely to be identified as having ED than those living in communities with a greater minority population.
Zhang and Katsiyannis (2002)	SLD, ED, ID	Multiple race categories	ANOVA, Pearson correlation coefficient	U.S. Department of Education data, national	Across all states, Black and American Indian/Alaskan students were overrepresented in SLD, ED, and ID; Latino and Asian students were underrepresented in these categories. Disproportionality by race in each category varied by region. In states with high poverty rates, fewer Latino students were identified for SLD and fewer Black, American Indian/Alaskan, Asian, and White students were identified as having ED.
Hosp and Reschly (2004)	SLD, ED, ID	Multiple race categories	Relative risk ratios, least squares regression	Elementary and Secondary Schools Civil Rights Compliance Report, Common Core of Data, district-level achievement data from each state’s Department of Education; national	The economic block was the strongest predictor of ID across racial/ethnic groups, the demographic block was the strongest predictor for ED, and the academic block was the strongest predictor for SLD. Overall, both the economic and demographic blocks predicted disproportionate representation across racial/ethnic groups, and the academic block was the weakest as a predictor.
Yeh, Forness, Ho, McCabe, and Hough (2004)	ED	Multiple race categories	Odds ratios, logistic regression	Patterns of Youth Mental Health Care in Public Service Systems (Patterns of Care); municipality	Latino and Asian/Pacific Islander children were underrepresented in ED. Parents of Black and Asian/Pacific Islander children were more likely than White parents to see prejudice as related to their children’s difficulties. Parental beliefs about the causes (i.e., personality and relational difficulties) of their children’s problems partially accounted for the relationship between race and ED identification for Asian/Pacific Islander and Latino youths.
Skiba, Poloni-Staudinger, Simmons, Feggins-Azziz, and Chung (2005)	ID, ED, SLD, and SLI	Black/White	Odds ratios, ordinary least squares regression and logistic regression	Indiana Department of Education; district	When controlling for poverty, Black students were more likely to be identified as ID and ED, but less likely to be identified as SLI and SLD. Race was more predictive of special education identification than low income across disability categories.
Ong-Dean (2006)	SLD	Black/White, Latino/White	Odds ratios, logistic regression	Office of Civil Rights Compliance Report; state (California)	In 1976, Latino and Black students were less likely to be identified for SLD than White students. By 1998, Latino students were nearly as likely as White students to be identified as SLD, and Black students were much more likely.
Mann, McCartney, and Park (2007)	Receipt of special education services	Black/White	Odds ratios, ANOVA, multinomial logistic regression	Early Child Care and Youth Development; national and longitudinal	After including sociodemographic covariates, race/ethnicity was not associated with either special education or remedial services.
Hibel, Faircloth, and Farkas (2008)	Receipt of special education services	Multiple race categories	Odds ratios, multilevel modeling	ECLS-K; national	When controlling for kindergarten test scores, Black, Latino, and Asian students were significantly underrepresented in special education, and the overrepresentation of AI/AN students was no longer statistically significant. An increase in a school’s proportion of minority students (not including AI/AN) was associated with lower odds of placement in special education. An increase in AI/AN student enrollment in a school did not have a significant effect on the odds of AI/AN students being placed in special education. After controlling for school, family, and individual characteristics, Black, Latino, and Asian students had lower odds of special education placement in third grade than White students.
Hibel, Farkas, and Morgan (2010)	Placement with any disability, SLD, SLI, and ID	Multiple race categories	Multilevel modeling	ECLS-K; national	When controlling for SES, Latino and Asian students were less likely to be placed in special education generally. When controlling for student’s initial level of academic achievement, Black, Latino, and Asian students were significantly underrepresented in special education, though some of these effects were reduced in magnitude after adding school-level variables to the model.
Morrier and Gallagher (2012)	SLD, SLI, ID, ED, and additional low-incidence categories	Multiple race categories	Risk ratios, ANOVA	U.S. Department of Education and U.S. Census Bureau (five southern states). Preschool students only.	American Indian preschoolers were overrepresented in one of the states. Asian, Latino, and White preschoolers were represented proportionately. Black preschoolers were overrepresented in only one of the five states. The authors noted that aggregating data across states may mask state-by-state variation in disproportionality.
Shifrer, Muller, and Callahan (2011)	SLD	Multiple race categories	Odds ratios, multivariate, multilevel modeling	Education Longitudinal Study (ELS); national	When controlling for SES (parent education and family income), all significant race effects were explained except for lower odds of identification for Asian students.
Sullivan and Artiles (2011)	ID, SLD, SLI, ED, and additional low-incidence categories	Multiple race categories	Relative risk ratios, ordinary least squares (OLS) regression	Arizona Department of Education, state and district (Arizona)	Substantial disproportionality was not observed at the state level, though risk was elevated for Native American and Black students in some categories. At the district level, Black and Native American students had elevated risk in ID and SLD; Latino students were underrepresented in many categories. The authors noted that identification practices in a small number of large districts drove state-level patterns. Proportion of students receiving free or reduced lunch was not a significant predictor in regression analysis except for predicting the relative risk of Black students identified as SLD and in special education. As district minority enrollment increased, minority students’ relative risk of identification decreased.
Talbott, Fleming, Karabatsos, and Dobria (2011)	ID, ED, SLD	Multiple race categories.	Multilevel linear modeling	Office of Civil Rights; state (Illinois)	Once students were nested within schools and school-level variables were included in the model, race did not significantly predict student identification in ID, ED, or SLD.
Morgan, Farkas, Hillemeier, and Maczuga (2012)	Early Intervention or Early Childhood Special Education	Multiple race categories	Odds ratios, logistic regression	Early Childhood Longitudinal Study—Birth Cohort (ECLS-B), Included measures for children at 48 months only	After controlling for SES, sociodemographic factors, residential characteristics, gestational and birth-related factors, numeracy and receptive language scores, externalizing and internalizing problem behaviors, and low number of well-child checkups, Black and Asian children were significantly underrepresented in EI/ECSE.
Bal, Sullivan, and Harper (2014)	SLD, ED ID, OHI and additional low-incidence categories	Multiple race categories	Risk indices and relative risk ratios	Flen School District’s (Wisconsin) information system; district	Black and American Indian students had elevated relative risk compared with White students in each of the high-incidence categories but not in the low-incidence categories across all years studied. Latino students had generally lower relative risk in high-incidence categories.
Sullivan and Bal (2013)	SLD, ID, ED, SLI, OHI, and low-incidence categories	Multiple race categories	Risk indices, multilevel logistic regression	One school district in Wisconsin	When controlling for student free/reduced-price lunch and parent education, Black students were overrepresented for SLD only and Latino students were underrepresented for SLD. Asian/pacific islanders were underrepresented in special education generally.
Wiley, Brigham, Kauffman, and Bogan (2013)	ED	Black/White, Latino/White	Risk ratios, Pearson product correlation, logistic regression	Office of Special Education Programs (OSEP) and Children’s Defense Fund; national	Higher rates of child poverty were associated with lower rates of Black and Latino students identified as ED. Disproportionate poverty explained 36% of the variance in ED risk ratios for Latino students. Overall, the authors suggest that poverty and underidentification may cause disproportionate representation of minority students in the ED category.
Travers, Krezmien, Mulcahy, and Tincani (2014)	Autism	Black/White, Latino/White	Odds ratios, logistic regression	IDEA Data Accountability Center and NCES; national	Black and Latino students were significantly less likely than White students to be identified with autism. Prevalence rates by state and by racial category varied substantially.
Zhang, Katsiyannis, Ju, and Roberts (2014)	SLD, ID, and ED	Multiple race categories	Multilevel modeling, growth models	U.S. Census Bureau’s Small Area Income and Poverty Estimates; U.S. Department of Education Data Center; national and longitudinal	Black and American Indian students had the highest rates in SLD, ID, and ED. Black students had greater representation in ID than all other racial groups, though there was a decrease over time. There was an increase in Latino students identified as SLD over time. In poorer states, Black, Latino, and White students’ identification as SLD decreased, and Black and White students’ representation in ED and ID decreased.
Anderson, Howland, and McCoach (2015)	Receipt of special education services	Multiple race categories.	Multilevel multivariate regression	ECLS-K; national and longitudinal	Race/ethnicity was not associated with identification rates with the exception of Asian students, who were less likely to be identified for special education.
Kincaid and Sullvian (2017)	High-incidence disability (SLD, ID, or ED)	Multiple race categories.	Odds ratios, multilevel modeling	ECLS-K; national	When controlling for SES using tripartite measures, there was no evidence of racial disproportionality except that Asian students were significantly underrepresented.

Note. ED = emotional disturbance; ID = intellectual disability; SES = socioeconomic status; SLD = specific learning disability; ANOVA = analysis of variance; NCES = National Center for Education Statistics; SLI = speech/language impairment; ECLS-K = Early Childhood Longitudinal Study; AI/AN = American Indian/Alaska Native; OHI = other health impairment; IDEA = Individuals With Disabilities Education Act.

Socioeconomic status

In the years leading up to 2000, studies that controlled for SES often used aggregated community-level variables. For example, Oswald et al. (1999) used median housing value for a community-level SES measure, and found that Black students were 2.5 times more likely to be identified as having ID and 1.5 times more likely to be identified as ED than other students. They found that within a community with lower poverty rates, overrepresentation of Black students in ID was actually higher. Additional studies that controlled for SES in a similar way found that as a community or state became more affluent, students from racially and ethnically diverse backgrounds became more disproportionately labeled as having a disability (Coutinho et al., 2002; Zhang & Katsiyannis, 2002; Zhang et al., 2014).

When attaching SES measures to the student or family, results showed that coming from a family with fewer resources may increase one’s odds of identification for special education, but the extent to which SES accounts for disproportionality is unclear. When free/reduced-price lunch was used as a measure of individual/family disaggregated SES, studies showed that Black students had elevated risk in high-incidence categories (Bal et al., 2014; Sullivan & Bal, 2013). Studies that controlled for SES using disaggregated student/family variables like parent income and education often produced results that eliminated race as a predictor for disability identification or showed that some racial/ethnic categories were actually underrepresented. Mann, McCartney, and Park (2007) found that race/ethnicity was not associated with either special education or remedial services after including sociodemographic covariates into the multinomial logistic regression model; Hibel et al. (2008) found that Black, Latino, and Asian students were significantly underrepresented in special education; and Shifrer, Muller, and Callahan (2011) found that when controlling for parent education and family income, all significant race effects were explained except for lower odds of identification for Asian students. Sullivan and Bal (2013) found that when controlling for student free/reduced-price lunch and parent education, Black students were overrepresented for SLD, Latino students were underrepresented for SLD, and Asian/Pacific Islander students were underrepresented in special education generally.

Often, free and reduced lunch data are the only way to measure SES, but some studies noted that this might not be the most precise measure of SES due to administrative issues that reduce accuracy or reduced participation because of issues with stigma associated with such programs (Sullivan & Artiles, 2011). National data sets, such as the National Educational Longitudinal Study (NELS, 1988) and Early Childhood Longitudinal Study (ECLS, 1998), allowed for use of a composite SES indicator using such measures as family income and parent education levels, and these measures were disaggregated at the student level. Studies using these measures also indicated that students in lower socioeconomic strata were more likely to be identified for special education than higher SES counterparts, but race was often not a predictor of disability identification. For example, Kincaid and Sullivan’s (2017) study used tripartite measures of SES, including parent income, educational attainment, and occupational prestige and found that when all other net controls were considered, the only statistically significant predictor of special education identification in high-incidence categories was parent educational attainment.

The ways in which each study conceptualized SES varied, and, thus, results varied. When studies used aggregated measures for SES, they tended to report overrepresentation of students from racially and ethnically diverse backgrounds in more affluent areas. When considering SES by disaggregated free and reduced lunch measures, results were mixed. When using more specific continuous or composite indicators, much of the variability in special education identification could be attributed to SES. This provides evidence of the complex ways in which economic disadvantage can affect the development of a child, including the access a student has to family, school, and community resources, and the importance of examining context in the study of disproportionality.

Academic factors

Many studies did not include an academic measure as a covariate. It is reasonable to assume that a student having lower performance on academic measures such as standardized test scores and grades would be associated with higher propensity to be placed in special education when compared with a student with higher standardized test scores, as it is required by law that a student’s disability adversely affect educational performance in order for the student to be eligible for special education (IDEA, 2004). In fact, Skiba, Poloni-Staudinger, Simmons, Feggins-Azziz, and Chung (2005) state that

. . . to prove that poverty contributes significantly to special education disproportionality, it would be necessary to show that economic disadvantage increases the risk not merely of underachievement but also of the specific types of learning and behavior problems defined by IDEA as disability. (p. 132)

This further confounds the use of explicitly conceptualized measures of academic variables as these measures do not directly measure disability, and one might erroneously assume that a student at risk for academic failure is at risk because of a disability, and not because of lack of opportunity and access. This presents a challenge in that the use of academic measures in studies examining disproportionality are subject to effects that are highly collinear with the outcome variable of special education placement.

A few studies explored the use of academic measures that included kindergarten readiness cognitive scores. Morgan, Farkas, Hillemeier, and Maczuga (2012) found that lower SES was strongly related to lower levels of preacademic knowledge, and, all net controls considered, Black and Asian children were statistically underrepresented in special education when considering the ECLS-K measures of reading and math readiness upon entering kindergarten. The authors posit that because these early academic measures are a significant predictor of fifth-grade special education placement, students from racially and ethnically diverse backgrounds are actually underrepresented in special education. However, this does not explain the ways in which these young children come to be in this risk category in the first place, nor does it equate to students from racially and ethnically diverse backgrounds having higher incidence of disability. It merely highlights the fact that upon entering kindergarten with fewer academic skills, students often struggle to gain ground on more affluent and experienced peers. This is known as the Matthew Effect, and provides more evidence of the societal ways in which inequities are reproduced in our schools (Van Steensel, 2006). Several additional studies discussed the need for using academic measures; however, endogeneity was listed as a concern (Sullivan & Bal, 2013), as well as lacking a standardized, norm-referenced measure of academic standing that is culturally relevant, reliable, and validated for the group of students being measured (Rueda & Windmueller, 2006).

Age of diagnosis

As aforementioned, few studies have examined the issue of disproportionality from a longitudinal perspective. Hibel et al. (2010) studied a cohort of children in kindergarten to examine special education placement levels at fifth grade, and found lower levels of disproportionality than studies examining data sets that included students from kindergarten through high school. This shorter timeline may account for some unexplained differences in disproportionality rates, such as students identified for SLD in middle school.

Yeargin-Allsopp, Drews, Decouflé, and Murphy (1995) studied overrepresentation of Black students in the category of ID, and matched family address to neighborhood income averages. The study controlled for family demographics, such as mother’s age at the time of birth and birth order of the child, and assessed whether a child born in a low-income community might be more susceptible to mild cognitive impairments, making their risk for special education higher. The study found that prevalence of ID was higher among Black children, but odds were mitigated when controlling for SES. In addition, the authors found “little difference in the prevalence of mild mental retardation between Black and White children diagnosed before reaching school age, but Black children were overrepresented among children diagnosed after they entered school” (p. 327). As this study only examined patterns in ID, but found differences in disproportionality rates when considering a student’s age at diagnosis, there is a need for further longitudinal study among disabilities that are typically found in later years of schooling, especially SLD. Longitudinal studies would better enable researchers to understand whether or not disproportionality varies at different ages of diagnosis and with various disability categories.

The Need to Examine Local Contexts

Review of the included studies shed light on the stark contrast between studies that used a national data set and hierarchical modeling as opposed to those that examined district-level data with a focus on local contexts. For studies that used a national data set and multilevel modeling, which allows for an estimation of the unique influences of the school environment, race was most often not a significant predictor of identification in special education categories under study, and some racial/ethnic groups were found to be underrepresented in special education. Sullivan and Bal (2013) used multivariate hierarchical models, but did so with one school district in Wisconsin and found that Black students were overrepresented for SLD. In a recent study by Morgan et al. (2017), it was noted that Sullivan and Bal’s (2013) finding of overrepresentation of Black children in the SLD category in a school district with high parental education levels was unrepresentative, and the sample was considered an “unusual school district” (p. 191). We would argue that overrepresentation in any school district should not be overlooked considering the implications misidentification may have for students from racially and ethnically diverse backgrounds, and studies that use district-level data are perhaps better suited to elucidate the interplay of contextual factors.

Though studies with national data sets and hierarchical models are able to account for state, district, and school-level factors, there are limitations to the use of national data sets, as noted in several studies. First, large school districts are included in national data sets, but only a sample of smaller districts, which may mask district-by-district variation. Hibel et al. (2010) noted that the ECLS-K has substantial missing data, which required multiple imputation procedures to retain cases in the sample. In addition, the authors noted that ID and ED are diagnosed less frequently, making “a reliable application of inferential statistical methods problematic” (p. 320). An examination of data at the national level certainly provides valuable information that contributes to the ongoing discussion of equity in schools, but the results are less helpful in understanding the lived experiences of students and practitioners within their local contexts.

It is also important to consider that disability definitions and interpretations are not static. Only one study in this review discussed the changing nature of federal categories over time. Using a sociohistorical perspective of disability and diagnosis, Ong-Dean (2006) noted that Latino and Black students were less likely to be identified for SLD than White students in 1976, but Black students were much more likely to be identified as SLD than White students in 1998. Perspectives on disability, the meaning attached to labels, and cultural attitudes toward disability and services are dynamic and situated within local contexts. Policies change, as do local adaptations of policy and programs. As so aptly stated by Bal et al. (2014), “Researchers who aim to address disproportionality should study local professional structures, practices, and perspectives . . . [including] how students . . . , families, and teachers negotiate, orchestrate, and innovate within their immediate contexts” (p. 10). Hibel et al. (2010) also discuss the teacher’s referent group as normative, and teacher judgments of student achievement and behavior are relative to the performance of peers to which the teacher has been exposed. Thus, the study of local contexts, especially those that highlight adaptations of policy and processes over time, may do more to inform potential reform.

Discussion

As evidenced by the studies in this review, there are a range of factors in a research design that may impact results, and, thus, our understanding of disproportionality. It is difficult to make conclusive statements regarding disproportionality by race and ethnicity because studies use a variety of data sets at different levels, include samples of students at different ages and grades, and apply statistical analyses that may produce conflicting results. Though studies may appear to diverge on the surface, we would argue that each serves to provide a piece of information that contributes to the larger picture of who we refer and serve in special education. The various results produced in each study may in fact reflect that the issue of disproportionality is not a simple problem with a simple solution. There are several race categories with subgroups of individuals who represent a variety of ethnic and cultural identities within the larger category. The high-incidence disability categories are each different, and the labels are not defined, interpreted, or assessed in the same way across states, districts, and schools. The relationship between families, school professionals, students, and teachers may look vastly different from one school to another, and even from one classroom to another. The studies in this review highlight that overrepresentation for one racial group in one disability category may occur in one state, and underrepresentation may occur for that same racial group in a different disability category in a different locale.

It is also evident from this review that any study on disproportionality needs a theoretical framework, especially as an impetus for selecting and operationalizing variables within the analytical design. If, for example, one values understanding the ways in which students experiencing poverty function within a community, community-level covariates must be considered in the system of analysis. If a researcher believes that special education classrooms continue to focus on segregation and otherness, and that students experiencing this label are treated to lower expectations and stigma, then an analysis of disproportionality must examine the historical contexts of inequity that place already marginalized students at a disadvantage over more affluent, mostly White peers. In this context, understanding that bivariate risk ratio results show high levels of disproportionality should be alarming regardless of covariates, as it may lead to de facto segregation and further widening of opportunities to learn in a typical classroom setting.

Understanding the difference between biological differences and the social construction of our interpretation of differences as disabling may best be addressed by examining larger social problems, such as the ways in which socioeconomic inequality is reproduced in our nation’s schools. Moreover, historical and macro-level forces that intersect in the education arena must be considered at a more fine-grained level in future research. It is clear that the massive current body of research on the topic of disproportionality in special education remains insufficient for our understanding of the ways in which special education laws and policies might be tailored to either perpetuate or truncate racial and socioeconomic inequities in education.

Limitations

This review did not include studies that examined disproportionality of students who are linguistically diverse. There are numerous studies that explore issues related to the identification and placement of linguistically diverse students in special education, and those studies contribute to the discussion of equity in our school systems. While this review focused on culturally diverse students, the discourse surrounding disproportionality should also include referral and assessment of linguistically diverse students for high-incidence categories, especially SLD.

This review is limited in that it did not include books, book chapters, dissertations, or other published works apart from peer-reviewed journal articles. This review also did not include qualitative studies, which may be especially helpful in better understanding root causes. Additional information related to disproportionality is certainly available from a wide range of sources. Also, the presentation of key findings across studies was limited to race and focused on the covariates of SES and academics. Many of the studies in this review were quite complex and included additional community- and school-level factors that uniquely contribute to risk of identification. Discussing all findings for each study was beyond the scope of this review, but future reviews could include information on additional variables and their contribution to outcomes.

Future Directions

Future research in the area of disproportionality should include longitudinal studies that are able to capture the various points at which a student may enter or exit special education, and should also include an examination of local contexts. As aforementioned, studies should be grounded in a theoretical or conceptual framework that provides a rationale for the selected analytic design and included covariates. In addition, the ways in which each covariate is operationalized should be made explicit. Academic measures, for example, are a relatively recent addition to study designs, and, should a study use an academic measure as a control, the justification for the inclusion of the measure should be well articulated. We would especially caution against the inclusion of academic measures without a strong rationale considering the fact that low academic performance does not equate to disability. There are a variety of reasons why a student may be underperforming academically, and we cannot assume that students who are at-risk academically or behaviorally, especially in their early years of learning, have a disability and should receive special education services.

Given the mixed results related to racial/ethnic categories across high-incidence categories, and the increasingly complex statistical approaches related to the study of disability, we should consider the ways in which to study this issue that honors the experiences of the individuals who live within it. Our questions should dictate our methods, and we should not fear simpler methods that can provide elegant answers to our important questions (Peterson, 2009).

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Anderson

J. A.

Howland

A. A.

McCoach

D. B.

(2015). Parental characteristics and resiliency in identification rates for special education. Preventing School Failure: Alternative Education for Children and Youth, 59, 63–72. doi:10.1080/1045988X.2013.837811

Artiles

A. J.

Aguirre-Muñoz

Abedi

(1998). Predicting placement in learning disabilities programs: Do predictors vary by ethnic group? Exceptional Children, 64, 543–559. doi:10.1177/001440299806400409

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. doi:10.1177/002246699402700404

Bal

Sullivan

A. L.

Harper

(2014). A situated analysis of special education disproportionality for systemic transformation in an urban school district. Remedial and Special Education, 35, 3–14. doi:10.1177/0741932513507754

Carter

P. L.

Reardon

S. F.

(2014). Inequality matters (William T. Grant Foundation paper). New York, NY: William T. Grant Foundation.

Coutinho

M. J.

Oswald

D. P.

(2000). Disproportionate representation in special education: A synthesis and recommendations. Journal of Child and Family Studies, 9, 135–156.

Coutinho

M. J.

Oswald

D. P.

Best

A. M.

(2002). The influence of sociodemographics and gender on the disproportionate identification of minority students as having learning disabilities. Remedial and Special Education, 23, 49–59. doi:10.1177/074193250202300107

Coutinho

M. J.

Oswald

D. P.

Best

A. M.

Forness

S. R.

(2002). Gender and sociodemographic factors and the disproportionate identification of culturally and linguistically diverse students with emotional disturbance. Behavioral Disorders, 27, 109–125.

Diana v. State Board of Education, Civil Action No. C-70-38 (N.D. Cal., 1970, further order, 1973).

10.

Donovan

M. S.

Cross

C. T.

(Eds.). (2002). Minority students in gifted and special education. Washington, DC: National Academy Press.

11.

Dunn

L. M.

(1968). Special education for the mildly retarded: Is much of it justifiable? Exceptional Children, 35, 5–22.

12.

Hibel

Faircloth

Farkas

(2008). Unpacking the placement of American Indian and Alaska native students in special education programs and services in the early grades: School readiness as a predictive variable. Harvard Educational Review, 78, 498–528. doi:10.17763/haer.78.3.8w010nq4u83348q5

13.

Hibel

Farkas

Morgan

P. L.

(2010). Who is placed into special education? Sociology of Education, 83, 312–332. doi:10.1177/0038040710383518

14.

Hosp

J. L.

Reschly

D. J.

(2004). Disproportionate representation of minority students in special education: Academic, demographic, and economic predictors. Exceptional Children, 70, 185–199. doi:10.1177/001440290407000204

15.

Individuals With Disabilities Education Act Data Center. (2015, April). Navigating coordinated early intervening services: Frequently asked questions. Rockville, MD: Westat.

16.

Individuals With Disabilities Education Improvement Act of 2004, 20 U.S.C. § 1400 (2004).

17.

Kincaid

A. P.

Sullivan

A. L.

(2017). Parsing the relations of race and socioeconomic status in special education disproportionality. Remedial and Special Education, 38, 159–170. doi:10.1177/0741932516671199

18.

Larry P. v. Riles, 343 F. Supp. 1306 (N.D. Cal. 1972), aff’d, 502 F.2d 963 (9th Cir. 1974); 495 F. Supp. 926 (N.D. Cal. 1979), aff’d, 793 F.2d 969 (9th Cir. 1984).

19.

Mann

E. A.

McCartney

Park

J. M.

(2007). Preschool predictors of the need for early remedial and special education services. The Elementary School Journal, 107, 273–285. doi:10.1086/511707

20.

Mora

G. C.

(2014). Cross-field effects and ethnic classification: The institutionalization of Hispanic panethnicity. American Sociological Review, 79, 183–210.

21.

Morgan

P. L.

Farkas

(2016). Are we helping all the children that we are supposed to be helping? Educational Researcher, 45, 226–228.

22.

Morgan

P. L.

Farkas

Cook

Strassfeld

N. M.

Hillemeier

M. M.

Pun

W. H.

Schussler

D. L.

(2017). Are black children disproportionality overrepresented in special education? A best-evidence synthesis. Exceptional Children, 83, 181–198. doi:10.1177/0014402916664042

23.

Morgan

P. L.

Farkas

Hillemeier

M. M.

Maczuga

(2012). Are minority children disproportionately represented in early intervention and early childhood special education? Educational Researcher, 41, 339–351.

24.

Morgan

P. L.

Farkas

Hillemeier

M. M.

Mattison

Maczuga

Cook

(2015). Minorities are disproportionately underrepresented in special education: Longitudinal evidence across five disability conditions. Educational Researcher, 44, 278–292.

25.

Morrier

M. J.

Gallagher

P. A.

(2012). Racial disparities in preschool special education eligibility for five southern states. The Journal of Special Education, 46, 152–169.

26.

Musu-Gillette

de Brey

McFarland

Hussar

Sonnenberg

Wilkinson-Flicker

(2017). Status and trends in the education of racial and ethnic groups (NCES 2017-051). National Center for Education Statistics.

27.

National Education Association. (2007). Truth in labeling: Disproportionality in special education. Retrieved from http://www.nea.org/assets/docs/HE/EW-TruthInLabeling.pdf

28.

Ong-Dean

(2006). High roads and low roads: Learning disabilities in California, 1976–1998. Sociological Perspectives, 49, 91–113. doi:10.1525/sop.2006.49.1.91

29.

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. doi:10.1177/002246699903200401

30.

PASE v. Hannon, 506 F. Supp. 831 (D.N.D.I. 1980).

31.

Peterson

(2009). Minimally sufficient research. Perspectives on Psychological Science, 4(1), 7–9. doi:10.1111/j.1745-6924.2009.01089.x

32.

Raudenbush

S. W.

Bryk

A. S.

(2002). Hierarchical linear models: Applications and data analysis methods (Vol. 1). SAGE.

33.

Rocha

R. R.

Hawes

D. P.

(2009). Racial diversity, representative bureaucracy, and equity in multiracial school districts. Social Science Quarterly, 90, 326–344. doi:10.1111/j.1540-6237.2009.00620.x

34.

Rueda

Windmueller

M. P.

(2006). English language learners, LD, and overrepresentation: A multiple-level analysis. Journal of Learning Disabilities, 39, 99–107. doi:10.1177/00222194060390020801

35.

Rumbaut

R. G.

(2005). Sites of belonging: Acculturation, discrimination, and ethnic identity among children of immigrants. In Weiner

T. S.

(Ed.), Discovering successful pathways in children’s development: Mixed methods in the study of childhood and family life (pp. 111–164).

36.

Serwatka

T. S.

Deering

Grant

(1995). Disproportionate representation of African Americans in emotionally handicapped classes. Journal of Black Studies, 25, 492–506.

37.

Shifrer

Muller

Callahan

(2011). Disproportionality and learning disabilities: Parsing apart race, socioeconomic status, and language. Journal of Learning Disabilities, 44, 246–257. doi:10.1177/0022219410374236

38.

Skiba

R. J.

(2012). CCBD’s position summary on federal policy on disproportionality in special education. Behavioral Disorders, 38, 108–120.

39.

Skiba

R. J.

Artiles

A. J.

Kozleski

E. B.

Losen

D. J.

Harry

E. G.

(2016). Risks and consequences of oversimplifying educational inequities: A response to Morgan et al.(2015). Educational Researcher, 45, 221–225.

40.

Skiba

R. J.

Poloni-Staudinger

Simmons

A. B.

Feggins-Azziz

L. R.

Chung

C. G.

(2005). Unproven links: Can poverty explain ethnic disproportionality in special education? The Journal of Special Education, 39, 130–144.

41.

Sleeter

(2010). Why is there learning disabilities? A critical analysis of the birth of the field in its social context. Disabilities Studies Quarterly, 30, 210–235.

42.

Sullivan

A. L.

Artiles

A. J.

(2011). Theorizing racial inequity in special education: Applying structural inequity theory to disproportionality. Urban Education, 46, 1526–1552. doi:10.1177/0042085911416014

43.

Sullivan

A. L.

Bal

(2013). Disproportionality in special education: Effects of individual and school variables on disability risk. Exceptional Children, 79, 475–494.

44.

Talbott

Fleming

Karabatsos

Dobria

(2011). Making sense of minority student identification in special education: School context matters. International Journal of Special Education, 26, 150–170.

45.

Torraco

R. J.

(2005). Writing integrative literature reviews: Guidelines and examples. Human Resource Development Review, 4, 356–367. doi:10.1177/1534484305278283

46.

Travers

J. C.

Krezmien

M. P.

Mulcahy

Tincani

(2014). Racial disparity in administrative autism identification across the United States during 2000 and 2007. The Journal of Special Education, 48, 155–166.

47.

Van Steensel

. (2006). Relations between socio-cultural factors, the home literacy environment and children’s literacy development in the first years of primary education. Journal of Research in Reading, 29, 367–382. doi:10.1111/j.1467-9817.2006.00301.x

48.

Waitoller

Artiles

A. J.

Cheney

(2010). The miner’s canary: A review of overrepresentation research and explanations. Journal of Special Education, 44, 29–49. doi:10.1177/0022466908329226

49.

Wiley

A. L.

Brigham

F. J.

Kauffman

J. M.

Bogan

J. E.

(2013). Disproportionate poverty, conservatism, and the disproportionate identification of minority students with emotional and behavioral disorders. Education and Treatment of Children, 36(4), 29–50. doi:10.1353/etc.2013.0033

50.

Yeargin-Allsopp

Drews

C. D.

Decouflé

Murphy

C. C.

(1995). Mild mental retardation in black and white children in metropolitan Atlanta: A case-control study. American Journal of Public Health, 85, 324–328. doi:10.2105/AJPH.85.3.324

51.

Yeh

Forness

S. R.

McCabe

Hough

R. L.

(2004). Parental etiological explanations and disproportionate racial/ethnic representation in special education services for youths with emotional disturbance. Behavioral Disorders, 29, 348–358.

52.

Zhang

Katsiyannis

(2002). Minority representation in special education: A persistent challenge. Remedial and Special Education, 23, 180–187. doi:10.1177/07419325020230030601

53.

Zhang

Katsiyannis

Roberts

(2014). Minority representation in special education: 5-year trends. Journal of Child and Family Studies, 23, 118–127. doi:10.1007/s10826-012-9698-6