U.S. Commission on Civil Rights

Employment Rehabilitation Services in Michigan

Chapter 4 

Data Analysis on the Provision of Rehabilitation Services

The Research and Training Center for Access to Rehabilitation and Economic Opportunity, Howard University, in 1995 studied the variables related to the provision of rehabilitation services.[1] The study was supported by a grant from the U.S. Department of Education, National Institute on Disability and Rehabilitation Research, and the Rehabilitation Services Administration.

The study addressed several topics pertinent to the Michigan Advisory Committee study on the provision of rehabilitation services. Specific research topics included: (1) how much of the variance in cost of service is a result of severity and type of disability, age, sex, and race, across States and within each State, and (2) what services show variation in delivery according to race/ethnicity.[2]

Following the study by Howard University, the Michigan Advisory Committee analyzed the data of rehabilitation service delivery to individuals in Michigan's tricounty area: Clinton, Eaton, and Ingham.[3] The Michigan Jobs Commission, Rehabilitation Services, has 25 service delivery areas; the Lansing office serves residents living in Clinton, Eaton, and Ingham Counties. Of interest to the Advisory Committee were (1) the delivery of services on the basis of severity of disability, and (2) the delivery of rehabilitation services on the basis of minority/ethnic status.

Howard University Study

The Howard University Research and Training Center examined the factors that influenced variations in the cost of purchased services in the delivery of rehabilitation services to racial/ethnic groups. Data analyses confirmed significant variations in services and costs within regions among racial/ethnic groups.

Two variables, time spent in VR program and the number of services, accounted for 19 percent of the variance in the cost of purchased services. Further analyses indicated a cluster of 15 other factors, including the client's race/ethnicity, which accounted for an additional 2 percent of the variance in the cost of purchased services. In general, individuals from African American, Hispanic American, and American Indian groups do not have as much money spent on their services as do persons from white or Asian groups.

The study further indicated that the individual's educational level and economic independence upon entrance into the program were the two major influences on the individual's earnings at closure. The inferences from this finding are (1) consumers from racial/ethnic minority groups, especially African Americans, American Indians, and Hispanic Americans, enter vocational rehabilitation programs with relatively less education; (2) since education is closely linked to economic status, these consumers are likely to be economically worse off; and (3) it is possible that these minority consumers may not be fully aware of their rights to request services, where such services are available, and how to access such services.[4]

If funding is proportionate among groups, then the percentage of funds spent on one particular subgroup should be the same as its percentage representation in the total group. However, the specific examination of the data shows that for the category of purchased services costing $2000 or less, African Americans were overrepresented (19.2 percent) relative to their percentage in the total population (17.4 percent). The proportion of white Americans was lower (70.5 percent) relative to their representation in the population (72.4 percent).

For the category of services costing more than $5,000, the relative percentages are reversed, with whites being disproportionately overrepresented (74.7 percent) compared with their total population (72.4 percent), and blacks (14.9 percent) being underrepresented compared with their representation (17.4 percent) in the population. A similar pattern was noted for American Indians and for Hispanic Americans, but to a much lesser extent. The representation of Asian Americans was more comparable to their population proportions.

An examination of the remaining cost of services for the category of service costing between $5,000 and $8,000 shows a similar pattern as noted for the above-$5,000 category. White clients were always represented relative to their population representation; in general the other groups were underrepresented relative to their population.[5]

A saturated model was employed to identify significant variables that account for variance in the cost of services. The dependent variable was the amount of money spent for purchased services. The list of independent variables included demographic characteristics, major and secondary disabilities, and other considered variables. The total number of variables reaching significance for the saturated model was in order of influence as follows:

1. Time in VR 

2. Number of purchased services 

3. Orthopedic, limbs

4. Blindness

5. Educational facility

6. Major disability

7. Rehabilitation facility

8. Psychotic conditions

9. Hearing impairment

10. White

11. Orthopedic, absence one upper, one lower

12. Drug abuse

13. Deafness

14. Orthopedic, lower[6]

Analysis of Rehabilitation Service Delivery in the Tricounty Area

The Michigan Advisory Committee undertook its own independent study to determine if, in Michigan, severity of disability and/or race were associated with the provision of rehabilitation services. The Advisory Committee analysis was limited to the tricounty area of Clinton, Eaton, and Ingham Counties, served by the Michigan Jobs Commission Lansing Rehabilitation Services (RS) office. This area was selected as a representative area because of its central location, relatively sizable population that is racially and ethnically diverse, and the area's diversity of rural and urban populations. Services provided by the Michigan Commission for the Blind were not analyzed as part of this study.

The population of Clinton, Eaton, and Ingham Counties is estimated in 1997 by the U.S. Census Bureau to be 371,275 (86.4 percent) white; 30,021 (7 percent) African American; 17,037 (4 percent) Hispanic; 8,254 (1.9 percent) Asian; and 2,912 (0.7 percent) American Indian.

Three sets of analyses were intended. The first analysis was to be a study of the relationship between severity of disability and successful rehabilitation. A multivariate logistic analysis was planned with successful rehabilitation the dependent variable and severity of disability, i.e., most severe, severe, or not severe; cost of services; race and ethnicity; education; age; and months of rehabilitation separately and interactively the independent variables.

Michigan Rehabilitation Services codes severity of disability under seven codes: 0 not reported, 1 severe SSDI eligible, 2 severe SSI eligible, 3 severe disability, 4 severe disability plus qualifying condition, 5 severe functional limitation, and 6 not severe. Among all clients in the Lansing RS office, 37 had code 0 and 48 had code 6.  The other 1,849 cases were designated with a severe disability code, i.e., codes 1, 2, 3, 4, or 5.

It is evident that the current format of the data does not allow for an analysis of service on the basis of severity of disability because (1) there is no clear delineation between severe and most severe in the data set, and (2) the number of not severe is too small for a valid statistical procedure.

Two other sets of analyses were undertaken to examine the impact of race and ethnicity on rehabilitation services. First, the case status, e.g., closed're habed and closed not rehabed, for the major racial and ethnic groups was analyzed to determine if the case status was independent of race and ethnicity. That is, the analysis examined whether a relationship existed between race and ethnicity and the status of a case. In concert with this analysis, the rate of successfully closed cases, i.e., employment, was analyzed for differences along racial and ethnic lines.

Second, the types of services provided to the different racial and ethnic groups were analyzed to determine if the type of service provided was independent of race and ethnicity. That is, the study examined whether a relationship existed between an individual's race and/or ethnicity and the type of rehabilitation service he or she received.

Data

Information on clients served by the Lansing RS office during 1998 was obtained by the Committee. The relevant variables obtained for each client included: (1) race/ethnicity, (2) case status, (3) primary disability, (4) secondary disability, (5) severity of disability, (6) beginning service date, (7) case cost, (8) reason case not successfully closed, (9) work status, (10) age, (11) education, and (12) types of services provided. The race and ethnicity groups are white, black, Hispanic, Asian, and American Indian. Case status refers to cases opened and closed and the reasons for case closure. 

Thirteen types of services can be provided to clients and information on these services was also obtained for each client. These services can include: diagnostic services, restoration, on-the-job training, placement, college, business/vocational training, maintenance, counseling, adjustment, job referral, miscellaneous, and other.

Racial/Ethnic Overview

In 1998 the Lansing RS office had 1,934 active clients. The racial and ethnic breakdown was 1,503 (77.7 percent) white; 333 (17.2 percent) African American; 62 (3.2 percent) Hispanic; 23 (1.2 percent) Asian; and 13 (0.7) American Indian. Minorities, as a group, are receiving rehabilitation services at a higher rate than their proportion of the population. The minority population percentage in the tricounty area is estimated to be 13.6 percent of the population; yet minorities are 22.2 percent of the Lansing RS office clientele. However, African Americans are the only racial/ethnic minority group to receive a disproportionate level of services relative to their proportion of the population. African Americans are 7 percent of the area population, and 17.2 percent of the Lansing RS office clientele (see table 5).

Ages of clients receiving rehabilitation services at the Lansing RS office are similar along racial and ethnic groups. Among white, African American, and American Indian clients, the average age is 35. The Asian clientele has the lowest average age, 26, and Hispanics have the second lowest average age, 31.

Similar to age, there is no significant difference among the racial and ethnic groups in the mean education level of the clientele receiving rehabilitation services at the Lansing RS office. American Indians have the highest average level of education, 12.5 years. Asians and African Americans average 11.3 years of schooling. The average level of education for whites is 10.8 years, and for Hispanics it is 10.6 years.

Analysis of Case Status

The case status indicates the progress of a client toward employment. There are 19 different status codes that may be assigned to a case. The Advisory Committee analysis considered 11 codes to test for independence from race and ethnicity: (1) extended evaluation, (2) closed during processing, (3) eligibility development, (4) physical and mental restoration, (5) training, (6) readiness for employment, (7) in employment, (8) services interrupted, (9) closed're habed, (10) closed not rehabed after 12 months, and (11) closed not rehabed after 10 months. For the analysis, variables (10) and (11) were combined.

Six of the seven excluded codes had no activity: referral, preservice, counseling and guidance, post-employment service, closed from preservice, and closed due to no application. The other code excluded from the analysis was applicant. The number of clients by race and ethnicity for the 10 included status variables are in table 7.

A chi-square test procedure was employed to analyze the frequency of occurrence of observations in the observed sample of services provided and the expected frequencies of such services obtained from an hypothesized distribution. Comparing observed frequencies with corresponding expected frequencies demonstrates whether differences in case status are associated with race and ethnicity. The employed procedure, a goodness-of-fit test, is based on the quantity:

 C = i=1  {(oi - ei) / ei }     (eq. 4.1)

where C is a value of the random variable X whose sampling distribution is approximated very closely by the chi-square distribution. The symbols o and e represent the observed and expected frequencies for the ith cell.[7]

Computing equation 4.1, C = 39.2, while X (a = 0.05) with 36 degrees of freedom equals 43.7. Since C (39.2) < X (43.7), the null hypothesis is rejected at the 0.05 level of significance, and race and/or ethnicity and the status of a case are not related. In other words, the race and ethnicity of clients are independent of the particular status of a case, everything else held constant.

A subsequent analysis was undertaken to examine specifically whether the proportion of minorities in the rehabilitation services program obtaining employment was significantly different from the proportion of nonminorities in the program obtaining employment. Examining table 7, 28.3 percent of whites have obtained employment, i.e., status of in employment or closed're habed, while only 21.8 percent of minorities have obtained employment.

To control for the outlier effect of extraneous case status and condensed timeframes, only four case status variables were considered: (variable 7) in employment; (variable 8) services interrupted; (variable 9) closed're habed; and (vari ables 10 and 11) closed not rehabed. By racial/ethnic group, these case status variables affected 679 whites, 140 African Americans, 29 Hispanics, 6 Asians, and 3 American Indians.

The first two variables, closed're habed and in employment, were combined into a successful rehab variable, while the two variables closed not rehabed and services interrupted were combined into one unsuccessful rehab variable. From table 7, 392 whites had a successful rehab, 66 African Americans had a successful rehab, 15 Hispanics had a successful rehab, 5 Asians had a successful rehab, and no American Indians had a successful rehab.

Combining the four minority groups, in the Lansing RS office white clients have a success ful rehab rate of 57.7 percent. Minorities have a lower successful rehab rate of 48.3 percent. Using a point estimation technique of the observed proportion in a binomial distribution, a confidence interval can be established for p to determine if the observed successful rehab rates for minorities is significantly different than expected. The point estimator of the observed proportion, p, is determined by:

p - za/2 {(p q)/n} < p < p + za/2 {(p q)/n}      (eq. 4.2)

Computing equation 4.2, the confidence interval for p is:

52.7% < p < 55.8%     (eq. 4.2a)

where a = 0.05. This means, other things constant, with 95 percent probability the observed percentage of minorities with a successful rehab should be between 52.7 percent and 55.8 percent. However, the observed proportion of minorities is 48.3 percent, a rate significantly lower than what would be expected absent unusual circumstances.

A possible explanatory variable is that the unemployment rate for minorities in the Lansing RS area is significantly higher than the unemployment rate of whites. If minorities, independently of rehabilitation services, are less likely to be employed than nonminorities, then the results observed in equation 4.2a are not unexpected.

Analysis of Types of Services Provided

The number of types of services provided to clients was evaluated for independence from race and ethnicity. The 13 types of services provided: (1) diagnostic services, (2) restoration, (3) on-the-job training, (4) placement, (5) financial aid for college, (6) business/vocational training, (7) transportation, (8) maintenance, (9) counseling, (10) adjustment, (11) job referral, (12) miscellaneous, and (13) other, by race and ethnicity are shown in table 8.

A chi-square test procedure was employed to analyze the frequency of occurrence of observations in the observed sample of services provided and the expected frequencies of such services obtained from an hypothesized distribution. By comparing the observed frequencies with the corresponding expected frequencies, it can be determined whether differences in services among racial and ethnic groups are a result of sampling chance or the result of a nonuniform distribution.

The procedure, a goodness-of-fit test, is based on the quantity:

C = i=1  {(oi - ei) / ei }     (eq. 4.3)

where C is a value of the random variable X whose sampling distribution is approximated very closely by the chi-square distribution. The symbols o and e represent the observed and expected frequencies for the ith cell.[8]

Computing equation 4.3, C = 47.4, while X (a = 0.05) with 48 degrees of freedom equals 43.7. Since C (47.4) > X (43.7), the null hypothesis is rejected at the .05 level of significance. Race and ethnicity and the types of services provided are not independent. That is, the above establishes a statistically significant relationship between an individual's race and/or ethnicity and the receipt of specific rehabilitation services.

Equity of Resource Distribution

For several years appropriations of general funds by the Michigan State Legislature have been insufficient to capture all Federal dollars available for vocational rehabilitation. The Rehabilitation Services Administration requires match by the State or local public entities at a ratio of 21.3 percent State or local to 78.7 percent Federal dollars.[9]

MRS attempted to make up the shortfall in general funds by approaching public community partners. Most of the partners with available funds were community mental health agencies or local school districts. These entities and others have contributed a total of more than $5 million that supports a total of about $18.3 million. These funds represent about 20 percent of the total budget.[10]

The remaining 80 percent of the MRS budget goes to support the Michigan Career and Technical Institute ($7.7 million); operations, including staff, travel, supplies, rent, etc. ($40.3 million); title I grants ($6.3 million), non-title I grants ($3.6 million), and case service dollars for the general disabled population. Most of this 80 percent is distributed equitably based on the 1990 census data for the disabled working-age population. Grants are awarded on a variety of bases, some competitive, some by population formula.[11]

The 20 percent, or $18.3 million, is what MRS calls the cash match portion of its case service budget. This is where the equity problem presents itself in that opportunities for cash match contributions vary from area to area around the State. Historically, communities on the west side of Michigan have been more able or willing to put forth local match dollars than those on the eastern side especially the southeast Michigan area.[12]

MRS senior managers recognized this as a growing problem several years ago and instituted a system to mitigate this trend and achieve greater equity over time. The process of managing this 20 percent of the budget toward equity is as follows:

1. A determination is made before the fiscal year as to how much local match is necessary and how much Federal funding that match would capture.

2. Population-based targets are established for each district office. MRS district offices cover specific counties. U.S. census population data for disabled working-age persons are used. Managers are instructed to manage their local cash match agreements toward their population-based portion of the potential cash match case service funds.

3. This equity target is compared with the actual and potential value of cash match agreements. The ratio of actual and potential agreement value to the population-based proportion of the cash match total is computed. Equity is the percentage of the population-based figure represented by the actual and potential value. For example, Oakland County's figure is 99 percent. This means that its actual and potential agreements are 99 percent of what they would be entitled to if the total cash match was distributed entirely on a disability population-based basis.

4. Managers are instructed they should work toward having their equity ratios between 80 and 120 percent. One hundred percent would represent perfect population-based equity. Offices over 120 percent are not allowed to commit to any agreements over the 120 mark. Offices less than 80 percent are given a period of time to garner more resources in their current agreements or establish new agreements to bring their figure up. A deadline is established.

5. Once the deadline has been reached, the total value of agreements is completed. If it falls short of capturing all Federal dollars available, those offices at more than 120 percent are given the green light to pursue more resources. The overall objective is to have a movement toward 100 percent, but there is acknowledgment that a gradual process is needed to avoid disruption to services in the communities involved.

6. An additional measure MRS undertook to achieve greater equity was to provide the offices with the lowest amount of cash match (those in Wayne and Macomb Counties) with supplemental funds to help balance their needs. For example, in 1998, an extra amount of money (approximately $1.1 million) was given to Wayne and Macomb counties. The plan was for this supplemental amount to be reduced on a regularly scheduled basis as Wayne and Macomb garnered more cash match. (The equity figures listed for those two counties do not include that supplement. Inclusion would bring them closer to equity.)[13]

Table 9 shows the change in equity figures for each of the MRS district offices over the years 1997, 1998, and 1999. MRS states that its goal was to see movement toward the 100 percent number. An examination of table 9 shows there has essentially been no change in Flint, Port Huron has come closer to equity, Oakland has increased and is now close to perfect equity, and Macomb has increased. Overall, progress has been made in the majority of communities. A few offices have moved away from equity, but they are the exception. MRS has measures being instituted to change that trend.[14]