Relationships among states' fiscal and demographic data and the implementation of P.L. 94-142.
This study used national data sources to explore the relationships among certain state-level special education, fiscal, and demographic variables. The primary purposes of the research were (a) to discover to what extent differences in special education identification and integration rates were associated with specific states' fiscal and demographic characteristics, and (b) to demonstrate the feasibility of using existing national databases in special education policy research.
Federal education agencies are the repositories for numerous large-scale databases, which have been underused for research (Burstein, 1984). The chief national database in special education is contained in the Annual Reports to Congress. Since 1977, states have been reporting information such as the number of students they serve in special education programs, by category of disability; the types of educational placements in which students are served; and the number of teachers and other professionals employed. This data set represents, on a macroscopic level, the status of our nation's implementation of special education programs.
The analytic studies that have used portions of these data have been descriptive for the most part (Brinker & Thorpe, 1985; Forness, 1985; Gerber, 1984; Gerber & Levine-Donnelson, 1989; Hallahan et al., 1986). Recently, Danielson and Bellamy (1989) used data from the Tenth Annual Report to Congress to analyze, across states, the use of various placement options for students with disabilities. This study demonstrated that considerable variation among states existed in the implementation of the least restrictive environment provision of Public Law 94-142, now called the Individuals with Disabilities Education Act (IDEA). As a result, the authors suggested that further research on state characteristics was warranted to examine factors influencing implementation of special education policies.
A study by Noel and Fuller (1985) examined the Annual Report data for the 1976-77 and 1980-81 school years to explore relationships to selected demographic and fiscal characteristics of states. That research revealed the existence of some tenuous relationships between a state's relative wealth and proportion of rural and minority populations and the identification rates for specific disabilities. The present study expands on that preliminary work by exploring the interrelationships among a broader number of both fiscal and demographic characteristics and data regarding states' special education implementation.
This research has attempted to assess whether state-level data obtained from existing and diverse data sources could be used to investigate selected features of individual states' implementation of P.L. 94-142. The research examined whether variability in the selected fiscal or demographic characteristics was systematically related to identification and integration rates. To the extent that meaningful relationships could be identified and understood, the effort expended in collecting, maintaining, and bringing together extant data sources can benefit both program evaluation efforts and policy analyses.
THE DATA SET
The information for this study was drawn from a large database, created by compiling and merging numerous data sets containing information on the economic, social, and general and special educational characteristics of the 50 states and the District of Columbia. The information came from The Annual Reports to Congress on the Implementation of PL. 94-142 and from government reports, tabulations, and data summaries available in the public record or through federal agencies like the National Center for Education Statistics (NCES) and the U.S. Census Bureau. Collectively, the data set is a comprehensive collection of national statistics.
Much of the extant data was reformatted and subsetted to create a new database consisting of a total of 366 variables. With few exceptions, data on all variables were available for each of 3 school years (1976-77, 1980-81, and 1983-84). These years were chosen because they represented: * The first year that data were reported on the implementation of P.L. 94-142. * A midpoint in the implementation process and a time when reporting procedures should have been routinized within the states (as well as a time when major federal educational policies were changed as a result of the consolidation of several large funding programs). * The most recent data available at the time of the study for certain statistics, specifically data obtained from NCES.
The data provided a means for placing special education variables within a state context and examining how they operate within, and are influenced by, the broader spectrum of state-level socioeconomic characteristics.
Description of Dependent Measures
Identification. Because the identification and integration of students with disabilities are major issues in the implementation of special education legislation, they were chosen as the dependent measures in these analyses. The identification rates were calculated for each of the 3 years under study for total special education students and for students within three categories. The specific disability categories were learning disabled (LD), seriously emotionally disturbed (ED), and multiply disabled (ML). Identification rates were calculated by dividing the student count by the total K-12 public school enrollment for a given year and converting to a percentage.
The LD category was selected because it is the largest group of students with disabilities and has also been the subject of much controversy regarding the identification criteria. The inconsistency in identification for students with serious emotional disturbance also suggested that this category might be reactive to contextual factors. Finally, the category of multiple disabilities was selected to represent students with more severe disabilities. However, this category was added after passage of P.L. 94-142; thus, child count data were unavailable before 1980.
Integration. The measures of integration were cumulative placement rates calculated according to the procedures used by Danielson and Bellamy (1989). This statistic represented the number of children in a given placement plus those in all more restrictive placements, as a function of the total school-age population within that state. The authors suggested:
The cumulative placement rate statistic allows
one to ask what percentage of school-aged
students in a state are served in a particular
educational placement [regular classrooms,
separate classrooms, separate day schools, and
other environments (including separate
residential schools and home or hospital
environments)] and all more segregated
placements. (Danielson & Bellamy, 1989, p.
449) The authors reported figures in rates per million, as does the present study.
Description of Independent Variables
The state context variables included in this study were four measures of state financial resources and three measures of state demographic conditions. These seven variables were selected as independent variables from among the large number of fiscal and demographic variables in the data set after extensive consideration of the dimensions of interest. For example, several measures of educational revenues and expenditures were available, but per-capita income and per-pupil expenditures were selected for analyses. The former represented a state's available wealth, and the latter was more indicative of a state's financial commitment to education. In addition, a variable related to a state's reliance on targeted federal education aid (e.g., economically disadvantaged, bilingual, etc.), excluding the P.L. 94-142, Part B dollars, was selected as a measure indicative of the extent of the state's special populations and the relative strength of the federal education programs within a state. Finally, to obtain an additional measure of state fiscal support for education, a variable was created by computing the percentage of all nonfederal educational revenues that were from state sources.
In addition to the fiscal variables, three demographic variables were selected: rural school-age population, minority public school enrollments, and percentage of children enrolled in school who are living in poverty. These three were chosen because of the acknowledged importance of these factors in educational programs.
To examine whether variation in state-level identification and integration rates was related to finance or demographic characteristics, bivariate correlations between each of the dependent (special education) and independent (state context) variables were produced. Correlational analyses were aimed at ascertaining whether systematic, linear relationships existed between states' implementation efforts and their fiscal or demographic characteristics. The presence of a nonzero bivariate relationship indicates that the relative position of states on one measure is similar to their relative position on another. Within-year correlations between each dependent and each independent measure were computed and tested for significance.
In addition, states were ranked on each of the independent variables and assigned to quartiles based on their relative position. Using rankings to group states with similar positions on the contextual variables provided a straightforward way of comparing states on their implementation efforts. The rationale for grouping states is based on the criticism that a mere ranking of states does not take into account state contextual differences.
Ginsberg, Noell, and Plisko (1988) highlighted some of the criticisms of attempts to make state-by-state comparisons in a discussion of the response of the educational establishment to the publication of the Secretary of Education's "Wall Chart," which compares certain test scores of students across the United States. They recommended ways to improve state comparisons, most particularly by moving beyond rankings alone to including state context measures and grouping states using these measures. The authors of the present study believed that, by grouping states with similar contextual features, a better understanding of variations among states in their implementation of IDEA would result.
For the purpose of this study, it was of interest to test whether the quartiles, which consisted of groups of states with similar relative positions, differed significantly on any of the special education measures. Analyses of variance (ANOVAS) were performed to test the mean differences among the quartiles on each of the special education variables for each year. When ANOVA was significant, a test of linearity was conducted to answer the question of whether the quartile means could be described with a linear function. Because this research was exploratory, significance level was set at .05.
Intercorrelations showing the relationships among the state context variables appear in Table 1. Table 2 presents the means and standard deviations of the variables; statistics are presented for each of the 3 years. (The demographic data were available for the 1980 Census only.) A more detailed description of the categorical independent variables follows. [TABULAR DATA 1 OMITTED]
Means and Standard Deviations for Independent Variables, 1976-77, 1980-81, and 1983-84
Variable 1976-77 1980-81 1983-84
Mean 1,589 2,458 3,197 SD 395 661 1,031 N 50 50 50
Mean 6,423 9,540 11,590 SD 1,088 1,379 1,852 N 50 50 50
Mean 9.76 9.40 6.65 SD 4.15 3.91 2.92 N 50 50 50
Mean 50.34 53.19 53.15 SD 17.75 17.39 17.44 N 50 50 50
Mean -- 36.00 -- SD -- 15.25 -- N -- 50 --
Mean -- 19.44 -- SD -- 15.85 -- N -- 50 --
Mean -- 14.35 -- SD -- 4.55 -- N -- 50 --
Note: PPEXP = per-pupil expenditures; PIPC = per-capita personal income; ADJFER = percentage of total educational revenue from federal sources; STPCT = state share of nonfederal revenue; RURAL = percentage of school-age children living in rural areas; MINORITY = percentage of school-age children who are minority; POVERTY = percentage of related school-age children living in poverty.
Per-Pupil Expenditures (PPEXP). These data, obtained from the Digests of Education Statistics and NCES, represented the annual expenditures per pupil for elementary and secondary public education. Figures were in current dollars.
States in the lowest quartile on this measure averaged between $1,090 and $1,305 in 1976, whereas those in the highest quartile averaged from $1,784 to $3,389 in that year. By the 19884 school year, these figures had doubled, nevertheless maintaining a large discrepancy in financial resources between states at the highest and lowest end of the scale.
Per-Capita Personal Income (PIPC). This variable, obtained from the Survey of Current Business, U.S. Department of Commerce, represented the average taxable income for each citizen within a state, in current dollars. The lowest quartile on this measure averaged between $4,662 and $5,513 in 1976, whereas the highest quartile averaged from $7,004 to $11,599. The figures for 1983-84 were about double the 1976 figures.
Percentage of Total Educational Revenue From Federal Sources (ADJFER). This variable was created by subtracting the funds awarded under Part B, obtained from the Annual Reports to Congress, from the federal funds awarded to a state, as reported in the Digests of Education Statistics. The federal contribution to state educational revenue ranged from 4.6% to 22.9% in 1976. States in the lowest quartile received 6.3% or less of their revenue from the federal government. For the highest quartile, this figure was at least 12.1%. By 1983-84, these percentages had dropped by about one third.
State Share of Nonfederal Revenue (STPCT). The relative contributions of state and local governments to state educational revenue can differ markedly, depending on historical trends or perceptions held by each regarding their role in supporting education. Differences in tax bases and funding priorities are also influencing factors. The measure used in these analyses represented the percentage of the state's nonfederal education revenue receipts for public elementary and secondary schools obtained from state sources. Data were taken from The Digest of Education Statistics and are reported in the Condition of Education. In 1976, the states in the lowest quartile on this measure received between 8.82% and 38.94% of their nonfederal revenue from state sources, and those at the highest level received from 65.37% to 100%. These figures did not substantially change by 1983.
The fiscal variables were obtained from the extant databases and reported in current dollars. The dollar figures were not adjusted for inflation for several reasons. First, the research examined relationships of the fiscal variables to the special education variables and then analyzed mean differences among quartiles, which were created by grouping states with similar rankings on the independent variables. Therefore, it was the relative position of states, not the absolute dollar values of the fiscal variables, that was examined. Further, the usual inflation adjustment used by the Consumer Department would apply a constant factor to each of the figures, thus not altering the relative standing of states. Finally, as shown in Table 1, correlations among the fiscal variables across the 3 years were extremely stable.
Percentage of School-Age Children Living in Rural Areas (RURAL). This variable, based on the 1980 Census, represents the number of persons 3-17 years of age living in rural areas as a percentage of the total number of people in that same age group. The lowest quartile states had between 9.33% and 21.10% of their child population living in rural areas. For the highest quartile, these figures ranged between 50% and 70.12%.
Percentage of School-Age Children Who Are Minority (MINORITY). Census data also provided figures for the number of related children 3-17 years old who were enrolled in public schools, with breakdowns by minority status. To compute the percentage of children having minority status, the categories (a) black; (b) Spanish origin; (c) Asian and Pacific Islander; and (d) American Indian, Eskimo, and Aleut were added, then taken as a percentage of the total. In the lowest quartile, the percentage of minority children ranged between zero and 6.8%; the highest quartile ranged from 30.75% to 75.14%.
Percentage of Related School-Age Children Living in Poverty (POVERTY). This variable, also from the 1980 Census, represented the percentage of related children enrolled in school and living in families below the poverty level. The lowest quartile had 7.4%-10.4% of their children living in poverty; the highest quartile figures ranged between 17.6% and 29.8%.
Nationally, average identification rates for students served in special education programs increased over the 8-year period by about 35%, from 7.70% in 1976 to 10.42% in 1983. For the learning disabled category alone, the average identification rate more than doubled between 1976 and 1983 (2.12% vs. 4.62%, for each of the years, respectively). Average identification rates for the emotionally disturbed category increased 59%, from.532% to.846%. For the multiply disabled category, average identification rates held steady between 1980 and 1983, at about .13%.
Between 1976 and 1983, the national average for cumulative placement rates of students in special education classes (plus all more restrictive environments) increased 27%, from 25,211 students per million to 32,064 per million. The average cumulative placement rates in separate schools (plus the most restrictive environments) increased by 24% during the same time period, from 5,984 students per million to 7,388 per million. Average placement rates in "other" environments (i.e., most restrictive, including residential schools, institutions, and homes/hospitals), decreased nationally by 23%, from 1,684 students per million in 1976 to 1,306 per million in 1983.
RESULTS OF CORRELATIONAL ANALYSES
Bivariate correlations between the identification rates for total special education and the three disability categories and the finance and demographic variables are shown in Table 3. Overall, a number of correlations were significant; however, most were below .40, indicating weak to moderate relationships between the independent and dependent variables. [TABULAR DATA 3 OMITTED]
Total special education identification rates demonstrated a moderate relationship to only one of the seven independent variables, and this correlation was significant only in 1976. In that year, there was a tendency for states with high percentages of rural school-age population to identify fewer special education students (r = -.36). Analyses of variance of the total special education identification rates by RURAL quartiles was significant for 1976, F(3,46) = 4.12, p < .01. The test of linearity was also significant for that year, F(1,46) = 8.74; p <.01. In 1976, the average identification rate for total special education for the lowest rural quartile was higher than the average rate for states in the highest quartile (8.28% vs. 6.32%). These differences had all but disappeared by 1983-84, when the mean identification rate for the highest quartile was 10.61% versus 10.01 % for states in the lowest quartile (r = -.10).
There were a greater number of significant correlations between identification rates for the specific disability categories and the independent variables. This was particularly the case for the learning disabled category. States with greater financial resources (higher PPEXP and PIPC) tended to identify more learning disabled students (correlations ranged from .31 to .43). Table 4 shows that, in the 3 years under study, LD identification rates for states in the highest quartiles on PPEXP and PIPC were between 20% and 50% higher than those states in the lowest quartiles. The ANOVA revealed that the PIPC quartile means differed in 1980, F(3,46) = 2.83, p <.05, and in 1983, F(3,46) = 3.97, p < .05. The tests of linearity for each of those 2 years also were significant, F(1,46) = 4.87, p < .05; F(1,46) = 11.27, p < .0 1. Similar results were obtained for the analysis of PPEXP quartile differences on LD identification rates. The overall ANOVA was significant for 1980, F(3,46) = 4.88, p < .01, and for 1983, F(3,46) = 4.86, p < .01, as were the tests of linearity for each of those 2 years, F(1,46) 13.39, p < .01; F(1,46) = 11.72, p < .01. [TABULAR DATA 4 OMITTED]
Negative correlations observed between LD and ED identification rates and ADJFER were weak but significant (range from -.24 to -.35) and suggested that states with greater federal educational assistance tended to identify fewer LD and ED students. However, the ANOVAS of LD and ED identification rates by ADJFER were not significant.
In terms of demographic variables, significant negative correlations were observed for LD and ED identification and RURAL (range from -.32 to -.43). For all 3 years, states with higher percentages of rural children identified fewer learning disabled and emotionally disturbed students (see Table 3). The quartile means (Table 4) suggested that the average LD identification rate for states in the highest rural quartile was lower than that of the lowest quartile. However, analyses of variance of LD identification rates by rural quartiles were not significant.
Disparities between the highest and lowest rural states were more pronounced for the ED identification rates. The average ED identification rate for states in the highest quartile of rural children was lower than that of the lowest quartile. The ANOVA revealed an overall significant difference for 1976, F(3,46) = 4.24, p < .01, and 1980, F(3,46) = 3.20, p < .05. Tests of linearity for each of these 2 years were also significant: 1976, F(1,46) = 12.42, p < .01; 1980, F(1,46) 9.50, p < .0 1.
POVERTY was also significantly and inversely correlated with LD identification rates in 1976 and 1980, indicating a tendency for states with higher percentages of children living in poverty to identify relatively fewer LD students (r = -.43; r = -.40). In 1983 there was a weak but significant relationship between ED identification rates and POVERTY (r = -.25); however, results of these ANOVAS were nonsignificant.
No marked relationships between identification rates for the multiply disabled category and any of the independent variables were evident for the 2 years for which those child-count data were available.
Correlations between the integration variables and the finance and demographic variables are presented in Table 5. Quartile means and standard deviations for cumulative placement rates in special classes and separate schools and selected independent variables are presented in Table 6. Because cumulative placements in regular classes are confounded with identification rates, the significant inverse correlation observed in 1976 between RURAL and cumulative regular class placement would be interpreted similarly to the inverse correlation between RURAL and total special education identification rates. The ANOVA was nonsignificant for 1976. [TABULAR DATA 5 OMITTED]
Cumulative placement rates for more restrictive placements were related to the independent variables, RURAL and PIPC. In 1980 and 1983, significant inverse correlations, in the weak to moderate range, were observed between percentage of rural child population and use of special classes and all more restrictive placements (r = -.37 for both years), and use of separate schools, plus most restrictive placements (r = -.25 and -.30, respectively). The ANOVAS of cumulative special class placement by RURAL was significant in 1980, F(3,46) = 6.17, p <.01, and 1983, F(3,46) = 3.89, p < .05, as were tests of linearity for both years, F(1,46) = 7.24,p <.01; F(1,46) = 8.05,p < .01. Analyses of variance of cumulative placements in separate schools was significant only for 1983, F(3,46) = 2.5 3, p < .05, as was the test of linearity, F(1,46) = 5.24, p < .05.
Significant positive relationships were also observed in 1980 and 1983 between PIPC and the cumulative placement rates for special classes and more restrictive environments (r = .29 and .28, respectively). Analyses of variance of cumulative placements in special classes by PIPC was significant for 1980, F(3,46) = 2.91, p < .05, and 1983, F(3,46) = 3.10,p < .05, as were tests of linearity for both years, F(1,46) = 6.50, p < .05; F(1,46) = 8.49, p < .01. Finally, in 1983 only, there was a weak but significant inverse correlation between PPEXP and cumulative placement in "other" environments (hospitals, home, etc.). Analysis of these quartile means was nonsignificant.
The purpose of this study was to investigate what bivariate relationships exist among certain state-level fiscal and demographic variables and special education identification and integration rates. As such, the research was exploratory. Because use of such contextual data is increasing within education to provide information about inputs or resources contributing to specific educational outcomes (Ginsburg et al., 1988; Oakes, 1989), it was also important to examine whether broad contextual variables could be useful in examining policy trends in special education. As noted by Keogh (1988),
Information ... has documented that referral, identification, and instructional decisions are influenced by a variety of conditions within both educational systems and governmental agencies, as well as by broad sociopolitical considerations. Thus, it is essential that program
analyses be broad-based and include sources of
influence on many levels. (p. 245)
The present study analyzed the implementation of two specific indicators of federal special education policy, using a broad set of fiscal and demographic variables. To the extent that relationships were demonstrated and understood, the percent research could be considered promising for future policy analyses in special education.
It is important to acknowledge, however, that local determination and local resources are powerful in directing implementation of special education policy. Substantial variation in wealth and educational practice exists within a state's borders, and this variation cannot always be accurately represented at the state or national level. Thus, state-level data provide only global indicators for policy analyses. Nonetheless, if such state variables are shown to relate to specific implementation features, such as identification rates or use of specific placements, they may prove to be useful in explaining local variation, as well as anticipating future implementation of statewide policies.
Before discussing the findings specific to identification and placement, we will make a few observations about the fiscal and demographic contextual variables. First, correlations remain strong across the 3 years for a given variable. Also, the independent variables do not represent unrelated phenomena. As might be anticipated, per-capita income and per-pupil expenditures were highly correlated. Accordingly, there was a moderate but inverse correlation between per-capita income and the amount of nonspecial education federal aid received within a state. This makes sense given that most of the federal aid is targeted to economically disadvantaged students. This federal aid variable was also associated with higher minority populations and states with higher proportions of children living in poverty. States in the latter group also had lower per-pupil expenditures (and, of course, lower per-capita incomes). States with high federal aid also made relatively higher contributions of their state monies to local education.
A state's rural status was inversely and moderately related to per-capita income in each of the 3 years, but demonstrated a weak relationship to per-pupil expenditures only in 1976. Thus, although rural states have a consistently lower tax base, per-pupil expenditures, or actual outputs for education, were not tied to rural states.
Identification and Demographic Variables
The findings of the present study indicate that two state-level demographic characteristics were related to identification of special education students in the beginning years of implementation of P.L. 94-142. The percentage of rural school-age population was the only state-level variable related to total special education identification rates, and that was for the earliest year, 1976, only. This inverse relationship indicated that the less populated states identified fewer special education students in the first report year, but quickly caught up in total numbers of students identified by 1980 and 1983. Interestingly, however, inverse relationships between rural school-age population and LD and ED identification rates remained consistent for each of the 3 years, indicating that rural states continued to identify fewer of each of these types of students than did their more populous counterparts.
These data suggest that rural states were slower to identify or report students as disabled, perhaps due to less well-developed special education systems and fewer programs for students with various disabilities. Because LD was a new category of disability in 1976, more rural school districts may have had neither opportunity nor impetus to, begin such programs. Further, given the problems of high costs, stigma, and general lack of trained personnel associated with establishing programs for students with emotional disturbance, it is also likely that more rural systems did not begin such categorical programs and perhaps chose to identify fewer such students.
The relationships between the percentage of school-age children living in poverty and identification rates for LD and ED differed somewhat from those for rural states. Although there was some tendency for states with more poor children to identify fewer LD in 1976 and 1980, a significant relationship was not observed in 1983. Poverty was not particularly related to the percentage of rural population, yet the "catch up" phenomena is again suggested. Though LD rates climbed nationally during the same period of time, it may be that poorer states were slower to establish programs for this newer category of disability.
Fiscal Variables and Identification
The two fiscal variables demonstrating the strongest and most consistent relationship to any of the identification rates were per-capita income and per-pupil expenditures. These variables, which were highly correlated, provide an indication of general state wealth, as well as the money given to education. These variables were positively related to LD identification for each of the 3 years. These results are supported by the earlier discussion regarding the negative relationship between poverty and LD identification, which together suggest that LD was most reactive to overall state wealth and to a lesser degree influenced by the amount of money expended on education.
A related wealth factor is the amount of nonspecial education federal money a state receives. This variable was inversely related to both LD and ED rates; but the relationships were weak and the analyses of quartile means nonsignificant. Nonetheless, the direction of the relationships suggests that states with greater availability of federal support for programs, which is targeted primarily for economically disadvantaged students, tended to identify fewer students as LD or ED. This could perhaps suggest that having other programs focusing on ameliorating failure for students at risk may slightly reduce identification into special education programs. However, it should be noted that per-pupil expenditure and per-capita income are significantly related only to LD identification. That is, irrespective of how much a state receives for remedial or extra programs, states with greater wealth identify more LD students.
A number of factors could contribute to this phenomena. For example, richer states that may have more school districts that are affluent and suburban may also have more challenging school programs, which create greater opportunities for students to experience difficulty and perhaps end up in LD programs. The finding could also reflect a culture of referring to special education students who experience any academic or behavioral problem. This is likely a result of a long history of special education programs in a state or district, which in turn may be related to more affluence in the state. At one level, these relationships appear to support the notion that LD identification is to some degree a middle-class phenomenon.
Contextual Variables and Placement Rates
Perhaps the most interesting observation regarding the relationships among the demographic or fiscal factors and cumulative placement rates is that, though all correlations were weak, they were significant only in 1980 and 1983. That is, placement data, because of either data reporting problems or actual placements used, did not appear to respond to contextual factors in the first report year.
There is much difficulty in attempting to interpret the meaning of the relationships, regardless of strength, due to construction of the cumulative placement variable. A cumulative placement rate means placement in one type of class and all more restrictive placements. Thus, cumulative placement in regular classes and other settings is highly confounded with identification rates. This likely accounts for the direction of the relationships between cumulative regular-class-placement rates and the contextual variables.
Cumulative placement rates in special classes or separate schools and all more restrictive placements would appear to lower student/teacher ratios, increase use of separate physical facilities, and result in higher costs. This may explain why states with higher per-capita income tended to have higher cumulative placement rates in special classes in both 1980 and 1983. Given the relationship of income to rural population, cost may also explain the negative relationship between that variable and cumulative special-class-placement rates. In addition, rural districts may not have the sufficient numbers or critical mass of students required to establish a special school or special class, thus facing use of less restrictive placements.
Observations Regarding Nonrelationships
Several independent variables demonstrated no relationship to either identification rates or cumulative placement rates. For example, measures of minority student population essentially did not relate to any of the dependent measures. Other research suggests that minorities are overrepresented in special education, but those trends were not reflected in these analyses. In this study, however, the measure of minority status included all nonwhite racial and ethnic categories reported by the U.S. Census. Also, these analyses did not examine all categories of disabilities. Perhaps if individual races or ethnic groups and disability categories were analyzed separately, the patterns would differ. For example, the category of mental retardation was not included in the present analyses; however, some (e.g., Heller, Holtzman, & Messick, 1982; Reschly, 1988) report the overrepresentation of African-American students in that particular disability category. In addition, it is important to remember that the data included in these analyses represented an aggregation of local school district data.
The relationships between the ML dependent variables and most of the independent variables were almost nonexistent. There may be several reasons for this. As a reporting category, this classification is newer than the others and has also undergone some definitional changes (e.g., the removal of the deaf-blind classification). Further, the low incidence may result in measures that are quite unstable. Thus, it is possible that the measures of ML may be limited in their ability to show systematic relationships with the independent variables. More data points are needed before any firm statements can be made regarding the relationships of ML variables to the independent variables.
Another matter of interest is that the variable related to state share of educational expenditure (STPCT), which is frequently used in educational finance as an important marker for state educational wealth (Odden, 1985), did not relate to special education programs. Perhaps the general wealth of a state, as measured by PIPC, is a proxy for some larger educational strength or other population or school characteristic that impinges uniquely on special education programs.
In designing this study, we gave a great deal of consideration to data quality issues. The database which was used for these analyses is virtually complete for the variables and the 3 years under study. Because the data generally behave as some would expect given what is known about major organizational, regulatory, service delivery, and budgetary changes since the mid- 1970s, it seems likely that they provide a valid view of state-level practices. As for reliability, the consistency of states in reporting and interpreting data is largely unknown. Some (e.g., Gerber, 1984) have argued that the special education data are so flawed as to be useless. However, while acknowledging the problems with program record data, we must note that many relationships evident in these analyses confirm what is already known about the evolution of special education practice in the United States. Thus, it is presumed that the flexibility in reporting has not been strong enough to conceal all true relationships. Moreover, the results reported here are interrelated. This study made an initial attempt at understanding the effects of the fiscal and demographic variables at a simple level. However, we recommend that future research use a multivariate approach to disentangle the effects of these variables and determine their independent influences.
Of further methodological interest are two issues, one related to the cumulative-placement statistic, and the other regarding the total and disaggregated special education categories. The measures of special education integration analyzed in this study were cumulative placement rates similar to those developed and analyzed by Danielson and Bellamy (1989). This statistic provides a useful means of ascertaining state variability in the use of more or less segregated environments, but it is less useful for making comparisons among placements. Because the various cumulative-placement measures confounded the less and more restrictive environments, comparisons between placements were not possible. This confounding also complicated the interpretation of relational statistics, because it was not clear whether the presence or absence of a relationship was due to the nature of the measure or the phenomenon under study. A better variable may be the actual placement rate per the number of students identified as having a specific disability. A final consideration is the use of aggregated special education categories. Variables based on total special education populations necessarily combined students with widely different educational needs. The fact that these measures showed very little relationship with other variables may be due to the effects of some categories' diminishing or nullifying others. This appears to be supported by the observation that individual categories, such as LD and ED, when analyzed separately, were more reactive to the independent variables than the total special education variables. Also, the ML category, when analyzed alone, showed few systematic relationships with the independent variables. Despite the small sample sizes, however, the inclusion of this category of students with the greatest educational needs in the total perhaps contributed to confounding relationships with the total special education category.
These results appear to confirm common understandings regarding the overall capacity of states to respond to special education demands, as well as the spurious nature of LD, and to a lesser degree ED, identification. Given the extreme subjectivity inherent in the identification criteria for both categories, as well as the political nature of the identification process, these two categories can be more reactive to contextual variables. That is, states with greater wealth may in fact have larger, well-established special education systems that allowed for identification of more LD and ED students from the inception of special education programs. Similarly, these states may also be able to use the more expensive segregated placements.
The results of this study support the notion that existing data, collected to monitor the operation of the federal special education programs, as well as other extant national statistics, can be used to examine implementation of public policy. Although the results are inconclusive, they point to some interesting and potentially significant issues regarding contextual factors that have influenced special education programs. This research represents only a preliminary step in understanding the influence of state-level socioeconomic factors on identifying and serving the nation's children with disabilities. Further exploration of these relationships, especially in a multivariate context, is warranted.
REFERENCES Brinker, R. P., & Thorpe, M. E. (1985). Some empirically derived hypotheses about the influence of state policy on degree of integration of severely handicapped students. Remedial and Special Education, 6(3), 18-26. Burstein, L. (1984). The use of existing data bases in program evaluation and school improvement. Educational Evaluation and Policy Analysis, 6(3), 307-318. Danielson, L. C., & Bellamy, G. T. (1989). State variation in placement of children with handicaps in segregated environments. Exceptional Children, 55, 448-455. Forness. S. R. (1985). Effects of public policy at the state level: California's impact on MR, LD and ED categories. Remedial and Special Education, 6(3), 36-43. Gerber, M. M. (1984). The Department of Education's Sixth Annual Report to Congress on P.L. 94-142: Is Congress getting the full story? Exceptional Children, 51, 209-224. Gerber, M. M. & Levine-Donnelson, D. (1989). Educating all children: Ten years later. Exceptional Children, 56, 17-29. Gerry, M. H. (1985). Policy development by state and local education agencies: The context, challenge, and rewards of policy leadership. Remedial and Special Education, 6(3), 9-17. Ginsburg, A. L., Noell, J., & Plisko, V. W. (1988). Lessons from the Wall Chart. Educational Evaluation and Policy Analysis, 10(1), 1-12. Hallahan, D. P., Keller, C. E., & Ball, D. W. (1986). A comparison of prevalence rate variability from state to state for each of the categories of special education. Remedial and Special Education, 7(2), 8-14. Heller, K. A., Holtzman, W. H., & Messick, S. (Eds.). (1982). Placing children in special education: A strategy for equity. Washington, DC: National Academy Press. Keogh, B. K. (1988). Learning disability: Diversity in search of order. In M. C. Wang, M. C. Reynolds, & H. J. Walberg (Eds.), Handbook of special education: Research and practice (Vol. 2, pp. 221-224). New York: Pergamon Press. Noel, M. M., & Fuller, B. C. (1985). The social policy construction of special education: The impact of state characteristics on identification and integration of handicapped children. Remedial and Special Education, 6(3), 27-35. Oakes, J. (1989). What educational indicators? The case of assessing school context. Educational Evaluation and Policy Analysis, 11(2), 181-199. Odden, A. (1985). Educational finance in 1985: A rising tide or steady fiscal state. Educational Evaluation and Policy, Analysis, 7(4), 395-408.
Reschly, D. J. (1988). Minority mild mental retardation overrepresentation: Legal issues, research findings, and reform trends. In M. C. Wang, M. C. Reynolds, & H. J. Walberg (Eds.), Handbook of special education: Research and practice. (Vol. 2, pp. 23-41). New York: Pergamon Press.
MARGARET J. MCLAUGHLIN (CEC #263) is an Associate Director and MARIA F. OWINGS is a Statistical Consultant at the Institute for the Study of Exceptional Children and Youth at the University of Maryland, College Park.
This research was supported in part by grant GOO8630144 awarded by the U.S. Department of Education, Office of Special Education Programs.
|Printer friendly Cite/link Email Feedback|
|Author:||McLaughlin, Margaret J.; Owings, Maria F.|
|Date:||Dec 1, 1992|
|Previous Article:||Learning disabilities and social competence: a social ecological perspective.|
|Next Article:||Are special educators' career paths special? Reports from a 13-year longitudinal study.|