Ethnic disproportionality among students with mental retardation and relationships between disproportionality and sociodemographic factors were investigated. Using national data, we examined the effects of gender, ethnicity, and sociodemographic factors on the proportion of students identified as having mental retardation. Results indicate a clear association among ethnicity, gender, and mental retardation. Sociodemographic variables were also strongly associated with the proportion of students identified. A logistic regression model that included sociodemographic predictors was significantly better than models with gender and race alone. Findings indicate that both individual student characteristics and district sociodemographic characteristics were important in determining the likelihood of identification of mental retardation and that the impact of the sociodemographic characteristics is different for each gender/ethnicity group. Recommendations for future research are provided.
Editor in charge: Steven J. Taylor
The disproportionate overrepresentation of minority children as having mental retardation is a longstanding issue in educational policy and practice. Overrepresentation of African American students, in particular, has led to significant litigation (Larry P. v. Riles, 1986; Marshall et al. v. Georgia, 1984; Pase v. Hannon, 1980; S-1 v. Turlington, 1986), broad legislative protections (Section 504 of the Rehabilitation Act of 1973, the Individuals With Disabilities Education Act, 1997 [IDEA]; the Americans With Disabilities Act, 1990), and a continuing watchdog role for the U.S. Office for Civil Rights, the agency established to monitor and oversee compliance with antidiscrimination laws.
Despite the many safeguards, for the second time in 20 years the U.S. Congress has mandated a comprehensive national study and examination of disproportionate representation of minority students in special education. National concern about the educational experiences of minority children was also prominent in the 1997 Amendments to IDEA (P.L. 105-17). The overidentification of African American students as having mental retardation was cited specifically, and states were given new requirements to collect data and respond appropriately to disproportionate representation.
Historically, estimates of disproportionate representation have varied widely, often reflecting definitions or calculations of disproportionality that are difficult to interpret (Coutinho & Oswald, 2000; Coutinho, Rice, & Oswald, 1999; Oswald, Coutinho, Best, & Singh, 1999; Utley & Obiakor, 2001). In many cases, those making estimates based them on regional samples, omitted or combined ethnic groups, or used confusing percentage figures to characterize differences. Recently, Oswald and Coutinho (2001) completed a comprehensive descriptive analysis of the national data base on ethnicity in special education that is collected biannually by the U.S. Office for Civil Rights. For the nation as a whole, Black and American Indian students were overidentified as having mild mental retardation throughout the 14-year period between 1980 and 1994 (as compared to White students). The extent of overrepresentation has declined from an odds ratio of about 3.2 to 2.3 for Black students and from about 1.5 to 1.3 for American Indian students. Asian Pacific Islander students were consistently underrepresented as having mild mental retardation throughout the time period, reflected by odds ratios of under .5.
The analyses pointed to the importance of disaggregating information by both disability condition and ethnicity. For example, although Black and American Indian children were also overrepresented among students with serious emotional disturbance through most of the time period, there was virtually no overrepresentation of Blacks and American Indians among students with learning disabilities, especially since 1986 (Oswald & Coutinho, 2001).
Policymakers and scholars are divided about how to respond to ethnic disproportionate representation among students with mental retardation. The contentious debate over what constitutes an effective response reflects basic assumptions about proportionality across ethnic groups (Kauffman, Hallahan, & Ford, 1998) and complex, sometimes not fully articulated, positions about what forces influence special education identification rates for minority children. Interpretations of disproportionate representation span a complex continuum. For some, disproportionate representation is evidence of discriminatory practices in the referral, assessment, and placement process and should be understood as a problem with historical and sociopolitical roots (Ford, Obiakor, & Patton, 1995; Harry, Rueda, & Kalyanpur, 1999; Larry P. v. Riles, 1986; Obiakor, 1999; Patton, 1998). Dunn (1968) was one of the first educators to decry the disproportionate labeling of children living in poverty as having mental retardation and providing services in segregated special classes. He pointed out the overrepresentation of minority students in these classes, called for an end to practices that he believed were obsolete, and raised serious educational and civil rights issues. More recently, Harry and Anderson (1995) emphasized the extent to which the special education process mitigates against positive outcomes for minority students, particularly African American students. These authors were critical of many aspects of the special education process of referral, assessment, and identification as well as the quality and appropriateness of instruction.
From a somewhat different perspective, the first Congressionally mandated study of disproportionate representation, conducted by the National Academy of Sciences in the 1980s, and other related works have been focused more on the educational opportunities available to minority students. From this perspective, the absence of adequate regular and special education programs may be the core problem rather than disproportionality per se (Heller, Holtzman, & Messick, 1982; Reschly, 1988a, 1988b, 1991). Despite strong disagreement about the causes or significance of disproportionate representation, there is agreement that:
Equal educational opportunities for students of diverse cultural, ethnic, linguistic, and socioeconomic backgrounds seem to be deferred dreams. The historic educational reality of these students is that their education mostly occurs in urban environments with little commitment to individuality, respect for differences, equal opportunity for all, freedom of discourse, and opportunities for upward mobility. (Utley & Obiakor, 2001, p. 3)
Increasingly, investigators have called for research that is conceptually guided and ethically informed (cf. Artiles, 1998). Conceptually based, empirical research may be expected to provide educators and the public with information that can improve the outcomes of minority students and ensure educational equity. Studies are needed in which researchers (a) consider alternative hypotheses regarding overrepresentation to improve our understanding of how school, fiscal, and community factors influence special education identification and (b) examine the policy and practice options that will improve the minority students' experiences.
Consistent with these purposes, Coutinho and Oswald (2000) have offered two alternative hypotheses summarizing the causes of disproportionality. The first hypothesis is that the special education referral, assessment, and eligibility processes measure and interpret the ability, achievement, and behavior of students differently across ethnic groups. An alternative hypothesis is that the underlying distribution of educational disability may vary across ethnic groups and that social and demographic factors, which are associated with ethnicity as well as with disability, may result in differential susceptibility to mental retardation.
There are many examples of social and demographic factors that may result in disproportionate representation, including poverty (U.S. Department of Education, 1998b; Utley & Obiakor, 2001), access to adequate health care or counseling (Baumeister, Kupstas, Klindworth, & Zanthos, 1991), and access to appropriate regular education programs (Messick, 1984b). Poverty, in particular, has been advanced as a critical factor affecting the ethnic composition of programs serving students with disabilities. Therefore, disentangling the effects of ethnicity and poverty is critical (Fujiura & Yamaki, 2000). Poor children are more likely than wealthy children to receive special education services, and ethnicity and poverty are correlated (U.S. Department of Education, 1998b). Using data from the National Longitudinal Transition Study, Wagner (1995) observed that the preponderance of poor students in special education was the primary contributor to disproportionate representation. Other investigators have also found that poverty may account for disproportionate representation of minority children with mental retardation (Baumeister et al., 1991; Gelb & Mizokawa, 1986; J. Gottlieb & Alter, 1994; Hodapp, Burack, & Zigler, 1990; Seifer & Sameroff, 1987; Yeargin-Allsopp, Drews, Decoufle, & Murphy, 1995). Yeargin-Allsopp et al. studied the extent to which sociodemographic variables accounted for the observed disproportionate representation of Black children with mild mental retardation in a sample of children living in a southern, metropolitan area. A Black to White odds ratio of 2.6 was reduced to 1.8 when investigators simultaneously controlled for race, gender, maternal age, birth order, maternal education, and economic status. Recently, the U.S. Department of Education (1998b) concluded: “Poverty rather than race/ethnicity, may account for some of the overrepresentation of minorities in special education programs. Therefore, without attention to poverty and its effects on children, the use of unbiased assessment alone will not eradicate the disproportionate representation” (Section II, p. 23).
Two other studies have been conducted recently to examine a number of demographic, economic, and educational variables that might impact rates of identification of minority students with mild mental retardation. Halfon and Newacheck (1999) provided a national profile of the prevalence of mild mental retardation and the impact of sociodemographic factors. The sample included children whose parents participated in the National Health Interview Survey (NHIS), an ongoing, nationwide household survey that relies on parental report of disabling conditions. The sample was constructed to be representative of the United States civilian noninstitutional population. Mild mental retardation was slightly more prevalent in Black children than in White children, but this difference disappeared when the effects of other sociodemographic variables were considered. Sociodemographic variables included age, gender, family income, family structure, family size, region, and educational level of the head of household. The adjusted odds ratio for identification as having mild mental retardation for African American and Hispanic children were 1.01 and .60 once the effects of social factors were accounted for.
In a study of African American students using U.S. Office for Civil Rights data, Coutinho, Oswald, and Best (1999) reported that 35.6% of the variation in identification rate could be accounted for by the variation in sociodemographic variables alone. Two race-related variables were then added to the model to investigate whether the racial/ethnic distribution of the school district and race of the individual students influenced the likelihood that a student would be identified as having mild mental retardation. That model yielded an improvement over the first model, which had included only sociodemographic variables, indicating that students' ethnicity was important in identification of mental retardation. Although this improvement was statistically significant, it was, in absolute terms, very small, and the final model that included the race-related variables accounted for only 36% of the variation in identification rates.
In studies of ethnic representation in special education, researchers generally fail to address the impact of gender on the probability of being identified as having mild mental retardation, but gender disproportionality in special education is well-known. More than two thirds of all students with disabilities are male; for youth with mental retardation in secondary schools, approximately 58% are male (U.S. Department of Education, 1998b; Wagner et al., 1991). The reasons for gender disproportionality, however, are not clear. Maturational and physiological differences (Halpern, 1992; Harmon, Stockton, & Contrucci, 1992; Hayden-McPeak, Gaskin, & Gaughan, 1993) and higher rates of disruptive behavior by boys (Del'Homme, Kasari, Forness, & Bagley, 1996; Sadker & Sadker, 1994) have been proposed as explanations, but most scholars regard these as insufficient to account for the extent of disproportionate representation (Phipps, 1982).
Gender bias in referral and assessment has also been suggested as a possible cause of the disparity (Gregory, 1977; U.S. Department of Education, 1998b). Teachers' opinions about a student's need for special services may be gender-biased because (a) society places higher emphasis on the achievement of boys (Richardson, Katz, & Koller, 1986); (b) the teaching force is predominantly female (U.S. Department of Education, 1998b); (c) assessment measures may overlook disability conditions that are more prevalent in girls, such as internalizing disorders (McIntyre, 1990; U.S. Department of Education, 1998b); or (d) society may hold lower expectations for girls, leading to the underreferral and underidentification of girls (B. Gottlieb, 1987). Appropriate responses to ethnic disproportionate representation among students identified as having mental retardation will also require consideration of gender disproportionality.
Sufficient and objective evidence of disproportionate minority representation is no longer debated. Unclear, however, are the appropriate policy and practice responses that will improve the minority student experience and provide equity. Appropriate responses to disproportionate representation require accurate estimates of proportionality (by ethnicity and gender) and conceptually based, empirical research related to the variation in identification rates for mild mental retardation among ethnic groups (Artiles, 1998; Coutinho & Oswald, 2000; Utley & Obiakor, 2001). As described previously, there is evidence to support the role of biased or inappropriate referral, assessment, and instruction of minority students as well as the influence of sociodemographic variables, such as poverty, per pupil expenditure. The available information makes clear that the process underlying disproportionality is complex, and focused investigations are needed to clarify the full range and influence of variables that affect susceptibility to educational disability and proportionality in identification rates by ethnicity and by gender.
Our purpose in the present study was to investigate (a) the extent of disproportionality among students with mental retardation in the most recent Office of Civil Rights survey and (b) the extent to which a set of sociodemographic variables was related to the disproportionate representation of minority students as having mental retardation. We hypothesized that mental retardation would be found to be more common, for example, in districts with higher poverty. If higher poverty districts also have a higher proportion of minority students, then some part of the disproportionate representation attributed to race or ethnicity may be explained by differences in poverty rates between communities. We also examined variability in mental retardation prevalence across communities having varying levels of minority student enrollment. We suggest that a significant relationship between mental retardation identification and minority enrollment, which varied across ethnic groups, would provide support for an hypothesis of systemic bias.
Every 2 years, the U.S. Department of Education Office for Civil Rights collects information on a nationally representative sample of school districts that is used to compile an Elementary and Secondary School Civil Rights Compliance Report, the chief source of data on the status of civil rights in the nation's schools. Approximately one third of the nation's school districts are included in the sample (U.S. Department of Education, 1998a). A stratified random sampling scheme is used so that state and national figures may be projected from the survey data. Those districts “forced” into the sample—usually for regulatory or monitoring reasons—were excluded from the analyses we report here. The final data set included school districts in the 50 states and the District of Columbia.
The Office of Civil Rights data set is marked by some important limitations. Disability definitions provided in the survey document are somewhat different from those included in IDEA; further, state definitions vary and the extent to which students in the various disability categories are similar across districts and states is unknown. In addition, the smaller districts included in the sample are different for each survey. Finally, the ethnicity categories employed in the survey have been criticized sharply, as has the very notion of reliably classifying people by constructs as elusive as race or ethnicity (Wright, 1994). Nonetheless, the Office of Civil Rights survey represents the only available dataset for investigating the ethnicity distribution of children with disabilities in the nation's schools system. As a result of a congressional mandate in the 1997 IDEA amendments, complete ethnicity data for all children with disabilities will be provided in the 1999 Office of Special Education Programs Child Count; however, these data have not yet been released to the public.
For this report, we considered only the information on enrollment and disability categories from the school year 1994–1995. District enrollment was broken down by ethnicity (five categories) and by gender (two categories). The number of students in three disability categories (mental retardation, serious emotional disturbance, and learning disabilities) was also broken down by ethnicity and by gender. The fourth disability category in this study, “None,” included students with lower incidence disability conditions as well as all regular education students. Because mild mental retardation data were not disaggregated by gender, the analyses reported here included students with mild, moderate, and severe mental retardation.
District weights were used in all analyses. These weights reflect the fact that, although all of the nation's large school districts are included in the data set, only a sample of the smaller districts were surveyed. The district weights are proportional to the probability of selection for districts in each stratum of the sample. This sampling and weighting scheme allows for projections to state and national levels (U.S. Department of Education, 1998a). The weighted number of students in each disability category provided in Table 1 estimates the figures for the nation as a whole.
The original data file had 44,276 observations (schools) and 431 variables. In order to summarize information at the district level, we collapsed the data by accumulating the data for all of the schools in a district. This accumulation to the district level was necessary because the predictor variables used in this study were only available for each district. Each district is uniquely identified by a code number to allow matching of information across data sets. Included in the sample were 4,151 school districts serving over 24 million students.
The National Center for Educational Statistics, Common Core of Data CD-ROM (NCES CCD93 Disc) has information on 15,041 school districts in the 50 states and DC. The information in this data set was matched with the Office of Civil Rights data so that only those districts surveyed in the 1994 Office of Civil Rights survey were included in this study. In the process of merging the two data sets, we discovered about 30 districts for which the ID number was inconsistent. Through a process of comparing the districts' name and city and state, we were able to correct all but 2 of these inconsistencies; these mismatches were dropped from the sample.
Nine sociodemographic variables were chosen as predictor variables for this study (see Table 2). Additional variables were considered but excluded due to a substantial number of missing values (e.g., districts' dropout rates). In a few cases, information was not available for the most recent school year, and data from a previous year were used to fill out the data set. As Table 2 shows, the variable with the most missing values, student–teacher ratio, was available for 95.2% of the districts and 94.5% of all students in the sample.
The predictor variables included in the analyses were clearly related; Spearman rank order correlation coefficients for predictor variables are provided in Table 3. POVERTY and INCOME, for example, showed a predictable strong, inverse relationship.
We examined the relationship among gender, ethnicity, and sociodemographic factors and the proportion of students in a school district identified as having mental retardation. The models use the proportion of students in the mental retardation disability category as the dependent variable and two sets of variables as predictors. The predictor variables were the district-level sociodemographic continuous variables and the child-level categorical variables of gender and ethnicity. The predictor variables are combined into a model that includes main effects and interaction. Categorical responses (such as having mental retardation: yes/no) may be modeled using logistic regression (Hosmer & Lemeshow, 1989), and such models include the proportion in each disability category as the response variable. The logistic regression procedure reported later models the probability of a child being classified as having mental retardation as a function of gender, ethnicity, and sociodemographic predictors.
The model used in this study actually examined the probability of a child being in one of the four disability conditions; however, in the interest of clarity, here we report only the mental retardation findings. All analyses were done with the districts weighted by the number of students (as well as by sample weight) so that the models simulate using the student as the unit of analysis rather than the district.
The predictor variables are of two types. The categorical predictor variables were gender (F, M) and ethnicity (American Indian [AI], Asian Pacific Islander [AS], Black [BL], Hispanic [HI], and White [WH]). To maintain consistency, we have retained the ethnicity and disability category labels used in the Office of Civil Rights survey. The continuous, sociodemographic predictors are listed in Table 2. The continuous predictor, or covariate, effects in the model included a linear and quadratic trend for each covariate, all possible two-way interactions for the linear and quadratic trends for each covariate, and gender, ethnicity, and Gender × Ethnicity interaction effects crossed with the linear and quadratic trends for each covariate. The net effect of this model is the possibility of a separate linear and quadratic trend for each gender and ethnicity combination. All continuous covariates were centered and scaled to avoid problems with ill conditioning and collinearity in analyses (Draper & Smith, 1998). Because of the large sample size and the complexity of the model, we decided to use .0005 as the cut-off for significance.
For the sample as a whole, 1.38% of all students were identified as having some form of mental retardation. This was slightly higher than the 1996 Annual Report to Congress, which placed the mental retardation identification rate at 1.14% of the population of enrolled students for the 1994–1995 school year. Table 4 illustrates, however, that the mental retardation identification rate varied from a low of .44% of Asian females to 3.15% of Black males. These identification rates clearly demonstrate disproportionate identification across the gender/ethnicity groups.
A simple chi-square analysis of the data in Table 1 shows a clear association between ethnicity and gender and the disabling conditions, χ2 (27, N = 41,819,191) = 628,912, p < .0001, verifying that the mental retardation identification rates are not the same for all of the 10 gender/ethnicity groups. To clarify this finding, we constructed odds ratios for each of the gender/ethnicity groups, with White females as the comparison group. These odds ratios provide an estimate of the likelihood of being identified as having mental retardation compared to the likelihood for White female students. Thus, the odds ratio for White female students is, by definition, 1.0. In this sample, White males were 1.36 times as likely as White females to be identified as having mental retardation, whereas Black females were 2.02 times as likely (see Table 4). Black males displayed the largest disproportionality, with an odds ratio of 3.26. These data starkly represent the extent of the problem of disproportionality in mental retardation identification across gender and ethnic groups.
A logistic regression analysis with mental retardation identification as the response variable and only the nine sociodemographic variables as predictors (including linear, quadratic, and interaction effects) was significant, χ2(162, N = 41,726,796) = 345,130, p < .0001. Thus, the sociodemographic conditions of a school district are strongly associated with the proportion of students identified as having mental retardation; some statistically significant portion of the variation in districts' mental retardation identification rates can be explained by this combination of predictor variables.
The bivariate correlation coefficients between identification rate for each gender/ethnicity group and the predictors are provided in Table 5. Spearman rank correlations are used for this purpose because of the skewed distributions of the variables and the presence of marked outliers. These correlation coefficients should be interpreted with caution because the relationships between identification rates and predictors are not linear, as demonstrated by the significant quadratic and interaction effects in the logistic model. Further, because these are bivariate correlations, the relationships do not take into account the effects of other predictors in the model.
Overall, mental retardation identification rate is positively associated with student–teacher ratio (STR), POVERTY, ATRISK, NONWHITE, and NO DIPLOMA, and negatively associated with per pupil expenditure (PPE), HOUSING, INCOME, and limited English proficiency (LEP). The direction of the relationships holds for most of the individual gender/ethnicity groups for all predictor variables. An examination of the correlation coefficients offers a first look at the evidence for the alternative hypotheses regarding disproportionality. POVERTY, for example, is weakly-to-moderately, positively associated with identification rates for each of the gender/ethnicity groups except female Asian/Pacific Islander students (where the relationship is essentially nonexistent). This would appear to provide some support for the differential susceptibility hypothesis: Increased poverty is associated with increased mental retardation identification in most of the gender/ethnicity groups. As will be seen shortly, however, this simple bivariate relationship is somewhat misleading; the nature of the relationships changes, in some cases substantially, in the full model where the predictor effects and their interactions are all considered simultaneously.
Given that both student characteristics (gender and ethnicity) and sociodemographic variables were each individually associated with the identification of students with mental retardation, we next sought to determine whether both were important in predicting mental retardation identification and whether the relationships between the predictor variables and mental retardation identification rates were the same for each gender/ethnicity group. A logistic regression analysis including the nine sociodemographic variables (and their linear, quadratic, and interaction effects), gender, race, and all possible interactions of the covariates with gender and race was found to be significantly better than the model with only the sociodemographic predictors, χ2(1485, N = 41,726,796) = 667,570, p < .0001, and significantly better than the model with only the gender and ethnicity groups, χ2(1620, N = 41,819,191) = 383,788, p < .0001. There was also a significant gender/ethnicity by sociodemographic interaction, χ2(1458, N = 41,726,796) = 86,224, p < .0001. These findings indicate that both individual student characteristics and district sociodemographic characteristics are important in determining the likelihood of mental retardation identification and that the impact of the sociodemographic characteristics is different for each of the various gender/ethnicity combinations.
To illustrate this finding, we plotted the predicted mental retardation identification rate for each gender/ethnicity group across the range of POVERTY represented in the sample. For the purpose of this illustration, the effects of all other sociodemographic variables are held constant (i.e., are set to the median value). Figure 1 shows that for African American males, American Indian males and females, and, to a lesser extent, African American females, mental retardation identification drops markedly as POVERTY increases. However, for White males and females, the rate of mental retardation identification increases slightly, beginning at about .8% and .5%, respectively, then leveling off at about 1% and .8%. Such a differential effect significantly influences disproportionality. Living in a community with a low rate of poverty markedly increases the odds of mental retardation identification for Blacks and Native Americans, both in absolute terms and in relation to the rates for their White counterparts. Conversely, in high poverty districts, the mental retardation identification rates are much closer for all ethnic/gender groups.
Similarly, when plotted across the full range of NONWHITE, Figure 2 shows that for African American males and females, mental retardation identification drops markedly as the percentage of nonwhite students increases. However, for all other students, the decline is less notable. Living in a largely White community markedly increases the odds of mental retardation identification for Blacks, but only slightly for other students.
To clarify how the odds ratios for the gender/ethnicity groups change when sociodemographic influences are taken into account, we calculated adjusted odds ratios, defined as the odds ratio for each gender/ethnicity group at the median value of each of the predictors. This odds ratio is actually a predicted value based on all of the effects included in the logistic regression model. As shown in Table 4, the odds ratios do change, but not necessarily in the expected direction. The adjusted odds ratio for African American male students, for example, actually increased to 4.03 (as compared to 3.26 for the raw values for the sample) at the median value of the sociodemographic variables.
Finally, we examined how the odds ratios change across the distribution of each of the predictor variables. We computed the 10th percentile and the 90th percentile for each predictor and computed the odds ratio for each of the gender/ethnicity groups at those points in the distribution, holding all of the other predictors at the median value. The results are shown in Table 6.
As an example, as HOUSING goes from the 10th percentile ($43,733) to the 90th percentile ($192,027), predicted disproportionality for Black male students increases from an odds ratio of 3.45 to an odds ratio of 6.88. Correspondingly, as POVERTY goes from the 10th percentile (5.5%) to the 90th percentile (33.6%), predicted disproportionality for Black male students decreases from an odds ratio of 6.33 to an odds ratio of 3.42.
The data in Table 6 also illustrate how the predictors differentially affect the gender/ethnicity groups. The disproportionality of mental retardation identification for American Indian male and female students, for example, declines as NONWHITE goes from the 10th to the 90th percentile, while it increases for Black and Hispanic male and female students.
The data support the position that both individual student characteristics, such as gender and ethnicity, and communities' sociodemographic characteristics influence the likelihood of being identified as a student with mental retardation. These findings offer some clues to clarify the validity of competing hypotheses regarding the causes of disproportionality.
Some of the predictor effects would seem to support the hypothesis that disproportionality is, in part, a result of differential susceptibility. The influence of poverty, for example, is validated in the bivariate analyses in that poverty is generally positively associated with mental retardation identification rates, both overall and within most of the gender/ethnicity groups. The association between minority status and poverty would lead to the conclusion that some disproportionality may be attributed to the differential effects of poverty. However, when all of the predictors are included in the model, the effect of poverty is reversed for most students of color, and overidentification for these students increases as poverty declines. These findings suggest that in low poverty districts, the increased rate of identification among students of color may be attributable to systemic bias.
The effects of other sociodemographic factors appear to offer more explicit support for a hypothesis of systemic bias. As noted earlier, Black students who live in a largely White community are more likely to be identified as having mental retardation than those who live in a community with a greater minority population. It is difficult to construct a scenario by which the ethnicity distribution of the community differentially affects students' susceptibility to disability. Indeed, in this example, the effect would appear to support the bias hypothesis that students who “stand out” by virtue of being a member of a small ethnic minority may be more likely to be the victims of discrimination.
Districts that spend more on education (higher per pupil expenditure) tend to have relatively less disproportionality in mental retardation services for African American students but, curiously, have substantially more disproportionality of Hispanic students. This finding may be interpreted as support for either hypothesis: (a) perhaps Hispanic students do not share equally in the benefits of increased spending on education so that their educational and cognitive development is not enhanced and a disproportionate number meet mental retardation criteria; alternately (b), perhaps districts with higher per pupil expenditure recruit relatively fewer Hispanic teachers and psychologists, and, as a result, cultural characteristics are misinterpreted as indications of mental retardation.
Recommendations for Further Research
The findings reported in this study illustrate the need for additional research to better understand disproportionate minority representation in special education. The diagnosis of mental retardation is a significant event with lifelong consequences. Children with mental retardation made up about 11% of all students served as disabled under the Individuals With Disabilities Education Act (IDEA-P.L. 105–17, as amended) in the 1998–1999 school year (U.S. Department of Education, 2000). The causes of mental retardation are complex, and the disability is sometimes difficult to distinguish from other problems (e.g., achievement deficits caused by poor instruction). Three to 5 years after leaving secondary school, only 7.7% of all young adults with mental retardation were productively engaged outside of the home, living independently, and socially active (Wagner, D'Amico, Marder, Newman, & Blackorby, 1992). Additional studies are needed to determine more clearly how both individual student characteristics (such as gender, ethnicity, IQ, and adaptive behavior) and community sociodemographic factors influence the likelihood that a minority student will be identified as having mental retardation. Effective responses to disproportionate representation will rely on a clearer understanding of how minority students may be differentially susceptible to the disability of mental retardation and the extent to which systemic bias influences identification and services. Specific recommendations for research are as follows:
Examination of Age Differences in Disproportionality
Age has been identified as an important factor in disproportionality rates (Yeargin-Allsop et al., 1995). A careful scrutiny of the impact of age on mental retardation identification rates should offer additional clues with which to better interpret disproportionality. A detailed study of the incidence of mental retardation, and other disability conditions, across gender/ethnic groups in the preschool as compared to school-age years can provide information about how disproportionality arises.
Ethnicity-Specific Studies of the Impact of Sociodemographic Factors on Identification Rates
The adjusted odds ratios obtained in this study (which indicate how the probability of being identified as having mental retardation changes when the sociodemographic influences are taken into account) illustrate how the factors affect gender/ethnic groups differentially, but also indicate that, in many instances, the changes are not in the expected direction and are difficult to interpret. Additional research is needed to understand the basis for why the adjusted odds ratios often go up. For example, it is not clear why Black and Hispanic students experience increased disproportionality at the median value of the sociodemographic variables. In addition, for some of the ethnic gender groups, changes in the odds ratios across the distribution of the predictor variables are difficult to understand. One might ask why disproportionality among Hispanic students decreases as the percentage of dropouts in the community increases, but for Black and American Indian students the effect is reversed.
Investigation of the Identification Process
Large scale studies of disproportionality and the relationship between this problem and school district characteristics provide an important foundation and impetus for more detailed investigation of the process of special education identification. Capitalizing on the present work will require additional investigation of that process to better understand how these macro-level variables impact the identification of children with disabilities. Such studies can offer a much more detailed picture of the influences on identification in individual districts and for individual students and can begin to tease out the importance of those influences for children of color.
Finally, underidentification among White students rather than disproportionate overidentification may be the underlying and most urgent problem to address to improve equity and educational outcomes for all students, across ethnic groups. Mental retardation identification rates are just under 1% of the estimated resident population for the 1996–1997 school year, and the actual number of students with mental retardation decreased by .8% between 1987–1988 and 1996–1997, a time period when the resident population increased by 7.1%. During the same period, the number of students with learning disabilities increased almost 38% (U.S. Department of Education, 1998b). The significant reduction in mental retardation identification rates over the last several years is not necessarily good news. Many educators have expressed concern that children formerly categorized as having mild mental retardation are labeled instead as students with learning disabilities or remain in general education as a largely unserved population (Coutinho & Oswald, 2000; J. Gottlieb, Alter, Gottlieb, & Wishner, 1994; MacMillan & Balow, 1991; MacMillan, Hendrick, & Watkins, 1988).
Implications for Practice
The findings presented in this study emphasize the importance of examining the data before attempting to intervene in a problem as complex as minority disproportionality in special education. The impact of sociodemographic factors may be counter-intuitive and will require thoughtful analysis and additional exploration. In important respects, the findings obtained in this study appear to support the hypothesis of systemic bias. For example, the high absolute levels of mental retardation identification among Black students (especially Black males) in low poverty communities suggests that many are identified inappropriately as having mental retardation. On the other hand, it is possible that in low poverty communities, some White students with mental retardation are assigned another disability label, thus artificially depressing the White rate. In high poverty communities, the system may not work at all, such that many students with mental retardation simply go unidentified or are identified under a different disability condition.
With respect to the marked overidentification of Black students as having mental retardation in communities with low nonWhite enrollment, again the hypothesis of systemic bias is supported. This is particularly true for male Black students. A direct implication for practice related to these two findings is that educators must carefully scrutinize mental retardation identification rates in low and high poverty and in low nonWhite enrollment communities. The findings obtained in this study stress the importance of actions to redouble efforts to improve the referral and assessment process in special education (Harry & Anderson, 1995).
Longer term, the findings reinforce the importance of improved teacher education programs, ones that provide sustained, supported opportunities for preservice educators to become competent in providing effective teacher–student interactions, regardless of the ethnic background of the teacher or the student and that successfully remove all harmful incongruities between the home and the school learning environment (Townsend, 2000). Sustained, well-structured, and well-monitored field experiences with children from diverse cultures may be regarded as the most promising factor in helping preservice educators develop the cultural sensitivity and intercultural competence essential for appropriate special education identification and educational experiences (Grant, 1994; Noordhoff & Kleinfeld, 1993). Direct engagement with students of diverse backgrounds allows prospective teachers to learn experientially from the students their culture, language, and behavior as well as their communities' economic, social, and political atmosphere. Having obtained this knowledge, student teachers can come to understand cultural context, consider educational goals, and implement instructional strategies that fit perceived context and meet public expectations for improving the minority student experience. In other words, the field experience provides cultural knowledge as well as experiential knowledge—practical information about multicultural classrooms—that cannot be obtained through formal course work or typical workshops (Cooper, Beare, & Thorman, 1990; Melnick & Zeichner, 1997; Powell, Zehm, & Garcia, 1996).
Based on the findings obtained here, there do not appear to be any immediate implications for practice as relates to per pupil expenditure. Disproportionality appeared to increase as expenditures increase for Black females and Hispanic students. However, there is not sufficient or clear enough evidence to recommend expenditures be increased or decreased as a response to disproportionality.
In summary, long-term policy and practice responses to disproportionate representation for each ethnic/gender group depend upon conceptually guided, empirical research at the community level to learn more specifically how individual characteristics and sociodemographic factors influence who is identified as having mental retardation. At this time, current practices at the community level must also be examined and altered, as appropriate, to address the likelihood of systemic bias in the special education referral and identification process. Empirically validated improvements in preservice teacher education programs are recommended to support instructional environments in regular and special education that are characterized by culturally competent teacher–student interactions and the identification of students who have disabilities, rather than differences due to gender or ethnicity.
NOTE: Preparation of this manuscript was supported in part by the Field-Initiated Studies Program of the National Institute on Educational Governance, Finance, Policymaking, and Management, Office of Educational Research and Improvement, U.S. Department of Education (Grant No. R308FG70020).
Authors:Donald P. Oswald, PhD, Associate Professor ( DOSWALD@VCU.EDU); Al. M. Best, PhD, Associate Professor; and Nu Nguyen, BA, Graduate Assistant, Department of Psychiatry, Medical College of Virginia, Virginia Commonwealth University, Box 980489, Richmond, VA 23298–0489. Martha J. Coutinho, PhD, Professor, Department of Human Development and Learning, East Tennessee State University, Box 70548, Johnson City, TN 37614.