Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 322
Pallet n Special Eclucation Placement as Revealed by the OCR Surveys JEREMY D. FINN Since 1968 the Office for Civil Rights (OCR) has surveyed schools and school districts regarding student enrollment and placements. This paper describes the results of an analysis of the OCR survey data pertaining to the 1978-1979 school year. The original data consist of simple counts of students obtained from school and district offices at one point in time: Oc- tober 1978. The data do not describe the processes whereby one student (or a group of students) is placed in special programs in any particular set- ting and, therefore, cannot explain how differences in placement rates are created. The data do, however, document the extent of disproportion in special programs by race/ethnicity and gender as well as the demographic conditions under which smaller or larger disproportions are found. It is clear that the placement rates in special education programs are very different both for minority and white students and for males and females. Table 1 gives nationwide percentages of students in each of five special programs. Minorities are classified as educably mentally retarded (EMR) at a rate that is substantially higher than that for white students both in absolute and relative terms. By comparison, the male-female ratio in EMR programs is smaller, but 3 times as many males as females are classified as emotionally disturbed, and almost 2.5 times as many males as females have specific learning disabilities. I am grateful to Robert Serfling, Amado Padilla, Reginald Jones, Richard Eyman' Lyle Jones, Ingram Olkin, and Miron Straf for their reactions and suggestions for improvements to an earlier draft of this paper. 322
OCR for page 323
Patterns in Placement as Revealed by the OCR Surveys 323 The purpose of this analysis is to illuminate the differences in place- ment rates and, to the extent possible from the survey data, to describe the context in which they arise. This paper summarizes the results of the data analysis in a progression from general to more specific findings. Dif- ferences between minority and nonminor~ty students in EMR placements are described, and the examination is specified by separate racial or eth- nic classifications and by special education programs other than EMR programs. THE 1978 OC R S U RVEY In its 1978 Elementary and Secondary School Civil Rights Survey, OCR sampled approximately 6,000 school districts, or about one third of the districts in the United States. Questionnaires were sent to all district of- fices and to every school in the 6,000 districts, requiring counts of the number of students enrolled, the number enrolled in special education programs, and additional global characteristics of the student population. All student counts were classified by racial or ethnic identity, and some were also classified by gender. Both racial/ethnic and gender classifica- tions were required for students in five special programs, which are, according to the general instructions (Form OS/CR 102), as follows: 1. Educable mentally retarded (or handicapped) a condition of mental retarda- tion which includes pupils who are educable in the academic, social, and occupa- tional areas even though moderate supervision may be necessary. 2. Trainable mentally retarded (or handicapped) a condition of mental retar- dation which includes pupils who are capable of only very limited meaningful achievement in the traditional basic academic skills but who are capable of profit- ing from programs of training in self-care and simple job or vocational skills. According to the "general instructions to the fall 1978 school survey" (Form OS/CR 102), the following racial or ethnic categories are recognized: American Indian or Alaskan Native: A person having origins in any of the original peoples of North America and who maintains cultural identification through tribal affiliation or community recognition. Asian or Pacific Islander: A person having origins in any of the original peoples of the Far East, Southeast Asia, the Pacific Islands, or the Indian subcontinent. This area includes, for example, China, India, Japan, Korea, the Philippine Islands, and Samoa. Hispanic: A person of Mexican, Puerto Rican, Cuban, Central or South American, or other Spanish culture or origin-regardless of race. Black, not of Hispanic origin: A person having origins in any of the black racial groups of Africa. White, not of Hispanic origin: A person having origins in any of the original peoples of Europe, North Africa, or the Middle East.
OCR for page 324
324 .~ C) ._ so o Ct :^ Ct x :^ C~ C) Ct o ._ Ct :: - Ct ._ C) U) ._ o ._ Ct z x C~ ; - ._ ._ 50 o C) CJ. Ct 00 ~ C~ r~ _ _ . . . . . - C~ 5 L~ O C) . - O ~ ~ _ o~ Ct s~ CD ~0 C~ ~> C) C~ ~: ¢) - Ct - Ct C~ ~ O 00 ._ ~S: ._ s~ o ._ o ._ C~ C~ ._ ._ C~ Ct - ~0 ~ ~q oO r~ _ ~ ~ . . . . . O O - O O c~ 0 ~ ~ ~ a~ _ ~ _ r~ . . . . . . - O O - - ~ a~ oo C~ O O c~l ~r ~l ~ ° . . . . . . _ O O a~ r~ oO - - O . . . . . ~ U~ ~ _ ~ oo ~ ~ O O . . . . . O O O O O 1 ~ c~ a~ 0 c~ ~ O - ~ ~ O - . . . . . . _ o o C~ a~ C~ ~ ~ O oo . . . . . . ~ O O ~q - ~ - _ ~ a ~ E~ ~ 50 <~S · _ Ct ~ Ct 5: ~ O ~ £ £~ ~ _ :~. 5 ~ ~ ~ .- O ~ ~ V) - .O :_ 5 C~ - V, _ t4 ~ o ·Ce: C~ Ct S C) ._ ·Q 5: 0 ) C.) C) ~ ~ =, O ~? ~ Z - ~' a - - .0 _ Ct . _ ~ O Ct Ct ._ _ ~ o D s~ ~ C) O o o ~ ~_ C~ s: ~ O Ct ~ s~ O _ ~ O 0 ,:: C) C,) CL ~ _ O C., 0~ C~ ._ .. ~ O ~ C~ ~
OCR for page 325
Patterns in Placement as Revealed by the OCR Surveys 325 3. Seriously emotionally disturbed a condition exhibiting one or more of the following characteristics over a long period of time and to a marked degree, which adversely affects educational performance: an inability to learn which cannot be explained by intellectual, sensory, or health factors; an inability to build or main- tain satisfactory interpersonal relationships with peers and teachers; inappropriate types of behavior or feelings under normal circumstances; a general pervasive mood of unhappiness or depression; or a tendency to develop physical symptoms or fears associated with personal or school problems. The term includes children who are schizophrenic or autistic. The term does not include children who are socially maladjusted, unless it is determined that they are seriously emotionally disturbed. 4. Specific learning disability a disorder in one or more of the basic psycho- logical processes involved in understanding or in using language, spoken or writ- ten, which may manifest itself in an imperfect ability to listen, think, speak, read, write, spell, or to do mathematical calculations. The term includes such conditions as perceptual handicaps, brain injury, minimal brain dysfunction, dyslexia, and developmental aphasia. The term does not include children who have learning problems which are primarily the result of visual, hearing, or motor handicaps, of mental retardation, or of environmental, cultural, or economic disadvantage. 5. Speech-impaired a communication disorder, such as stuttering, impaired articulation, a language impairment, or a voice impairment, which adversely af- fects a child's educational performance. The sample consists of both "forced" districts, which OCR required to be included because of their compliance status or because they had ap- plied for funds under the Emergency School Aid Act, and "drawn" dis- tricts, chosen at random from within a sampling frame organized by 13 demographic characteristics to ensure that all characteristics of impor- tance to OCR were represented.2 Of the total 6,079 districts sampled, 6,040 provided responses; of these, 2,146 districts were "forced." The total number of schools represented in the sample is 54,082. Because the data are not a simple random sample of the districts of a state or region, sampling weights are provided to allow estimates of state totals or aver- ages. Checks on the accuracy of these projections were made from the 1976 school survey (U.S. Department of Health, Education, and Welfare, 1978a), which followed a similar sampling plan and yielded very reason- able results. The number of districts actually responding to the survey is given for each state in Table 2.3 The District of Columbia and Hawaii each have a 2Details of the sampling design for 1978 are given in U.S. Department of Health, Education, and Welfare (1978b). 3The sampling plan caused the District of Columbia and eight states to be surveyed ex- haustively (Alabama, Florida, Georgia, Hawaii, Louisiana, Mississippi, North Carolina. and South Carolina).
OCR for page 326
326 TABLE 2 Districts Sampled and Responding to Question on EMR Programs FINN Approximate Number Percentage of DistrictsNumber with EMR Statein State*Sampled Programs Alabama125125 100.0 Alaska5022 100.0 Arizona22381 96.3 Arkansas348237 94.9 California1,022326 66.6 Colorado11758 91.4 Connecticut16274 90.5 Delaware2412 100.0 District of Columbia11 100.0 Florida6767 100.0 Georgia183187 99.5 Hawaii11 100.0 Idaho11142 95.2 Illinois958320 68.8 Indiana280123 91.9 Iowa389138 60.9 Kansas301105 79.0 Kentucky159108 95.4 Louisiana6666 100.0 Maine17272 93.1 Maryland2521 100.0 Massachusetts336126 47.6 Michigan565202 88.1 Minnesota405142 81.0 Mississippi150150 98.7 Missouri419197 95.4 Montana52162 82.3 Nebraska99666 92.4 Nevada179 100.0 New Hampshire14243 93.0 New Jersey556197 75.1 New Mexico8851 94.1 New York716263 62.0 North Carolina144145 100.0 North Dakota28753 62.3 Ohio566245 92.2 Oklahoma596193 92.7 Oregon32764 87.5 Pennsylvania479317 83.3 Rhode Island3927 88.9 South Carolina929.2 100.0 South Dakota17458 63.8
OCR for page 327
Patterns in Placement as Revealed by the OCR Surveys TABLE 2 (continued ~ 327 Approximate Number Percentage of Districts Number with EMR Statein State* Sampled Programs Tennessee140 110 99.1 Texas1,077 573 90.8 Utah39 19 100.0 Vermont232 58 56.9 Virginia132 101 99.0 Washington301 91 87.9 West Virginia49 28 100.0 Wisconsin406 143 85.3 Wyoming48 29 89.7 *From 1976 OCR survey, which surveyed districts exhaustively. single administrative school district. Elsewhere the number of districts in a state varies immensely, as do the ways in which districts are defined. A number of states that are predominantly rural have many small districts- e.g., Nebraska, which has a large number of one-school districts a situa- tion that creates unique problems both for the organization of special education programs and for studying enrollment patterns. These districts, which often have small proportions of minorities, cannot be readily com- pared with those with much larger enrollments. To date, OCR has not conducted any checks on the accuracy of the school or district reports. The 1976 survey requested data from all school districts in the country, and the response rate was at least 95 percent in every state. School districts are obligated under Title VI of the Civil Rights Act of 1964 to respond to the survey in a timely and accurate fashion and are reminded of this in the survey instruments. Thus, while the data have not been and should be verified, the conditions under which they are ob- tained suggest that respondents would take reasonable care with their reports. TECHNICAL ISSUES MEASURING GROUP DIFFERENCES IN PLACEMENTS The results by sex in Table 1 show that disproportions may appear larger or smaller depending on whether they are based on the differences of per- centages or on ratios. Because the percentage scale is bounded by zero at
OCR for page 328
328 FINN one end and 100 at the other, absolute differences between values close to either end are generally limited to being relatively small. In other words, a program for the seriously emotionally disturbed (SED), which has a small proportion of pupils enrolled in total, does not have a large absolute disproportion by gender, even though the process of classifying students as emotionally disturbed results in a male-female ratio of about 3:1. At the same time, there is a greater percentage of females who are not in spe- cial education. In comparison to the nonclassified group, the 3:1 dispro- portion is still more extreme. In the analysis presented in this paper, these problems were addressed by using an index of disproportion derived from recent statistical develop- ments termed log-linear analysis (Bishop et al., 19751. The basic element in the index is the "odds" of being assigned to a particular special educa- tion category. For example, a measurement of the odds of a minority stu- dent's being assigned to an EMR class is the percentage of minority stu- dents who are classified as EMR divided by the percentage of minorities who are not in special programs. From Table 1, this is 2.54/92.60 = 0.027. The odds of a white student's being designated EMR is 1.06/94.12 = 0.011. The disproportion index is the ratio of these two odds, scaled by being transformed to a natural logarithm;4 that is, loge(O.027/0.011) = 0.89. The log-odds index is positive because the EMR odds for minorities is larger than those for whites; it would be zero if the odds for minorities and whites were equal and negative if the odds for minorities were lower than those for whites. The index is not simple to interpret since the measure is unbounded, i.e., it can vary from-oo to +oo depending on the magni- tude of the disproportion. As a rough interpretive device, however, the log-odds index can be transformed to a measure of association, Yule's Q-statistic, which, like a correlation, is limited to values between-1 and + 1.s Thus the association of race or ethnicity (minority versus nonminor- ity) with placement (EMR versus none) is +.42. To see the degree of change in either the log-odds index or Q with a change in disproportion, suppose that the minority-white EMR ratio was 5:1 instead of the actual ratio of approximately 2.5:1.0. That is, suppose that 5.30 percent of racial/ethnic minorities were enrolled in EMR programs 4This is also equivalent to the difference of the logarithms of the two odds, i.e., In (0.027)- In (0.011) = 0.89. sThe relationship is given by ~ = (a-l )/(a + 1), where a = ex and x is the log-odds index. This transformation is the inverse of Fisher's z for correlations and maps x onto the zero-one interval. Q is normally distributed in large samples and attains a value of unity whenever either odds is zero.
OCR for page 329
Patterns in Placement as Revealed by the OCR Surveys 329 about double the Table 1 value and that 89.84 percent of minorities were not enrolled in any special program instead of the actual value of 92.6 percent. These hypothetical values would increase the log-odds index to 1.66 and the measure of association to Q = .68. DISAGGREGATION OF DATA A second technical issue is the extent to which data on disproportion should be disaggregated. For example, Table 3 presents the percentage of students in each special program for specific racial/ethnic populations. It is clear that the relatively large minority-white differences in EMR place- ments are even more extreme for black students alone (3.46 percent of black students are classified EMR), who also comprise the largest minor- ity population in this country. The disproportions in programs for the trainable mentally retarded and for emotionally disturbed children are also due in large part to the disproportionate representation of blacks in these classifications. At the same time, for Hispanic students the second largest minority group placement rates in EMR, TMR, and SED pro- grams are very close to those for non-Hispanic whites on a nationwide basis. Asian and Pacific Island students have the lowest placement rates of all groups in the same three programs. Table 3 also provides information on the apparent lack of difference be- tween minority and white placements in specific learning disabilities pro- grams. A slightly larger percentage of whites is classified as having spe- cific learning disabilities than blacks, unlike the difference in other special programs, while a still larger percentage of Hispanic students is classified as having specific learning disabilities. Disaggregation by race or ethnicity provides information that is not ap- parent in Table 1.6 To simplify the data presentation, this paper first presents results for all minorities combined; the results are then subdivided for separate racial/ethnic groups. It is an important characteristic of the log-odds index of disproportion that it can be validly computed for each minority group separately, by replacing the odds of placement for all mi- norities with the odds for a particular subpopulation (e.g., blacks or His- panics). Other approaches e.g., the comparison of the proportion of 6 Some further disaggregation by grade is possible with the OCR data, by locating schools within each district that serve only grades kindergarten-6, 7-8, and 9-12, respectively. About three fourths of the schools in the sample can be classified in this manner. While some age-related analysis is possible, the various levels are not comparable because of different dropout rates; dropout information was not gathered in the OCR's 1978 survey.
OCR for page 330
330 = _ = O ~ ~ ~ ~ ~O for ° == .- ret O ~ ~ ~ O ~ ~_ o o O ~ CN ~ ~ O ~ ~Cq so ct ret ~ ~ up ~ oO .O ~ ~ret 0 0 ~ _ Is ~1 . ~in ~.V' lid ~ ~ ~ ~ 00 ~ ~ Cat ~ ~O O O Hi _ ~ X es . e,: ° ~ E V) so ~ ~ ~ 2 ~ . ~ ~ ., o _ o o ~ _ _ ~ o ~ ~ it,, .¢ !,5
OCR for page 331
Patterns in Placement as Revealed by the OCR Surveys 331 EMR students who are black with the proportion of the total school popu- lation that is black do not give an accurate portrayal of disproportion in settings with multiple minority groups. This is because the denominators of the EMR proportion and of the total proportion are inflated differen- tially by the number of minority students included who are not black. Further disaggregation by geographic or administrative unit can reveal trends that would not be apparent if the number of children enrolled in each school or district was disregarded. For example, consider a hypo- thetical geographic unit (a state or nation) that has only two school dis- tricts. District 1 has a total enrollment of 3,000 students, consisting of 1,000 white and 2,000 minority students. Of these, 20 white (2 percent) and 20 minority students (1 percent) are classified as EMR. While the rate for minorities in District 1 is slightly lower than that for whites, the situa- tion is the opposite in District 2. The total enrollment is 600, consisting of 400 white and 200 minority students. Four of the white students (1 per- cent) and 18 minority students (9 percent) are enrolled in EMR classes, reflecting a relatively large disproportion. If the geographic unit's total alone is examined, there are 1,400 white students of whom 24 are assigned to EMR classes, yielding a 1.7 percent placement. There are 2,000 minority students, of whom 38 (also 1.7 per- cent) are in EMR classes. While the two percentages are the same at the state level, they disguise several more detailed outcomes the large dis- proportion in the small district and the variability between district prac- tices. This stems from the tendency of large districts to obscure data for small districts in aggregations. Districts that have no students classified in a special program inflate the state's total enrollment proportionate to the percentage of minority stu- dents in the district, distorting aggregate measures of disproportion fur- ther. For example, according to the 1978 OCR survey, in only 12 states did all districts report having EMR students. In 19 of the remaining states, more than 10 percent of the school districts reported having no EMR students at all, and in 8 states more than 25 percent of the districts reported having no EMR students. The average enrollment of 887 districts having no EMR students was 1,336, well below the average of 5,911 stu- dents in districts having EMR programs. While many smaller (often ru- ral) districts maintain other special programs, including those for trainable mentally retarded, emotionally disturbed, or specific learning-disabled students, about one third of the districts having no EMR programs do not operate any of these other programs. Thus, there are essentially two popu- lations of school districts represented in the survey data those with and those without EMR programs. Statistical information regarding racial or
OCR for page 332
332 FINN sex differences in EMR placement rates can be obtained only from the for- mer set. 7 Placement in special education programs is a district-by-district process, and a wide range of placement rates and racial disproportions may be found among districts operating under the same state guidelines. It is essential that an analysis of special education trends reflects this variability. SCORING DISPROPORTION FOR DISTRICTS AND STATES The 1978 OCR survey provides data from which placement rates and the disproportion index may be calculated for each school district. Of the 5,153 districts in the 1978 sample that have students enrolled in EMR pro grams, 236 districts do not have both whites and minorities in their stu- dent populations. The distribution of the log-odds index for EMR place- ments in the remaining 4,917 districts is given in Table 4 for all minorities combined. A log-odds index of 1.6 (Q = .66) separates 20 percent of the districts with the highest degree of disproportion from those with less dis- proportion; an index value of 2.1 (O = .78) separates 10 percent of the districts with the most extreme disproportions from the rest. The index values in every column of Table 4 have a nearly normal distribution and may be used in normal-theory statistical analysis (e.g., l-tests or F-tests). Small districts present a special problem to the investigation of special education placements, which is reflected in any measure of proportional- ity. A typical rural district or one in a small New England community, for example, may have 500 students of whom all but 20 are white. One stu- dent of the 20 classified as EMR results in a EMR rate of 5 percent for mi- norities. If two are classified as EMR, the minority rate is 10 percent, which is unusually high, and so on. In other words, in districts with a veer low number of minorities enrolled (or with a very low number of whites), small differences in the number of placements create large disproportions that may not reflect a serious problem of overrepresentation or underrep- resentation. Furthermore, if none of the minority students (or none of the whites) is in an EMR class, the odds for that group are zero, and the logarithm is not defined. Recent advances in the analysis of contingency tables provide methods for "smoothing" proportions so that they allow finer differences than the 5-10-15 percent values of the example above. The method of "iterative The proportion of the nation's school districts having no EMR programs may be larger than the OCR's 1978 survey indicates. In 1976, OCR surveyed all districts in the country, and ap- proximately 45 percent reported no EMR enrollment. The 1978 sampling plan may have tended to overrepresent those districts having EMR classes.
OCR for page 371
371 1 1 1 . C~ ._ 3 C~ o ~_ .o o~ & .= C~ 3 C~ ~_ - C~ C) s" = Ct - C) & V, ._ s~ - Ct C~ Ct Q) ~: ._ ~S .O 3-- - ,D Ct s~ ~o .~ & o _ & ~ ~o C) ~ ~, s" ~ o ._ = E ~ .~ cc, ~ = s s 2 s o ~ ~ _ CS~ ~ o ~ ~oo ~ _ ~ ~ ~ ~ oo oo l_ ~ ~ _ ~ ~ ~ _ _ o _ o o - ~ ~ ~ ~ '~ ~ - . ~ ~ o o o o o o o o o o o o o o o o o o o o o o 1 1 1 1 1 1 1 1 1 1 ~ ~ ~n ~ - ~ ~ ~ ~ oo O ~ u~ oo - - ~ ~ ~ ~ 0 r~ r~ 0 ~ oo ~ ~ 0 ro ~ 0 oo ~ ~ oo ~ ~ oo ~ ~ ~ ~ uO . . . . . . . . . . . . . . . . . . . . . . ~ ~ O ~ - - ~ - - ~C~ r~ r~ ~ ~ r~ r~ c~ ~ ~ ~ ~ oo ~ ~ ~ ~ _ oo ~ ~ ~ ~ ~ ~ ~ oo ~ ~ ~ ~o ~ ~ 0 ~ u~ ~ ~ ~ ~ _ ~ 0 ~ a~ ~ _ 0 u~ ~ Oo ~ 0 O ~ - - ~ - _ ~ r~ r~ ~ ~ ~ 0 ~ ~ ~ - - _ _ - ~ oo U~ ~ oo ~ oo 0 ~ ~ ~ ~ ~ ~ ~ ~ ~ ° O ~ ~ ~ - O ~ - ~ ~ ~ ~ r~ ~ ~ ~ 0 ~ u~ Lr . . . . . . . . . .. . . . . . . . . . .. O 0 0 0 0 0 0 - 0 00 0 0 0 - 0 0 - 0 0 0- 1 1 11 - ~ _ ~ ~ ~ ~ _ _ ~U) U~ _ ~ ~ 0 _ er ~ ~ ~ 0 0 oo ~ ~ _~ ~ ~ oo ~ _ ~ ~ U~ 0 _ . . . . . . . . . .. . . . . . . . . . .. _ O ~_ _ _ ~_ ~_ 0 0 0 0 0 - 0 0 0 - - 0 1 oo ~ ~ O oO ~ ~ ~ oO ~ ~ ~ 0 - 0 r~ ~ 0 0 0 - ~ ~ ~ ~ ~ ~ ~ ~ U~ ~ ~ ~ ~ 0 . . . . . . . . . . . . . . . . . . . . . . ~-----r ~- 0-0----~0--- O 00 ~ ~ ~ ~ - O ~ ~ ~ ~ ~D ~ ~ O ~ ~ ~ ~ u~ _ ~ 0 ~ ~ ~ Oo ~ 0 ~ cr 0 oo O O oO ~ - ~ ~ O -)t_ 00 `,0 t_ 00 _ ~ ~ ~1 ~ ~ ~- ~- ~1 r ~t- ~#_ _ r~ - ~ ~ ~ ~ O _ _ ~ \~' ~ ~ u: u~ C~ r~ ~ ~ ~oO C~ ~ O . . . . . . . . . . . . . Lr' _ 0 ~ ~ 0 ~ U~ 0 ~ ~ ~ 0 ~ ~ ~ ~ 0 - - r~ 0 ~o _ ~ uO ~ ~ _ O ~ 00 ~ 00 ~D ~ ~ O oo ~ - - - c~ ~ ~ ~ ~ - - ~ . u) ~ '~ . ~ ~ 'e o ~ o ~ ~ = ~ = ~ o u
OCR for page 372
372 FINN TABLE 19 Distribution of EMR Disproportion for Hispanic Students Minimum Maximum Number of Standard District Size Districts Mean Deviation Log-Odds Q Log-Odds Q Fewer than 1,000 students 124 1.08 1.71 - 4.30 - .97 7.41 .99+ 1,000 to 2,999 students 242 0.66 0.99 - 2.13 - .79 7.67 .99+ 3,000 to 9,999 students 232 0.47 0.85 - 2.11 - .78 6.94 .99+ 10,000 or more students 167 0.35 0.63 - 3.35 - .93 2.17 .80 All districts 765 0.64 1.12 - 4.30 - .97 7.67 .99+ NOTE: Except for the number of districts, which is the actual number in the sample, all results are the weighted projections to nationwide values; weights are the inverse of sampling probabilities. their Hispanic student participating in bilingual instruction. At the same time the largest districts have the lowest EMR disproportion for Hispanic students (see Table 19~. To explore the relationship of bilingual education with EMR place- ments further, districts in each of four size intervals were classified by geographic region and by the extent of EMR disproportion for Hispanic students. Three levels of disproportion were formed. The high group is composed of those districts whose disproportion was greater than one standard deviation above the mean for all districts in the size interval; the low group is composed of those districts whose disproportion was less than one standard deviation below the mean; and the medium group contains those districts in between. Mean scores for the three groups were com- pared by fitting a two-way fixed-effects analysis-of-variance model to the data, with the percentage of students in bilingual education as the cri- terion measure. The results are summarized in Table 21. Among districts in two size intervals the percentage of Hispanic stu- dents in bilingual programs is significantly related to disproportion, and a similar trend is seen in the smallest districts as well. In each case, districts with the highest disproportion levels have the smallest proportion of stu- dents in bilingual programs. It is possible that Hispanic students with poor English proficiency are misclassified as EMR when bilingual pro- grams are not available. It is apparent from the nationwide results (Table 3) that Hispanic stu- dents are placed in SLD programs to a somewhat greater extent than are
OCR for page 373
Patterns in Placement as Revealed by the OCR Surveys TABLE 20 Mean EMR Disproportion for Hispanic Students, by District Racial/Ethnic Composition 373 Percentage Percentage of Black Enrollment of Hispanic Enrollment O to 25 25 to SO 50 to 75 75 to 100 All Fewer than 1,000 students 0 to 20 0.86 2.40 0.91 20 to 40 1.010.55 0.99 40 to 60 1.09 1.09 60 to 80 1.07 1.07 80tolO0 3.73 3.73 All 1.080.55 2.40 1,000 to 9,999 students 0 to 20 0.510.78 0.13 - 0.55 0.51 20 to 40 0.630.10 0.62 40 to 60 0.81- 0.43 0.76 60 to 80 0.59 0.59 80 to 100 0.57 0.57 All 0.580.62 0.13 - 0.55 10,000 or more students 0 to 20 0.43- 0.03 0.02 - 1.17 0.32 20 to 40 0.440.43 0.31 0.43 40 to 60 0.30 0.30 60 to 80 0.69 0.69 80 to 100 0.23 0.23 All 0.430.13 0.09 - 1.17 NOTE: Average log-odds index is the weighted projection to all districts in the particular size category. Empty cells indicate fewer than two districts in the sample with the particular ra- cial/ethnic composition. nonminorities. The state-by-state results (Table 18) show that while the Hispanic percentage in SLD is lower than the nonminority percentage in many states, the reverse is true in states with high concentrations of His- panics (Texas and the southwestern states exclusive of California3. The dynamics that create the SLD difference are not apparent from the OCR data. It does not appear that SLD placements substitute for EMR place- ments, since a few states have high average disproportions in both classifi- cations simultaneously (New Mexico, Texas, and Wyoming. In fact, the correlation of SLD with EMR disproportion among Hispanic students is Hit is important to recall that these data represent only the 1978-1979 school year. If pressure increases to reduce EMR enrollments, it is possible that programs for specific learn- ing disabilities will become an alternative placement.
OCR for page 374
374 FINN TABLE 21 Mean Percentage of Hispanic Students in Bilingual Education Low Size of District Disproportion Medium High Disproportion Disproportion Fewer than 1,000 students 9.76 (11) 9.30 (101) 1,000 to 2,999 students 16.89* (29) 9.97 (184) 3,000 to 9,999 students 13.19 (37) 11.84 (166) 10,000 or more students 23.87* (18) 12.90 (123) 6.51 (12) 7.52 (29) 14.33 (29) 14.43 (26) NOTE: All percentages are the weighted projections to nationwide values; weights are the inverse of sampling probabilities. Actual sample sizes are in parentheses. *Significant differences among these three means at p < .05. +.33 for all districts combined,22 and close to this value for districts in each of four size intervals. Examination of the SLD rates for Hispanics and nonminorities (not shown) indicate that the correlation reflects dif- ferent placement rates for Hispanic students, while that for nonminorities is not related to EMR disproportion. The processes by which Hispanic students are referred and assessed for placement in both special programs need further investigation. In summary, the apparently similar EMR placement rates for Hispanic and nonminority students disguise enormous variation in practices among school districts. There are a number of districts in which Hispanic stu- dents are assigned to EMR programs in large proportions. They are dis- tinguished from other districts by having small enrollments that are of- ten but not always largely Hispanic; furthermore, they have small black enrollments, small or nonexistent bilingual programs, and high per- centages of Hispanic students in SLD classes as well. Among large dis- tricts with the greatest pool of resources, low EMR disproportion and low SLD disproportion occur where many Hispanic students participate in bi- lingual programs. Further research on factors affecting the availability and utilization of alternate programs for Hispanic students is certainly warranted. It would be important to determine to what extent specific learning difficulties are 22Statistically significant at p < .05, using a two-sided test.
OCR for page 375
Patterns in Placement as Revealed by the OCR Surveys 375 related to language or whether SLD programs, like EMR, may be used in some districts as a substitute for bilingual instruction. The criteria used for both EMR and SLD placements should be elucidated as well as the definition of these possibly amorphous categories and the actual instruc- tional programming that is provided. AMERICAN INDIANS OR ALASKAN NATIVES American natives comprise over 1 percent of the public school enrollment in 15 states, largely in the West. Their placement in EMR classes exceeds the rate for whites in all but three of the states; in Alaska the average log- odds index exceeds the 80th percentile value of 1.6. The largest racial dif- ferences in Alaska are in districts with fewer than 1,000 students, but the disproportion in larger districts is substantial as well. Also, higher degrees of disproportion are concentrated in districts of all sizes with 70 percent or more Alaskan-native enrollment. Other than in Alaska, the average log-odds index of disproportion is not sizable, and in several instances is zero or negative. In general, the dif- ference in the placement of American Indians and Alaskan natives in EMR classes is not large or even consistently positive throughout the states. For this group in particular, however, the OCR survey may not tell the complete story, since numerous American Indians are enrolled in special schools and special programs that are not represented in the usual public school sample. ASIAN OR PACIFIC ISLANDERS Students who are of Asian or Pacific Island origins are assigned to EMR programs at rates considerably below those of whites in 10 of the 12 states in which they comprise more than 1 percent of the school enrollment. The average log-odds index is negative in 8 of the 12 states, with most values substantially so. Thus, in general, overrepresentation of Asian or Pacific Island students in EMR classes is not a problem; these groups might even be studied to determine why their placement rates are low. Two states, however, have positive log-odds indexes of disproportion in excess of 1.6. In both Colorado and Nevada, larger disproportions occur in small school districts with low minority enrollment. Unfortunately, the OCR survey does not distinguish among Asian populations; it is possible that the students in these states are, for example, recent immigrants from Vietnam rather than Japanese or Korean children whose families have been established in the United States for longer periods of time. Newly ar
OCR for page 376
376 FINN rived immigrant populations present a unique opportunity to monitor special education placement rates as they develop. SOCIOECONOMIC STANDING AND SUSPENSIONS FOR SPECIFIC MINORITY GROUPS General relationships between socioeconomic status and suspensions with disproportion in EMR placements for all minorities combined are given in preceding sections. Correlations for each minority separately are pre- sented in Table 22. Suspensions are not correlated with disproportionate EMR assignment for any individual racial or ethnic minority. The same correlations for all minorities combined (Table 13) are positive. While EMR disproportions are accentuated by students of one minority group, it may be students of a different minority classification who are suspended. Thus, only an associ- ation for all minorities together is observed. There is a significant negative relationship between racial disproportion and socioeconomic status for each minority group except Asian or Pacific Islanders. The relationship is strongest for American Indian and Alaskan native students and least strong for students of Hispanic origin. However, the correlations for these groups, and blacks as well, are consistently negative. That is, disproportions even within a minority group tend to be smaller in districts serving populations with higher income levels. This relationship is worthy of further exploration to address such questions as whether individuals with higher income tend to live in suburban districts with lower overall EMR rates and also lower disproportional and whether the same behavior and school performance are treated differently in mid- dle- and lower-income districts. The answers to these questions may differ for particular minority groups and the attitudes and values associated with lower income for that population. DISPROPORTION AND STATE EMR CRITERIA To determine the extent to which state guidelines are associated with disproportion, information was obtained for 37 states on whether adaptive behavior assessments are required for EMR classification and the max- imum IQ score a child may have and still be labeled EMR. The states were classified by region and by whether adaptive behavior assessments were 23The correlation of socioeconomic status with the proportion of all students in EMR pro- grams for the total sample is-.3l, suggesting support for this three-variable hypothesis.
OCR for page 377
Patterns in Placement as Revealed by the OCR Surveys TABLE 22 Correlations of Log-Odds Index With Suspensions and Socioeconomic Status for Separate Minority Groups 377 Number of Suspensions of Proportion Racial/Ethnic Group Districts All Students Full-Price Lunches American Indian/Alaskan Native 817 - 0.03 - 0. 17* Asian/Pacific Islander 936 - 0.05 0.07 Black 3995 0.04 - 0.15* Hispanic 2681 0.00 - 0.10* NOTE: Correlations are the weighted projections to nationwide values; weights are the in- verse of sampling probabilities. *Statistically significant at p < .01 (two-sided test). required, and mean differences were tested by fitting a two-way analysis- of-variance model to the data, with several different criterion measures. The results are summarized in Table 23. There is no statistically significant difference between states that re- quire and those that do not require adaptive behavior assessment for EMR placement on any of the measures listed, including average IQ cut- off score, average size of the states' EMR programs (in terms of percent TABLE 23 Comparison of States Requiring and Not Requiring Adaptive Behavior (AB) Assessment for EMR Placement AB Required (20 States) AB Not Required (17 States) Correlation Correlation Standard With Standard With Variable Mean Deviation IQ Cutoff Mean Deviation IQ Cutoff IQ cutoff score U 73.10 3.92 74.42 3.97 Percentage of all students in EMR 1.61 0.94 0.11 1.43 0.69 0.0 EMR disproportion for race/ethnicity (log-odds) 0.44 0.93 - 0.15 0.59 0.90 - 0.59h EMR disproportion by sex 0.45 0.15 - 0.37 0.40 0.15 - 0.31 Percentage of white enrollment 73.65 25.44 0.33 79.56 13.44 o.50b U From Patrick and Reschly (1982). b Significant at p < .05 (two-sided test).
OCR for page 378
378 FINN age of students labeled EMR), disproportion either by sex or by race or ethnicity, or in terms of the average proportion of minority or nonminority students enrolled. Further, there is no interaction of region with the adap- tive behavior factor, indicating no exception to this generalization in any part of the country; also, when further control was added by employing "percentage of minority enrollment" in the state as a covariate, no signifi- cant differences appeared. Thus, the imposition of a state requirement that childrens' adaptive behavior be assessed as a necessary condition for EMR placement does not have a statistically noticeable impact on any of the outcomes investigated. This is due at least in part to the relatively wide variations in practice including the use or nonuse of adaptive behavior ratings among districts within the states. It contrasts strongly with find- ings that adaptive behavior limits EMR programs within individual school districts (Fisher, 1977; Mercer, 1973~. Two of the measures have a significant correlation with the state IQ cut- off but only in those states not requiring adaptive behavior assessments. EMR disproportion by race or ethnicity is correlated negatively with statewide IQ cutoff scores. That is, on the average, the lower the IQ cutoff score i.e., the more stringent the EMR criteria the greater is the rela- tive assignment of minority students to EMR classes. This is predictable for states in which adaptive behavior assessments are not made regularly, since EMR placements become more nearly a function of children's IQ scores. When adaptive behavior is included as an additional required assessment, however, the correlation with IQ cutoff score is reduced to . . ~ nonslgnl~lcance. The statewide IQ cutoff score is also correlated with the percentage of white enrollment in states not requiring adaptive behavior assessments. While this reflects a trend for states with greater proportions of white students to set higher cutoff scores for EMR classification, the motivation for this practice is not revealed from the survey data. REFERENCES Bishop, Y. M. M., Fienberg, S. E., and Holland, P. W. 1975 Discrete Multivariate Analysis: Theory and Practice. Cambridge, Mass.: MIT Press. Fisher, A. T. 1977 Adaptive Behavior in Non-Biased Assessments. Revised version of paper presented at the meeting of the American Psychological Association. ERIC Document Reproduction Service no. ED 150 514. Mercer, J. R. 1973 Labeling the Mentally Retarded: Clinical and Social System Perspectives oil Me''- tal Retardation. Berkeley, Calif.: University of California Press.
OCR for page 379
Patterns in Placement as Revealed by the OCR Surveys 379 Patrick, J. L., and Reschly, D. J. In Relationship of state education criteria and demographic variables to school system press prevalence of mental retardation. American Journal of Mental Deficiency 86. U.S. Department of Health, Education, and Welfare 1978a Fall 1976 Elementary and Secondary School Civil Rights Survey. Final File Documentation. Office for Civil Rights. June. Washington, D.C.: U.S. Depart- ment of Health, Education, and Welfare. 1978b 1978 Elementary and Secondary Civil Rights Survey. Sample Selection. Office for Civil Rights. February. Washington, D.C.: U.S. Department of Health, Educa- tion, and Welfare. Zoloth, B. S. 1976 Alternative measures of school segregation. Land Economics 52:278-298. APPENDIX EXAMPLES OF SMOOTHING DATA FOR SMALL DISTRICTS The 1978 OCR survey indicates 11 districts in Georgia with fewer than 100 minority students enrolled or fewer than 100 whites. When the numbers of students in these districts are summed, the proportions of students in EMR and in no special programs are as follows: Minority White EMR No special program 0.0065 0.0131 0.1054 0.8751 In the following table the EMR odds for minorities is 0.0065/0.1054 = 0.0617 and for whites 0.0131/0.8751 = 0.0150. One specific district has the following numbers of students: Minority White EMR 1 11 No special program 31 922 The EMR odds for minorities is 1/31 = 0.0323; one additional EMR stu- dent would bring the ratio to 0.0645, and no value between the two is pos- sible. The odds for whites is 11/922 = 0.0119, and the difference is 0.0323 = 0.0119 = 0.0204. The smoothed frequencies for this district are as follows:
OCR for page 380
380 FINN Minority White EMR No special program 1.0393 1 1.0119 31.5270 921.4218 From these values the EMR odds for minorities is 0.0330 (a small degree closer to the statewide value of 0.0617) and for whites 0.0120. The dif- ference of 0.0330 - 0.0120 is 0.0210, which is close to the original value. Another of the 16 districts with small enrollment has the following num- ber of students: Minority White EMR 21 0 No special program 194 65 The EMR odds for minorities is 21/194 = 0.1082, for whites 0/65 = 0, and the difference is 0.1082. The zero value raises such questions as does zero of 65 students, for example, mean as much as zero of 100 or of 500 students? Would the number remain zero if the white enrollment were in- creased, as may happen from one school year to another, or is this value a stable zero? A partial answer may be provided by examining the larger statewide data set, in which the odds for whites is small but is nonzero (0.0150~. Smoothing the district's frequencies yields the following results: Minority White EMR No special program 20.9588 0.0079 193.6464 65.3870 The odds for minorities is 0.1082, for whites 0.0001, and the difference is 0.1081, which is close to the original value.,~While the original zero value did not allow calculation of the log-odds index, the adjusted values yield In (0.1082)-In (0.0001) = 6.80.
OCR for page 381
Patterns in Placement as Revealed by the OCR Surveys 381 The smoothing procedure used in this analysis involves obtaining "pseudo Bayes estimates" of actual population frequencies in the manner described by Bishop et al. (1975: Section 12.1.1~. This method has distinct advantages over the widely used practice of adding 0.5 to each cell count, especially when the total number of observations in one or both columns is . ~ small.
Representative terms from entire chapter: