copyright 1992 by The Association for Persons with Severe Handicaps

JASH 1992, Vol. 17, No.2, 87-94

The Wage Effects of Supported Employment Lyke Thompson, Greg Powers, and Berenice Houchard Western Michigan University Noble and Conley (1987) compared a series ofprograms in various states with regard to costs and earnings. They found that earnings are clearly higher for employment in integrated, regular job settings. Kiernan, McGaughey, and Schalock (1986) have presented perhaps the most extensive data on increases in wages resulting from employment. Data on more than 85,000 persons in 1,119 adult service programs were supplied as part of the National Employment Survey for Adults with Developmental Disabilities. They found average quarterly wages for workers in sheltered workshops to be $402.75, compared with $786.01 for persons in SE. The average per-hour wage rate for persons in workshops was $1.31, compared with $2.59 for persons in SE. Wehman, Kregel, Banks, Hill, and Moon (1987) and Lagomarcino (1988) found that workers in SE made more than those in sheltered employment. Recently, Kregel, Wehman, and Banks (1989) found significant and substantial wage increases for persons at each level of retardation after they entered supported employment. They found that individual placements produced higher wages compared with other models. Our Michigan data (Thompson, 1990) showed that supported workers averaged more hours and higher wages per hour than their counterparts in sheltered settings. Such simple comparisons did not satisfy policy makers, facility staff, and other practitioners, however. Possibly, the results could be an artifact of SE projects picking higher functioning workers, or of other characteristics ofworkers, such as their gender or their living arrangements. These alternative explanations are worthy of consideration. Several recently published studies do not report efforts to control for levels of disabilities (Ellis, Rusch, Tu, & McCaughrin, 1990; Kregel, Wehman, Revell, & Hill, 1990) or do not include comparison groups. A further question of interest is whether wages differ on the basis of service delivery model. Three models that enjoy widespread use are individual placement, enclaves, and mobile crews. To date, the effects of these models on wages is relatively unknown. The current study examines alternative explanations by developing a more rigorous test of the wage effects of SE versus traditional rehabilitation options. Three steps were taken to ensure a more complete and rigorous analysis. First, the analysis uses regression analysis

This study investigated the effects of supported employment on wages earned by persons with disabilities. We sought to determine whether supported employment produced higher wages than sheltered employment, whether any effect was a result ofincreases in wage rates or hours, and whether one model of supported employment had more impact than another. The analysis showed that wages increased after entering supported employment, after correctingfor differences in employees' measured IQs. Entry into supported employment had a more consistent impact on wage rates than on hours of employment. Individual placement was shown to have the largest effect on wages, while participation in mobile crews had little effect. DESCRIPTORS: adults, developmental disabilities, modeling, public policy, research, supported employment Supported employment (SE) provides communitybased jobs for people with severe disabilities. Many outcomes of supported work were anticipated or promised, ranging from increased independence and integration to decreased need for financial support, but the key goal was to increase wages. For workers, increasing wages is not only an important goal in and of itself, but is also viewed as a means to greater independence and control over their lives. Evidence of the efficacy of community-based employment for persons with severe disabilities was first presented in 1982 (Brickey, Browning, & Campbell, 1982; Kraus & MacEachron, 1982; Revell, Arnold, Taylor, & Saitz-Blotner, 1982). Wehman (1986) was able to show that workers could maintain jobs for long periods, earn increased wages, and fade their use of extensive support fairly rapidly. Vogelsberg, Ashe, and Williams (1985) demonstrated similar results for several pilot programs in Vermont. Results in Washington State were also positive (Moss, Dineen, & Ford, 1986). This research was supported in part by the Michigan Developmental Disabilities Council and Michigan Rehabilitation Services. We gratefully acknowledge the work of former staff who aided in the collection of the data used here. We appreciate the comments of the anonymous reviewers who provided feedbacks on previous drafts of this article. 87

88

Thompson, Powers, and Houchard

to control for variables that might account for the wage effects of SE. Second, the use of a comparison sample permits the multivariate analysis to be structured to make direct contrasts of wages, wage rates, and hours between SE and traditional employment options (e.g., sheltered work). Third, using a statewide sample based on many projects broadens the available data base on the effects of SE. The current study was designed to answer several questions. First, does SE have a wage effect after accounting for the effects of other factors? Second, if it does have an effect, how is that effect producedthrough changes in number of hours worked or through wage rates or both? Third, are all SE models equal, or do some approaches have more impact on wages?

Table I Number of Cases within Each Analysis First quarter First Second Third Fourth Fifth Sixth Seventh Eighth Ninth Tenth Eleventh

Last quarter Fourth Fifth Sixth Seventh Eighth Ninth Tenth Eleventh Twelfth Thirteenth Fourteenth

Entered SE 26 58 71 47 49 109 51 25 40 39 32

Not in SE 913 1,069 1,103 913 947 921 726 726 716 511 531

Total 939 1,127 1,174 960 996 1,030 777 751 756 550 563

Setting Supported employment in Michigan was structured as an interagency initiative driven by the joint effects of Michigan Rehabilitation Services (MRS), the Department of Mental Health (DMH), the Special Education Services (SES) division of the Michigan Department of Education, and the Michigan Developmental Disabilities Council (DDC). The Michigan initiative began with a few projects and grew over 4 years to include nearly 42 projects. These projects ranged from ones that included just a few individuals to some that included more than 60 people. Some projects grew rapidly, diversifying and producing substantial outcomes; others started small and stayed that way. Supported employment included work in regular jobs in the community in groups of eight or fewer persons. The federal definition of SE calls for all employees to work at least 20 hr per week, but this definition was adopted at the federal level considerably after the initiation of the Michigan project. Therefore, many participants worked less than 20 hr per week.

demographics, activities, and work of a comparison sample drawn randomly from those people with disabilities remaining in the day and work activity settings. Table 1 presents the number of people entering SE and the number of the comparison group for the 11 analyses. The first quarter refers to the period before entry into SE, and the last quarter refers to a time 9 months later. The average number of workers entering was 49.70. The number of people entering SE for the first time was relatively small for any given quarter. The number entering was also small relative to the size of the comparison group. One major advantage of using the 11 analyses is that this replication helps offset the effects of having relatively small groups entering SE in anyone quarter. That is, it increases the reliability of the results. Table 2 presents the characteristics of the supported workers and of people in the comparison group for three representative periods. This table reveals some consistent patterns. First, males represented a higher proportion of the samples, both among those entering SE and in the comparison group. Second, a higher proportion of people entering SE were ages 25 to 34 years, compared with those not in SE, who were most represented in the 35-44 years category. Third, when asked about the disabilities of participants, respondents indicated that mental retardation was the most frequently occurring category. Note that respondents could mention more than one disability for a sample member. I Fourth, the intelligence quotient (IQ) scores of persons entering SE tended to be somewhat higher than the comparison group. For example, as shown in Table 2, the percentage of persons entering supported employment in quarters 6 through 9 that had IQs in the 55-69 range was 35.6, compared with 27% of the comparison group. Some people entering SE had IQ

Participants Each quarter (3 months) a new group of people with disabilities entered SE. Our project kept track of the characteristics of these and all other supported employees of participating projects. In addition, we tracked the

1 Respondents were not asked to designate a primary disability because, in the review procedure to determine what items would be included in our forms, practitioners argued that it is often difficult or inappropriate to designate one disability as primary.

Method This study involved 11 parallel quasi experiments using statistics to control for a series of factors that might affect the impact of SE on wages, hours, and wage rates. In each of the 11 analyses, the wages, hours, and wage rates ofpeople who entered SE were compared both with their own levels prior to entry and with those of a comparison sample. For example, one analysis estimated the effects of SE on wages three quarters (9 months) after entry into the job. Comparisons were made both to the supported workers' wages prior to entry and to the wages of sample members who remained in sheltered work settings.

I

32.0 8.0

8 2

4.0

56.0

4.0

I

14

16.0 80.0

4 20

8.3 29.2 58.3 4.2

3.8 3.8

I I

2 7 14 I

92.3 11.5 3.8 3.8

4.2 37.5 45.8 12.5

64.0 36.0

24 3 I I

I 9 II 3

16 9

%

33 2

313 14 469 23 3 46

86 800 12 10

3.7 0.2

34.7 1.6 51.9 2.5 0.3 5.1

9.5 88.1 1.3 1.1

8.0 23.3 31.5 25.7 6.7 4.8

4.1 3.1 2.3

37 28 21 69 202 273 223 58 42

88.2 8.7 1.1 3.6 6.9 6.7 0.8 3.6 8.7

0.4 1.8 27.6 35.0 18.1 10.4 6.7

56.7 43.3

805 79 10 33 63 61 7 33 79

5 16 247 313 162 93 60

512 391

Total

%

34 2

327 14 477 25 3 46

90 820 12 II

1.9 4.6

2 5

25.0 0.9 59.3 8.3

0.9

6.5 92.6

15.4 35.6 35.6 7.7 5.8

1.8 1.8 8.3

88.1 3.7 2.8 2.8 0.9 11.0 11.0 1.8 2.8

2.8 39.8 35.2 11.1 8.3 2.8

57.4 42.6

%

27 I 64 9

I

7 100

16 37 37 8 6

2 2 9

37 28 21 69 204 280 237 59 42

96 4 3 3 I 12 12 2 3

3 43 38 12 9 3

62 46

T\ltal

829 82 11 34 63 61 7 34 80

5 17 256 324 165 93 60

528 400

Total

29 2

288 21 501 20 I 47

105 782 13 II

84 195 261 235 56 38

22 23 39

807 77 5 5 9 70 39 34 64

15 289 335 151 75 44

504 401

Total

3.2 0.2

31.7 2.3 55.1 2.2 0.1 5.2

11.5 85.8 1.4 1.2

9.7 22.4 30.0 27.0 6.4 4.4

2.4 2.5 4.2

87.6 8.4 0.5 0.5 1.0 7.6 4.2 3.7 6.9

1.7 31.8 36.9 16.6 8.3 4.8

55.7 44.3

%

Not in SE

Entered SE

Not in SE

a Missing values not calculated into percentages. • Note: Numbers reflect persons who may have multiple disabilities.

Alternative immediate services Other Unknown

Primary living arrangement Family home Other relative's home Adult foster care/home Semi-independent living State hospital or facility

Race or ethnic origin Black Caucasian Hispanic Other

IQ score 0-19 20-39 40-54 55-69 70-80 80+

Mental retardation Seizure disorder Learning disability Autism Brain injury Speech impairment Chronic mental illness Cerebral palsy Other physical & orthopedic impairment Legally blind Auditory impairment Other

Disabilities"

Age 16-20 21-24 25-34 35-44 45-54 55-64 65 and over

Gender Male Female

Total

Entered SE

Quarter 6-quarter 9

Quarter l-quarter 4

Table 2 Consumer Characteristics"

34 2

315 22 565 29 I 49

112 882 13 12

84 211 298 272 64 44

24 25 48

903 81 8 8 10 82 51 36 67

0 18 332 373 163 84 47

566 447

Total

3.1 3.1 12.5

I I 4

2 I

16 5

8

6 26

6.3 3.1

50.0 15.6

25.0

18.8 81.3

6.7 20.0 43.3 16.7 13.3

3.1 12.5 15.6 3.1 9.4

I 4 5 I 3

2 6 13 5 4

68.8 6.3 15.6

3.1 3.1 46.9 25.0 15.6 6.3

59.4 40.6

%

16

II

176 12 304 II

58 462 8 2

24 86 176 169 44 22

8 15 34

8 49 26 24 30

477 44 7

I 8 176 186 92 50 18

276 250

Total

3.0

2.1

33.2 2.3 57.4 2.1

10.9 87.2 1.5 0.4

4.6 16.5 33.8 32.4 8.4 4.2

1.5 2.8 6.4

1.5 9.2 4.9 4.5 5.6

1.3

89.8 8.3

0.2 1.5 33.1 35.0 17.3 9.4 3.4

52.5 47.5

%

Not in SE

Quarter l l-quarter 14

22 2 5

I I 15 8 5 2

19 13

Total

Entered SE

18 I

320 16 0 II

12

184

64 488 8 2

24 88 182 182 49 26

9 16 38

499 46 12 0 9 53 31 25 33

2 9 191 194 97 52 18

295 263

Total

00 \0

a "a

~

"2-

a

"0-m

~

"8

"0

'"

Vl

0

a ..,

iii

.,~ """m

90

Thompson, Powers, and Houchard

scores higher than 70 because Michigan included in SE some workers who had physical disabilities or mental illness. Fifth, the racial characteristics of the samples were fairly similar. Sixth, a higher proportion of people entering SE tended to live in semi-independent settings. Experimental Design The quasi experimental design for this study relies upon an extended version of a nonequivalent control group design (Cook & Campbell, 1979), reinforced by statistical control. The nonequivalence of the control group resulted from the evaluators having no way to control which workers the study sites chose for inclusion in SE. Sites may have selected candidates for SE according to some implicit criterion that increased (or decreased) the programs' chances of success. To correct for this statistically, the evaluators collected data on level ofretardation and other factors that may be related to chance of success. It is generally not possible to know or measure all variables that might influence outcomes. However, as Cook and Campbell suggested, use of such "statistical control" represents a significant improvement on before-and-after measurement. The following symbolic diagram indicates the general plan of the design: In SE 0 X 000 Not in SE 0 000 In this representation, Os represent observations and the X indicates the intervention-entering SE. Each column of Os represents a new set of observations at another point in time. The top row shows the pattern of observation, intervention, and repeated follow-up observation for the workers entering SE. The bottom row represents the pattern of observations for the comparison group. Sampling Procedures The data for this analysis were provided by the evaluation of SE in Michigan. This evaluation collected basic data such as demographics, history, disabilities, benefits, job characteristics, wages, and movement in and out of various employment programs. The evaluation initiated collection of basic data in the last quarter of 1986. Data collection ended in the first quarter of 1990, resulting in a total of 14 quarters' worth of data. This data collection relied on agencies to submit data in two ways: completion of a packet of forms, or the use of the computerized Supported Employment Data Entry System (SEDES). Other than agency data, all data were collected on specific individuals. Most participating rehabilitation facilities submitted data on all their supported employees, and a random sample of persons in traditional programs. (This sample was drawn using procedures provided by the evaluation.) Some facilities provided complete data on all persons working through the facility, including those in

traditional and SE programs. Some facilities offering SE outside of the state-sponsored initiative chose not to provide data for the statewide evaluation.' Statistical Analysis Table 3 clearly shows the complexity faced by the evaluators of a field study of SE. The table shows that the wages of supported workers and their counterparts in the comparison group were not equal prior to entry. Sites did appear to choose their more productive workers for entry into SE; thus, subjects who became supported workers had higher average wages than their counterparts, even before entering. Because of this, simple before-and-after comparisons of wages will not accurately display the effects of SE. Second, the table shows that workers entering SE did increase their total quarterly wages after entering. These points suggest that straightforward averages will not accurately display the effects of SE. These findings clearly point to the necessity for controlling statistically the effect of initial differences between the groups in analyzing wage effects resulting from SE. This study used a series of II identical regression analyses to estimate the effect of entering supported employment upon wages, hours, and wage rates. The wage regression equation used is:

W= a

+ WP + G + R + ND + IQ + LA + S

where W =

a = WP =

G = R =

ND = IQ LA

total wages earned in the fourth quarter of a four-quarter period. This would be the third quarter after entry for those sample members entering SE. the constant in the regression equation. total wages in the first quarter of the fourquarter period. This would be the quarter before entry for workers entering SE. the gender of the sample member. the racial and ethnic background ofthe sample member, measured as minority versus nonminority. the number of disabilities as a proxy for severity. the IQ category of the sample member. a series of three dummy variables representing

2 Michigan did not require all facilities to report data. Therefore, at least four major facilities and a number of smaller ones chose not to provide data because of the considerable costs of completing forms. The exact characteristics of the nonparticipating agencies are not known, but our impression is that the agencies that did not participate were more involved in enclaves and work crews as compared with those agencies that reported data. The agencies that did participate included large, medium, and small facilities in every geographical region of the state.

91

Wage Effects of Supported Employment

Table 3 Wage Effect on Entering SE Before entering

After entering

Beginning quarter

Entered

Did not enter

Entered

Did not enter

1 2 3 4 5 6 7 8 9 10 11

$217.69 $313.60 $302.50 $339.77 $232.71 $276.81 $369.50 $320.54 $498.35 $659.90 $455.06

$158.49 $176.70 $210.79 $152.83 $208.91 $183.26 $231.61 $283.86 $211.30 $266.63 $234.24

$435.93 $593.39 $324.43 $490.65 $654.99 $494.88 $617.85 $593.09 $771.45 $894.18 $1,206.72

$161.41 $210.80 $220.87 $206.59 $257.36 $215.92 $248.22 $215.61 $249.23 $239.93 $250.86

living arrangements.' The three dummies were for family, semi-independent, and other living arrangements, respectively, with the base case being a group home case. S = whether the person shifted into SE or not. By including wages prior to entry into SE, the shift variable will only capture the effect of entering SE. Variables for gender, racial and ethnic background, and living arrangements were included because our previous bivariate analyses had indicated these might have an effect on wages. Number of disabilities was included based on the logic that a person who is legally blind may face more barriers than a person of the same intelligence but without the challenges imposed by blindness. This proxy for severity has been related to wages in bivariate analyses and has been used in other settings (Schalock, 1986). Obviously, this is an inexact measure given that some individuals with one severe disability may face more barriers than individuals with multiple less challenging disabilities. This variable and the variable for IQ were included to control for the alternative explanation that wage effects of SE are artifacts of a higher functioning population in SE. This equation was adapted to test the research question regarding models. Instead of a single variable indicating a shift into supported employment, three (or, in later quarters, four) dummy or indicator variables were used in order to measure the relative effect of the various models. It was also adapted to measure the effects on hours and wage rates of entering SE. In these cases, hours and wage rates were simply substituted in the appropriate places in the equation. s In other words, there were four cases being measured. The inclusion of these dummy variables sought to determine whether living in a family home, a semi-independent setting, or some other setting (including some varieties of independent of independent living) had a significant effect on wages as compared with the base case of living in a group home or adult foster care.

Results Wages

The results shown in Table 4 indicate that entering SE leads to increases in wages. In 10 of the 11 quarters, after adjusting for the effects of IQ, number of disabilities, and other factors, entry into SE significantly increased wages. On average, SE increased wagesby about $299.28 per quarter, or $1,197.12 per year, for workers with disabilities. This is likely to be a conservative estimate, given that the magnitude of difference appears to be increasing across time." The results indicate that in 7 of the 11 quarters IQ affected wage levels. Wages tended to increase with IQ. Given the distribution of IQ in the persons in SE as compared with those in traditional programs, failure to correct for IQ would have skewed the effects positively toward SE. The number of disabilities a person had was related to wageslevels in three quarters. Living arrangements were related to wages in five quarters; semiindependent settings, family homes, and other (independent living and other settings) were related to higher wages. Interview data indicate this was because group home providers tend to limit hours of work and, thus, total wages. They reported doing this because unusual working hours interrupt their normal routines. Race and gender were related to wagesonly occasionally (one quarter and two quarters, respectively). Hours It is important to understand how SE increased wages. This could have occurred because of either increases in hours or increases in wage rates or both. Table 5 shows the results of a series of regression 4 Because of space consideration, detailed analyses on which these numbers are based are not included here. For further information, please contact either Lyke Thompson or Berenice Houchard, Western Michigan University, Policy Studies Project, 159 Wood Hall, Kalamazoo, MI 49008; telephone number 616-387-4138.

92

Thompson, Powers, and Houchard Table 4 Effect of entering SE on Total Wages Last quarter

First quarter First Second Third Fourth Fifth Sixth Seventh Eighth Ninth Tenth Eleventh Averages for p < .05

Fourth Fifth Sixth Seventh Eighth Ninth Tenth Eleventh Twelfth Thirteenth Fourteenth

Table 5 Effect of SE on Hours after Entry First quarter First Second Third Fourth Fifth Sixth Seventh Eighth Ninth Tenth Eleventh Average for p < .05

Last quarter Fourth Fifth Sixth Seventh Eighth Ninth Tenth Eleventh Twelfth Thirteenth Fourteenth

Additional wages after entering SE

$160.33 $182.66 $216.33 $160.73 $210.22 $194.26 $241.06 $228.79 $226.28 $288.35 $243.57 $213.87

$215.42 $242.13 $37.17 $122.26 $336.72 $172.68 $235.71 $324.58 $280.03 $331.71 $731.53 $299.28

Significance 0.0000 0.0001 Not sig. 0.0049 0.0000 0.0000 0.0002 0.0000 0.0000 0.0000 0.0000

Table 6 Effect of Entry into SE on Wage Rates

Additional hours after Significance entering SE 6.64 7.27 14.53 41.26 60.85 58.95 37.84 63.37 31.67 39.89 88.34 58.77

Average in-house wages

Not sig. Not sig. Not sig. 0.0010 0.0000 0.0000 Not sig. 0.0002 Not sig. 0.0000 0.0000

analyses estimating the effect of SE on hours on the job. These results are less definitive. Supported employment had a significant effect on number of hours worked in 6 of the 11 analyses. For these quarters, however, the effect was an average increase of 58.77 hr per quarter, or approximately 4.52 hr per week. Note that these results were adjusted for the effects of IQ, number of disabilities, and other factors. Wage Rates If hours are an inconsistent and limited source of the wage effects, the major impact should be produced through increases in the wage rate. Table 6 shows the results of regression analyses of the effect of SE on wage rates. Again, results are adjusted for the effects of IQ, number of disabilities, and other factors. Wage rates were consistently affected by entering SE. In 9 of the 11 analyses, the effect of entering upon wage rates was significant. The average effect of moving into SE on wage rate was an increase of$1.85 per hour. Wage effects, then, resulted from a combination of changes in hours and wage rates, although wage rate

First quarter

First Second Third Fourth Fifth Sixth Seventh Eighth Ninth Tenth Eleventh Average for p < .05

Last quarter Fourth Fifth Sixth Seventh Eighth Ninth Tenth Eleventh Twelfth Thirteenth Fourteenth

Additional wage rates after Significance entering SE $0.88 $1.28 $0.15 $4.37 $3.19 $0.63 $1.18 ' $1.58 $1.43 $1.13 $1.82 $1.85

Not sig. 0.017 Not sig. 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

effects appear to the more consistent source. In three quarters, there was no significant effect on hours from entering SE, but there was a consistent effect upon wage rates. In six quarters, entry had a significant effect upon both hours and wage rates. Models Throughout the evaluation of SE in Michigan and nationally, there have been indications that wages varied by the type of model workers entered. Table 7 presents the results of an analysis structured as were the previous analyses, except that the intention is to measure the quarterly wage effects of each of the major models. Again, the regression analysis was adjusted for the effects of IQ, number of disabilities, and other factors. After correcting for the effects of other variables, persons in individual placements had substantial wage advantages over those in in-house settings. Persons in enclaves were shown to have had lesser wage advantages. People in individual placements, on average, earned $535.03 more per quarter than those in in-house work settings; persons in enclaves earned $243.83 more.

93

Wage Effects of Supported Employment

Table 7 Wage Effects of Entering SE by Model

*p =

First quarter

Last quarter

Individual placement

Enclave

Mobile crew

First Second Third Fourth Fifth Sixth Seventh Eighth Ninth Tenth Eleventh Averages

Fourth Fifth Sixth Seventh Eighth Ninth Tenth Eleventh Twelfth Thirteenth Fourteenth

$394.43* $588.82* $155.43 $469.97* $494.13* $369.60* $435.74* $583.77* $541.93* $621.80* $850.15* $535.03

$202.43* $140.79 $69.09 $174.56* $228.41* $137.14* $49.10 $143.81 $9.00 $140.24 $476.61* $243.83

$73.01 $134.16 $78.95 $112.23 $147.24 $192.47* $85.91 $186.59 -$193.98 -$5.08 Not sig. $192.47

.05 or better for the estimated coefficient.

The results for enclaves are based upon only five quarters in which this model produced significant differences. Mobile crews produced statistically significantly higher wages in only one quarter. 5

Summary and Conclusions The results of this study indicate that SE is likely to produce greater wages than in-house employment. More specifically, individual placements and enclaves appear to increase wages. In general, overall wages of supported workers are likely to be more than double those of workers in work activity and workshop programs after correcting for the effects of IQ, number of disabilities, and other factors. Some of this wage effect is produced by an increase in hours, but more consistent and larger increases result from the hourly rates earned by workers in the community as compared with their counterparts in segregated settings. The wage increases are most consistent and largest for workers in individual placements. There is an inconsistent but moderate increase in wages when workers enter enclaves. The effect on wages of participation in mobile crews was significant in only one analysis. There are limitations to this analysis. IQ is an insufficient indicator of functional ability. Other factors, such as social skill, may affect who gets chosen for SE and may account for a portion of the differential in wages found here. Further research must address this question. An additional limitation is that the sample included many, but not all, SE projects in Michigan. The results must be interpreted as representing a sample 5 There are two reasons that enclaves and mobile crews do not have as significant an impact. First, a smaller and smaller number of workers entered these models as it became apparent to sites that individual placements produced higher wages. This means that the statistical error is greater because of smaller sample sizes. Second, the effect of these models on wages does appear to be smaller, and for technical reasons it is more difficult for a small effect to be statistically significant.

of cases, albeit a large and geographically diverse sample. Measures of abilities, such as IQ levels, were found to be related to outcomes. The effect of IQ on wages was significant. In a majority of the analyses, workers with higher IQs made more income. As one moves up the IQ categories from one level to another, wages increase by $33.58 per quarter." Thus, an individual with an IQ in the range considered to represent profound retardation would average $412.39 per quarter, whereas a person with an IQ representing a mild level of retardation would make $546.73, a wage 32% higher. These findings have implications for selecting models of supported employment when the goal is to increase wages.Mobile crews may not be a viable way ofincreasing the wages of workers with disabilities, but this result may be a product of how this model is implemented in Michigan. Our impression from site visits is that enclaves and crews are more likely to pay systematically lower wages than are individual placements. Also, we are aware that some projects assign workers with more severe disabilities into mobile crewsand enclaveslargely because it is more efficient to coach them in a group context. Overall, these findings imply that individuals with severe retardation are likely to dramatically increase their income by entering SE, especiallyindividual placements. In Michigan, it is still the case that the vast majority of people with developmental disabilities who might be qualified for SE instead participate in work activity and day activity program. Our findings suggest many of these people could dramatically increase their incomes if they had the choice of entering SE, and if the funding of "follow-along" support can be expanded. Federal Medicaid policies that discourage SE and support the maintenance of large-scale facilities effectively 6 This number represents an average of the statistically significant IQ coefficients in the wage model. One could choose to use any of the quarters, but the average contains the most information.

94

Thompson, Powers, and Houchard

increase the dependency and lower the incomes of people with disabilities. Changes in policies to allow Medicaid to routinely fund "follow-along" support for workers are recommended. Finally, this research and a growing body of literature strongly indicates that SE, and especially individual placement, is a path to higher income for persons with severe disabilities. Our results indicate that it is SE, and not other factors, that produce this effect.

References Brickey, M., Browning, L., & Campbell, K. (1982). Vocational histories of sheltered workshop employees placed in projects with industrial and competitive jobs. Mental Retardation,

20,52-57. Cook, T. D., & Campbell, D. T. (1979). Quasi-experimentation: Design and analysis issuesfor field settings. Chicago: Rand McNally. Ellis, W. K., Rusch, F. R., Tu, J., & McCaughrin, W. (1990). Supported employment in Illinois. In F. R. Rusch (Ed.), Supported employment: Models, methods, and issues. Sycamore, IL: Sycamore Publishing Company. Kiernan, W. E., McGaughey, M. J., & Schalock, R. C. (1986).

National employment survey for adults with developmental disabilities. Boston: Children's Hospital, Developmental Evaluation Clinic. . Kraus, M., & MacEachron, A. (1982). Competitive employment training for mentally retarded adults: The supported work model. American Journal 0/ Mental Deficiency, 86, 650-653. Kregel, J., Wehman, P., and Banks, P. D. (1989). The effects of consumer characteristics and type of employment model on individual outcomes in supported employment. Journal 0/AppliedBehavior Analysis, 22, 407-415. Kregel, J., Wehman, P., Revell, W. G., & Hill, M. (1990). Supported employment in Virginia. In F. R. Rusch (Ed.),

Supported employment: Models, methods, and issues. Sycamore, IL: Sycamore Publishing Company. Lagomarcino, T. (1988). Self management: Facilitating employeeindependence in supported employment settings(Vol. 4). Champaign, IL: University of Illinois. Moss, J. W., Dineen, J. P., & Ford, L. H. (1986). University of Washington employment training program. In F. R. Rusch (Ed.), Competitive employmentissuesand strategies. Baltimore: Paul H. Brookes. Noble, J., & Conley, R. (1987). Accumulating evidence on the benefits and costs ofsupported and transitional employment for persons with severe disabilities. Journalofthe Association ofPersons withSevereHandicaps, 12(3), 163-174. Revell, G., Arnold, S., Taylor, B., & Saitz-Blotner, S. (1982). Project transition: Competitive employment services for the severely mentally retarded. Journal0/Rehabilitation, 48( I), 31-35. Schalock, R. (1986, May). Current approaches to assessing a person's qualityof life. Paper presented at the I 10th Annual Meeting of the American Association of Mental Deficiency, Denver, CO. Thompson, L. (1990, May). Supportedemploymentin Michigan:Fourteenth quarterly statisticalsummary. Kalamazoo: Western Michigan University, Supported Employment Evaluation Project. Vogelsberg, R. T., Ashe, W., & Williams, W. (1985). Community based service delivery in a rural state: Issues for development and implementation. In R. Horner, L. M. Voeltz, & B. Fredericks (Eds.), Education/or learners with severe handicaps: Exemplary service strategies. Baltimore: Paul H. Brookes. Wehman, P. (1986). Competitive employment in Virginia. In F. Rusch (Ed.), Competitive employment issues and strategies. Baltimore: Paul H. Brookes. Wehman, P., Kregel, J., Banks, P. D., Hill, M., & Moon, M. S. (1987). Sheltered versus supported work programs: A second look. Rehabilitation Counseling Bulletin, 31,42-53. Received: September 30, 1991 Final Acceptance: March 5, 1992 Editor in Charge: John A. Nietupski