Barry S. Delin and Ellie C. Hartman Stout Vocational Rehabilitation Institute University of Wisconsin Stout

Transcription

1 Characteristics of Joint Participants in the Wisconsin Medicaid Buy-in and SSDI Benefit Offset : A Comparison with the Population of SSDI Beneficiaries Enrolled in the Wisconsin Medicaid Buy-in Barry S. Delin and Ellie C. Hartman Stout Vocational Rehabilitation Institute University of Wisconsin Stout 2010 MIG/DMIE Employment Summit Pittsburgh PA April 19, 2010 The authors thank the managers and staff at Pathways to Independence, housed at the Office of Independence and Employment, Wisconsin Department of Health Services, for their cooperation and support. The authors also thank staff at Mathematica Policy Research, Inc. for providing data and technical assistance without which this paper could not have been written. This work was supported by the Centers for Medicare and Medicaid Services, Medicaid Infrastructure Grant CFDA No and Social Security Administration Contract No. SS The descriptions and interpretations in this paper are solely those of the authors

2 1 Slightly over half the participants in the Wisconsin SSDI Employment (SSDI- EP) also were enrolled in the Wisconsin Medicaid Buy-in for some period during the time they participated in the pilot. 1 The SSDI-EP involved a test of a benefit offset feature for the Social Security Disability Insurance (SSDI) program which would incent earnings above the Substantial Gainful Activity level. This paper presents information about how similar the joint SSDI-EP and Buy-in participants were to other Buy-in participants, especially those who appear to meet pilot eligibility requirements, on a range of selected characteristics. These characteristics include a range of static demographic, disability, and public participation variables and employment rates and earnings over a roughly two year period. BACKGROUND Those enrolled in a Medicaid Buy-in program (for workers with a disability) will often be participating in one or more other public programs providing income support, health care, and/or other long term supports. In particular, a large proportion of those enrolled in a Medicaid Buy-in also participate in the Social Security Disability Insurance (SSDI) program. In 2006, about 71% of those enrolled (nationally) in a Medicaid Buy-in were also SSDI beneficiaries. Approximately 95% of these individuals were SSDI only; the remaining 5% also had concurrent participation in the Supplemental Security Income 2 (SSI) program. Medicaid Buy-ins are intended as work incentive programs. However, there is some variation in eligibility criteria and program features across states, due to both the specific federal statute used to authorize a state s program and states own choices. 3 Nonetheless, participation in a Medicaid Buy-in allows persons with serious disabilities to maintain access to Medicaid benefits while having earnings above the monthly Substantial Gainful Activity (SGA) level. 4 Additionally, Medicaid Buy-ins are intended to support employment over a protracted time period, at least technically independent of whether an individual maintains attachment to other public programs such as SSDI. 1 The Wisconsin Buy-in is officially known as the Wisconsin Medical Assistance Purchase Plan (MAPP). 2 Gimm, Gilbert, Davis, Sarah K., Andrews, Kristin L., Ireys, Henry T. Liu, Su The Three E s: Enrollment, Employment, and Earnings in the Medicaid Buy-in Program, Washington DC: Mathematica Policy Research, Inc. (Report Reference # , April 11, 2008) pp Authority to establish Medicaid Buy-ins for disabled workers comes from two federal statutes: The Balanced Budget Act of 1997 and the Ticket to Work and Work Incentives Improvement Act of Though the general purpose and structure of both types of Medicaid Buy-in are similar, the authorizing statutes provide for different eligibility standards, program features and ranges of allowable variation in how states operate their Buy-ins. 4 For Medicaid Buy-ins, disability is generally defined by the same standards as for eligibility for Social Security disability programs. The main difference is that Buy-in participants do not need to have earnings below the SGA level to establish or maintain eligibility. In point of fact, there are ways in which persons in Social Security disability programs can earn above the SGA level and retain program eligibility. However the conditions for doing so differ substantially between the SSI and SSDI programs, with the latter program s conditions being far more restrictive and time constrained. The 2010 SGA level is $1,000 per month ($1640 for those who are categorized as blind).

3 2 Wisconsin was one of four states where the Social Security Administration (SSA) tested a benefit offset feature for the SSDI program; the effort was known as the Wisconsin SSDI Employment (SSDI-EP). Under current law, a SSDI beneficiary cannot receive a cash benefit for months when the beneficiary has earnings above the SGA level, except during a time limited Trial Work Period (TWP). After a TWP has been completed, earning above the SGA level in any month results in the complete loss of the SSDI cash benefit for that month. This is the so called cash cliff and constitutes a major disincentive to work effort unless an individual is able to earn enough above SGA to completely replace the lost benefit and to justify the time spent in doing so. Moreover, after the thirty-six month Extended Period of Eligibility (EPE) that follows a completed TWP, having SGA earnings generally results in losing eligibility to the program and, ultimately, to Medicare as well. The benefit offset feature tested in Wisconsin and the other pilot states was intended to insure that additional earnings over SGA would leave the beneficiary better off. Beneficiaries using the offset would lose only $1 of their SSDI cash benefit for every $2 of earnings over the SGA level during the pilot. Though all of these beneficiaries would eventually be returned to regular program rules, SSA agreed not to have any period of above SGA earnings during the pilot used to disqualify these participants from continued attachment to the SSDI program. Although all of the offset pilots utilized random assignment, this did not mean that the characteristics of pilot participants necessarily approximated those of working age SSDI beneficiaries in each pilot state. SSA excluded beneficiaries who also participated in SSI and those who had not qualified for their SSDI benefits under their own earnings records. Participants were also volunteers. It is quite likely that some of the SSA mandated program requirements affected who actually enrolled in a pilot, as did the differing recruitment strategies each pilot used. 5 Therefore, one must consider the possibility that the participant outcomes observed from the pilot might not reflect what would have been observed if participants were truly representative of the state s beneficiary population. Unfortunately, there were few characteristics and even fewer outcomes on which SSDI-EP participants could be directly compared to Wisconsin s population of adult SSDI beneficiaries. The SSDI-EP was a relatively small project with a total enrollment of 496 participants. 50.8% (252) of these individuals were enrolled in the Wisconsin Medicaid Buy-in for some period between the calendar quarter in which they enrolled in the pilot and the end of the eighth calendar quarter subsequent to the enrollment quarter (i.e., 5 Eligibility was limited to beneficiaries no more than seventy-two months past their TWP completion. If enrolled and assigned to the treatment group, a participant could only use the offset until the seventy-second month following TWP, no matter how soon that occurred after enrolling in the pilot. The SSDI-EP delivered the program through approximately twenty community based agencies. Through the enrollment period (August 2005 through October 2006) these agencies conducted recruitment activities, most often targeted toward persons with some current or prior involvement with an agency. Approximately midway during the enrollment period, the state project office arranged to have recruitment letters sent to selected individuals who were either current consumers of the state s Vocational Rehabilitation agency or enrolled in the Medicaid Buy-in.

4 3 Q0-Q8 ). 6 The proportion in the treatment group (53.0%) enrolled in the Buy-in was 4.7% higher than in the control group (48.3%). 7 Roughly three-quarters of SSDI-EP participants in the Buy-in had entered that program by the time they entered the pilot. Just as it would be useful to know how closely SSDI-EP participants match the characteristics of those who would have access to a benefit offset provision should it become law, so too would it be useful to know how similar those who jointly participated in the SSDI-EP and the Wisconsin Buy-in were to those in the overall Wisconsin Buy-in population who might have access to an SSDI offset feature in the future. Unlike for SSDI-EP participants as a whole, it has been possible to directly investigate this issue for the sub-group that was also enrolled in the Buy-in. This opportunity was the result of two factors. The first was the availability through the Wisconsin Medicaid Infrastructure Grant (MIG) of longitudinal employment and earnings data for a large random sample of working age Buy-in participants. The second factor was the potential availability of data about Buy-in participants from the Integrated Data Set maintained for the Centers for Medicare and Medicaid Services 8 (CMS) by Mathematica Policy Research, Inc. (MPR). These data would support identifying those in the Buy-in samples also in SSDI and, among those, the individuals highly likely to meet pilot eligibility requirements. The MPR maintained data would also provide a common source of demographic, disability, and public program participation data that would support comparisons across all three types of Buy-in subgroups. Finally, using data from Wisconsin unemployment insurance (UI) records, it would also be possible to compare employment rates and average earnings for the subgroups over time. METHOD This paper compares information about three subgroups of Wisconsin Buy-in participants: (1) offset pilot participants who were also enrolled in the Buy-in, (2) Buy-in participants that appear to have met pilot eligibility requirements, and (3) Buy-in 6 This Q0-Q8 period was the primary one used to compare the employment outcomes of those in the SSDI-EP intervention group to the outcomes of those assigned to the treatment group. All participants who did not withdraw or die completed this nine quarter period before SSA returned intervention group members who had not completed their TWPs back to regular SSDI program rules in January As by the start of 2009 many participants had been in the project for some period beyond Q8, it was possible for participants to have some period of Buy-in participation that is not captured in the Q0-Q8 data. We will return to this issue later in the paper. 7 Though a slightly higher proportion of treatment group members were in the Buy-in immediately prior to entering the SSDI-EP, it appears that most of the observed, albeit marginal, difference between the treatment and control groups Buy-in participation was a result of choices made after participants joined the pilot. 8 This data set includes Medicaid, Medicare, SSA, and Ticket to Work data. CMS provides MIG grantees with potential access to the Integrated Data Set through MIG-RATS, the technical assistance network CMS established to promote state based evaluation of MIG sponsored activities. MIG grantees and those designated to perform research on their behalf can apply for access to aggregated data on a semi-annual basis. This paper uses data generated as a result of a spring 2009 proposal.

5 4 participants who were SSDI beneficiaries and would have had access to an offset provision had one had been statutorily incorporated into the Social Security Act. The information presented in this paper was obtained from three separate data sets. The first data set is that for the SSDI-EP evaluation. The second data set is the Touched by MIG Outcome Tracking System that is maintained on behalf of Pathways to Independence, i.e. the Wisconsin MIG. The third data set is the Integrated Data Set of Medicaid Buy-in participants maintained by MPR. While all three data sets contain individual level data, it was only possible to get aggregated data from MPR. Thus, this paper is limited to univariate descriptive analyses. The first subgroup, that of joint SSDI-EP and Medicaid Buy-in participants, is defined somewhat differently than it is for the SSDI-EP evaluation. Instead of defining the subgroup based on participation in the Buy-in during the primary Q0-Q8 period, inclusion is based on any Wisconsin Medicaid Buy-in participation between January 1, 2005 and December 31, As a consequence, both the number and proportion of SSDI-EP participants included in the subgroup for this analysis is slightly higher than observed for the Q0-Q8 period (269 instead of 252 and 54% instead of 51%). 9 The members of other subgroups were identified by MPR from its Integrated Data Set using cases provided from the Wisconsin Touched by MIG Outcomes Tracking System. Every January since 2005, researchers working for the Wisconsin MIG (Pathways) take random samples of Buy-in participants and other Medicaid participants for reason of disability between ages twenty-one and sixty-four. 10 UI employment outcomes for the members of these cohorts are then followed over a nine calendar quarter period. We provided MPR with the identifiers for the Buy-in samples for both 2005 and 2006 as those roughly corresponded with the SSDI-EP s enrollment period. MPR then identified the cases from each of these samples that were SSDI beneficiaries, and, from that group, the smaller number of cases that appeared to meet the offset pilot eligibility criteria. The sample sizes for the joint Buy-in/SSDI groups are 3,546 for 2005 and 4,236 for The sample sizes for the pilot eligible subgroups 9 The subgroup thus includes a small number of individuals who either left the Buy-in by the start of the Q0-Q8 period or entered it after Q8 (though while still in the SSDI-EP). While it is quite likely that the seventeen additional cases, especially any where there was no Buy-in participation after SSDI-EP enrollment, decreased the absolute accuracy of the subgroup characteristics reported here, there is no reason to think they greatly altered the level or distribution of characteristics relative to the other subgroups (SSDI beneficiaries in the Buy-in and those in the Buy-in who putatively met SSDI-EP eligibility requirements). The reason that we defined the joint Buy-in and pilot participation group as we did was to expedite MPR s ability to perform our requested data draws and analyses. Recall that that the Q0-Q8 period for each individual is relative to that individual s enrollment date. SSDI participants enrolled over a fifteen month period beginning in August 2005; the final enrollees did not finish their Q8 until the end of However, the Integrated Data Base is structured on an annual basis. The data period was the most tightly defined way to encompass every SSDI-EP participant s full Q0-Q8 period. 10 The Buy-in samples are 50% of working age participants in the month of the data draw. The age range was chosen to match that defining working age in the Cornell University summaries of disability data from the American Community Survey.

6 5 were smaller, 2,026 for 2005 and 2,508 for 2006 (respectively 57% and 59% of the SSDI groups). Following identification of the relevant cases for each of the sub-groups, MPR queried its records for the desired static baseline variables that would be provided to the authors in aggregate form. 11 MPR also generated nine quarter long series of employment rates and earnings for each of the SSDI and putative offset pilot eligible groups based on the UI data we provided for the cases in the January 2005 and 2006 samples of Buy-in participants. We generated the nine quarter time series for the joint SSDI-EP and Buy-in subgroup based upon each individual s calendar quarter of pilot enrollment and restricted the cases to the 239 pilot participants who both completed the full Q0-Q8 period and were enrolled in the Buy-in for some part of that period. All monetary values, whether for static variables or for the earnings time series, were deflated using the August 2005 value of the Consumer Price Index. 12 Finally, it is important to observe that the data draws required resolution of a number of technical issues. This was particularly true for deciding how to identify those individuals putatively eligible to participate in the offset pilot. We thank MPR staff, both known and unknown, for their efforts. 13 Indeed, we think it important to point out that a key aspect of the process MIG-RATS developed for allowing MIG grantees access to data from the Integrated Data Set is the iterative consultation between the state based researchers and MPR staff that takes place after a proposal has been approved. FINDINGS In this section we present information about both the baseline characteristics of our subgroups and the nine quarters of employment and earnings outcomes data. Our comparisons focus on the similarities and differences between the SSDI-EP participants and the other sub-groups. Though there are three types of subgroups, data are provided for an actual total of five subgroups. While there is only one subgroup of pilot participants, there are two subgroups for Medicaid Buy-in participants in SSDI and those MPR identified as meeting offset pilot eligibility requirements. Recall that we provided samples of working age 14 Medicaid Buy-in participants for both January 2005 and Baseline characteristics 11 The values for the joint SSDI-EP and Buy-in subgroup differ slightly from those produced by the SSDI-EP evaluation records. These differences reflected (a) the somewhat different time points at which such data was collected (at SSDI-EP enrollment for the SSDI-EP evaluation rather than the annualized basis for the Integrated Data Set) and (b) the slightly different compositions of the joint pilot participant/buy-in groups already identified. 12 Specifically this was the =100 version of the CPI-U adjusted so that the August 2005 is 100. This formulation was used for the SSDI-EP evaluation and was adopted for the MPR analysis to insure comparability. 13 In particular, we note with appreciation the efforts of Kristin Andrews who coordinated the technical assistance process at MPR. 14 There is a substantial probability (nearing 50%) that an individual in one of these Buy-in participant samples will also be in the one for the other year. Additionally, it is almost certain that some of SSDI-EP participants are included in these subgroups, but as the likely proportion in the

7 6 reflect data for these two time points. The Q0s of the employment outcome data are those for the first calendar quarters containing the relevant sampling month. 15 Baseline characteristics for the SSDI-EP participant group are drawn from January 2006 data rather, than in the SSDI-EP evaluation, fixed to the actual enrollment date. The January 2006 date was roughly a third of the way through the pilot s enrollment period and served as a useful compromise given the structure of the Integrated Data Set. It is also critical to point out that SSDI-EP participation did not necessarily mean an individual actually used the offset feature. In point of fact, only a relatively small proportion did. Roughly half the participants were assigned to the control group and had no access whatsoever. Those in the treatment group had to complete their TWP and then had to earn above SGA in at least one month thereafter in order to actually use the offset. By mid-2009, only about 11% of those who entered the SSDI-EP had at least one month of offset use. Finally, readers are reminded that the descriptions that follow apply only to persons enrolled in the Wisconsin Medicaid Buy-in. When we provide information about or make comparisons between SSDI-EP participants, those who appear to have been eligible for SSDI-EP participation, and all working age SSDI, we always mean Medicaid Buy-in participants who are included in those categories. Subgroup Baseline Characteristics Age is widely identified as a factor that affects the probability of successful return to work. Younger individuals have a higher probability of a successful return to work, while older persons are more likely to remain on disability benefits. Table 1 exhibits mean and median age for our subgroups. It also exhibits distributions across four age categories. The SSDI-EP participants are substantially younger than the members of the four other sub-groups; these differences are found to be statistically significant at p-values of or less. For example, the median age of SSDI-EP participants is between three and five years lower than for the other subgroups. These differences are also reflected in the age distributions. In particular, the proportion of SSDI-EP participants in the 55 and older segment of the age distribution (13%) is less than one half that observed in the distributions for the other subgroups (a minimum of 28%). other sub-groups would be modest (e.g. about 3% for the SSDI groups, perhaps 6% of the putative eligible groups). We judged that the distorting effects of their chance inclusion would not significantly affect our ability to discern major differences across the subgroups. 15 Thus an individual included in both the 2005 and 2006 samples would have different time spans of employment outcomes data included in the aggregate statistics for each of those samples. For the 2005 subgroups the relevant period starts with the first quarter of 2005 and ends with the first quarter of For the 2006 subgroups the relevant period is first quarter 2006 through first quarter Significance test results were provided by MPR as only they had access to the individual level data from the Integrated Data Set.

8 7 Table 1: Age at Subgroup Entry Date by Participant Group, Eligible Groups, and All SSDI Groups Mean Age Median Age Under or Older Participants % 27.9% 38.3% 13.0% Eligible Eligible % 21.7% 34.7% 30.3% % 20.0% 36.9% 31.7% All SSDI % 24.1% 33.3% 27.7% 2005 All SSDI % 22.1% 35.8% 29.2% Source: MPR Integrated Data Set Note: There are 0.2% missing cases from both the 2005 and 2006 All SSDI subgroups. Table 2 provides information about the gender distribution and racial/ethnic composition of all five subgroups. The SSDI-EP participant group is more heavily male than the other subgroups. The proportion of males in the subgroup is almost 6.5% higher than for females. By contrast, the differences seen within the other subgroups are never more than 2%. This suggests that there was something about the SSDI-EP recruitment and enrollment dynamics that resulted in a somewhat unrepresentative distribution relative to the potential population of Buy-in users that might have entered the pilot or the population of Buy-in users who might be qualified for a statutory offset. Nonetheless, the differences are not found to be statistically significant. By contrast, the measured racial and ethnic characteristics of the offset pilot participants in the Buy-in appear to reasonably match those of the other subgroups. It appears that the proportion of black individuals in the SSDI-EP participant group is a bit larger than the others (perhaps reflecting the greater number of pilot sites in urban areas), but again the results are not statistically significant at the.05 level.

9 8 Table 2: Gender and Race/Ethnicity by Participant Group, Eligible Groups, and All SSDI Groups Female Male White (non- Hispanic) Black (non- Hispanic) Hispanic Participants All Others 46.8% 53.2% 86.3% 7.4% 2.6% 1.5% Eligible Eligible % 49.4% 88.8% 5.5% 2.6% 1.3% 51.0% 49.0% 89.4% 5.4% 2.2% 1.2% All SSDI % 50.2% 89.6% 4.7% 2.3% 1.0% 2005 All SSDI % 50.2% 89.6% 5.0% 1.9% 1.2% Source: MPR Integrated Data Set Note: Hispanic is included in the Race and Hispanic Status categories. Note: For Sex, there are 0.2% missing cases from both the 2005 and 2006 All SSDI subgroups. There are missing data for all five subgroups for Race and Hispanic Status. The percentage of missing data ranged from 1.8% to 2.5% Our next table displays information about educational attainment. In the general population, greater educational attainment is strongly associated with higher earnings and career development. Readers are alerted that the proportion of missing cases range from two-thirds to three quarters of cases in each subgroup. Table 3: Educational Attainment by Participant Group, Eligible Groups, and All SSDI Groups Mean Years Median Years 13 to 15 Years 16 Years or more Missing Data Participants % 5.6% 69.1% Eligible Eligible % 2.7% 66.0% % 2.9% 69.9% All SSDI % 1.8% 74.6% 2005 All SSDI % 2.0% 74.8% Source: MPR Integrated Data Set

10 9 Nonetheless, we think the data in table 3 can provide useful insight into how the pilot participants in the Buy-in differ from those in the other subgroups. While median years of educational attainment are identical across the five subgroups, the mean score for the SSDI-EP subgroup is a minimum of 0.6 years higher than for any of the other subgroups. Moreover, the largest difference between any of the Eligible and All SSDI subgroups is only 0.2 years. The (apparently) greater educational attainment of those in the pilot is most evident in the proportion with sixteen or more years of education. Roughly twice as high a proportion of offset participants are in this category than in any of the other subgroups. Table 4 displays data about the distribution of primary disabling conditions for the five sub-groups. MPR defined these categories by grouping similar SSA primary disability codes (DIGs) and has used these categories for some Buy-in related analyses. 17 While inclusion in any one of these categories may have little or no import as to the capacity to work, the service and support needs associated with different types of disabling conditions can vary greatly. Crucially, so too can the availability of the resources needed to meet service and support needs, especially across states and localities. Table 4: Distribution of Primary Disabling Conditions and Proportion Having Permanent SSA Disability by Participant Group, Eligible Groups, and All SSDI Groups Participants Mental Illness and Other Mental Disorders Mental Retardation Musculoskeletal Sensory Other or Missing Have Permanent SSA Disability 50.9% 5.9% 11.9% 3.7% 27.5% 30.9% Eligible Eligible % 6.0% 16.9% 2.2% 33.8% 39.4% 40.7% 6.0% 17.3% 2.0% 34.1% 38.8% All SSDI % 14.5% 12.0% 2.3% 35.7% 40.0% 2005 All SSDI % 13.8% 12.8% 2.2% 35.3% 38.6% Source: MPR Integrated Data Set Note: There is substantial missing data for the All SSDI subgroups for the Permanent SSA Disability variable, 10.2% for the 2005 sample and 9.8% for the 2006 sample. The percentage of cases with missing data for all other subgroups ranges from 2.2% to 2.9%. 17 Additionally, SSA asked that these categories (slightly recoded) be used by those evaluating each of the benefit offset pilots for describing participant characteristics and to specify disability subgroup models that were run for all four evaluations.

11 10 There are notable differences between SSDI-EP participants who were enrolled in the Buy-in and the other sub-groups. The difference between the distribution of SSDI- EP participants and that for any of the other subgroups is always statistically significant. Most notably, the proportion of pilot participants in the mental illness and other mental disorders category is about ten percentage points higher than for the pilot eligible subgroups and about fifteen percentage points higher than for all working age SSDI beneficiaries in the Buy-in. Restated, the relative prevalence of affective disabilities (at least as a primary disabling condition) appears to be much higher among offset pilot participants than among the other subgroups. Other observed differences were largely between the SSDI-EP participant group and either, but not both, of the other types of subgroups. participants in the Buy-in were less than half as likely as the overall SSDI beneficiary group to have primary disabilities categorized as mental retardation. By contrast, the proportions included in the mental retardation category were about the same for the pilot and the subgroups who met pilot eligibility standards. This pattern is reversed for the musculoskeletal category. SSDI-EP participants and the general beneficiary subgroups have similar proportions of individuals with a primary disability classified as musculoskeletal. This time it is the subgroups that are putatively eligible for the pilot who differ. About 17% of the members of these subgroups have a primary disability that is musculoskeletal as compared to about 12% for the pilot participants. Though our discussion stresses comparisons between pilot participants in the Buy-in and the other subgroups, the distributions of the pilot eligible and all SSDI subgroups are also significantly different from each other. While differences between the pilot participant and pilot eligible groups probably reflect differences in both project recruitment efforts and individual preferences, differences between the pilot eligible and all SSDI subgroups are rooted in SSA s choices about offset pilot eligibility requirements. In particular, the exclusion of both Disabled Adult Children (DACs) and concurrent beneficiaries almost certainly lowered the proportion of participants with certain kinds of cognitive impairments ( mental retardation ). The last column (to the right) in table 4 exhibits the proportion in each subgroup that SSA considers as having a permanent disability. Such conditions are understood as those for which no medical improvement is expected and, thus, where there is very little expectation at SSA (or at the Disability Determination Service which adjudicated the beneficiary s case) that the beneficiary will leave the SSDI program prior to full retirement age. About 31% of those in SSDI-EP subgroup are so classified. This is eight or more percentage points lower than for all the other subgroups. These differences are statistically significant at the.05 level and suggest that the offset pilot participants might be viewed as a less severely impacted group, a finding that would seem consistent with a willingness to enter a return to work program with the explicit goal of increasing earnings above the SGA level. Table 5 displays a number of SSDI program related characteristics. The proportion with Medicare A eligibility, which requires twenty-four months of SSDI eligibility), is meant to indicate the relative level of newer beneficiaries in the subgroups. The TWP and EPE elements are measures of work incentive usage and are aimed at

12 11 indicating serious involvement in return-to-work efforts. 18 Finally, as the Primary Insurance Amount (PIA) reflects prior earnings, differences in the average and median values across subgroups provides a relative indicator of labor market outcomes prior to entering SSDI. It is important to note certain limitations in using these data, particularly the TWP and EPE elements, to indicate baseline conditions for the SSDI-EP subgroup. The data in table 5 for SSDI-EP participants are from January Recall that this is about a third of the way through the enrollment period. We know that participating in the pilot increased the number who started or completed a TWP. As a consequence, the TWP started and in EPE data for pre-2006 SSDI-EP enrollees may be somewhat higher 19 than the true baseline values. Table 5: Selected SSDI Program Characteristics by Participant Group, Eligible Groups, and All SSDI Groups Eligible for Medicare A TWP Started but Not Completed In EPE Mean Monthly Primary Insurance Median Monthly Primary Insurance Participants Amount Amount 82.5% 24.9% 17.5% $917 $891 Eligible Eligible % 10.6% 7.6% $918 $ % 8.7% 5.9% $926 $905 All SSDI % 7.7% 5.1% $915 $ All SSDI % 6.9% 4.2% $920 $894 Source: MPR Integrated Data Set The vast majority of those in all five subgroups had at least twenty-four months of SSDI eligibility at the nominal baseline date for each subgroup. The data in table 5 indicate that the proportion of relatively new beneficiaries is higher in the all SSDI subgroups, but the differences with the offset pilot and pilot eligible subgroups are modest (yet generally statistically significant). 18 As participation in a Medicaid Buy-in requires employment (or, as in the case of the Wisconsin Buy-in, participation in a program expected to result in employment in a limited time period) one could argue that Buy-in participation demonstrates a similarly high level of involvement as TWP or EPE participation. This may be true in other states, but participation in the Wisconsin Buy-in does not require any earnings, let alone the level associated with expenditure of a TWP month (roughly 70% of SGA) % of pilot participants enrolled before January 2006.

13 12 Those in the pilot participant subgroup are far more likely than those in any of the other subgroups to have either started or completed a TWP by the nominal baseline date. Both the start and completion rates (the later indicated by the percentage in EPE ) are far higher for the pilot subgroup. Proportions are, at minimum, 2.3 times higher than the nearest of the other subgroups. Needless to say, the differences are statistically significant. Though the differences are not as stark, those in the pilot eligible subgroups are also more likely to have started or completed a TWP than those in the all SSDI subgroups. Though these results are also generally significant, all of these subgroups resembled each other more than they did the SSDI-EP participant subgroup. Finally, there are no significant differences between any pair of subgroups on the PIA measures. Indeed, the means and medians are surprisingly alike, suggesting little difference in pre-disability earnings. While this result is not unexpected for the pilot eligible subgroups, it is for the all SSDI subgroups which contain both DACs and persons with poor enough earnings histories to qualify for SSI. While we cannot fully explain the observed similarity, MPR has informed us that the PIAs for DACs in the all SSDI subgroups had been removed from the calculations due to uncertainty about whether their PIA values reflected their own earnings histories on those of a parent. 20 This exclusion should have increased mean and mean PIA values relative to those for pilot participants and for the pilot eligible sub-groups. However, it remains likely that the inclusion of PIA values for concurrent beneficiaries should still have resulted in lower means and medians for the all SSDI subgroups relative to those for all the other subgroups. A MPR staffer hypothesized that the equalizing factor might have been the larger proportion of older beneficiaries in the all SSDI subgroups compared to that for SSDI-EP participants (see table 1). All things being equal, those in the workforce longest should have the highest PIAs. 21 Unfortunately, this explanation is not completely satisfying. If it were, the mean and median values for the pilot eligible subgroups (which have about the same age distributions as the all SSDI subgroups, but include no DACs) should be higher than those for any of the other subgroups. 22 Subgroup Employment and Earnings Outcomes, Q0-Q8 We turn now to the employment and earnings trends data for our five subgroups of Wisconsin Medicaid Buy-in users. These data are from the baseline calendar quarter for each group through the eighth subsequent calendar quarter. For one pair of pilot eligible and all SSDI subgroups, the initial quarter (Q0) is the first calendar quarter of For the other pilot eligible and all SSDI subgroups, the initial quarter is the first three months of While everyone in these subgroups was a Buy-in participant during the month they entered the subgroup sample (i.e., January 2005 or January 20 We lack information about whether values for widows and widowers receiving benefits based on their deceased spouses earnings records were also excluded. 21 For this to be true, it would also be necessary that the older beneficiaries actually had longer work histories rather than being in the SSDI program for longer periods. The data in the Eligible for Medicare A column of table 5 is consistent with meeting this condition. 22 Another consideration is that The Wisconsin Buy-in has a premium structure that treats unearned income far more disadvantageously than earnings and treats SSDI payments as unearned income. Thus, if anything, beneficiaries with high PIAs are likely to face a disincentive for participating in the Buy-in.

14 ), continued Buy-in participation through Q8 varied. The vast majority either remained in the Buy-in for all or most all of the Q0-Q8 period. The Q0-Q8 periods for those in the pilot (SSDI-EP) subgroup begin in the calendar quarter of enrollment. For example, if a participant enrolled in April 2006, Q0 would be the second calendar quarter of Additionally, the Q0-Q8 trend lines reflect data for SSDI-EP participants who were enrolled in the Buy-in at some point in that period. Thus, unlike the other subgroups, one can t assume that all of these individuals were in the Buy-in in the first month of entering the pilot (though three-quarters were). Additionally, as already discussed, these cases are marginally different from the set MPR identified to generate the baseline characteristics presented above. The employment and earnings data presented are from Wisconsin Unemployment records. 23 There is high compliance with UI reporting requirements; however, there are jobs that do not need to be reported to the system. Important exclusions include self-employment, employment at firms located outside the state, and employment at very small or religious not-for-profit entities. Additionally, as the data are quarterly, it is not possible to determine whether an individual was employed throughout the three month period or how much was earned in any given month. 24 The five charts that follow compare the SSDI-EP subgroup to the pilot eligible subgroups, the all SSDI subgroups, or both. As the trend lines for the SSDI-EP participants combine data for individuals assigned to the treatment group and to the control group, it is important to know how large the differences in employment outcomes between these study assignment groups are. If the differences are large, it could be argued that differences between the offset pilot subgroup and the others result largely from the behavior of the treatment group. In point of fact, there are not large differences between the Q0-Q8 employment and earnings trends for those assigned to the intervention group and those not. Though Buy-in participants in the intervention group had slightly higher employment rates and mean earnings than those assigned to control in the final quarters of the Q0-Q8 period, 25 the overall trends for the period are not significantly different. These results mirrored those for the overall SSDI-EP participant group (i.e. both those in the Buy-in and those 23 Cases where there was no data in the UI records for any or all relevant quarters were imputed not to have UI employment or UI earnings for those time periods. 24 Thus it is impossible to determine from UI data whether an individual had monthly earnings that were equal or greater than either SGA or the TWP trigger amount. 25 See the data table in the appendix to this paper for more detailed information.

15 14 who did not use the Buy-in). 26 Consequently, we would not attribute much of the observed differences in employment related differences to pilot intervention effects. 27 Figure 1 displays the Q0-Q8 employment rate trends for the SSDI-EP participants and the two subgroups of those who are pilot eligible. It is interesting to note that the UI employment rates for all these subgroups are much lower than would be expected given that all Wisconsin Buy-in participants, save those in an approved employment preparation program or temporarily exempted from employment due to illness, will be employed. The difference between actual and expected employment rates mainly reflects differences in what counts as reportable employment for UI and the Wisconsin Medicaid Buy-in. 28 Throughout the Q0-Q8 period the employment rate for those in the SSDI-EP subgroup is always far higher than for the subgroups of those who appear to meet pilot eligibility requirements. UI Employment rates for the SSDI-EP subgroup hover around 50%, generally 20 percentage points higher than those of the pilot eligible subgroups. By contrast the UI employment rates for the pilot eligible subgroups are only modestly different from each other (always within five percentage points) It is of course possible that differences in the trends will become significant over longer periods. Though we intend to continue following both UI data and Buy-in participation until the end of 2011, two factors will make the analysis more challenging. First, the treatment group members who did not complete a TWP by the end of 2008 were immediately returned to regular SSDI program rules. Second, those in the treatment group who completed their TWPs will lose access to the offset feature after the seventy-second month following TWP completion. 27 Barely a fifth of SSDI-EP treatment group members had used the offset provision by mid Unfortunately, we do not have good information about the timing of the first month of usage. First usage could not occur until after the three month grace period that followed the TWP completion month, however the first use might occur later in some cases and some TWP completers may never use the offset before reaching the end of the seventy-second month following TWP completion. Moreover, whatever the actual rate of offset use in the Q0-Q8 period, it must be somewhat lower than the TWP completion rate as of 12/31/2008 when non-completers in the treatment group were returned to regular SSDI program rules. 28 We have already noted that Wisconsin UI records exclude certain kinds of employment. Additionally, the employment of those receiving only in-kind compensation (allowable under Wisconsin Buy-in rules) would not be included in UI records. 29 We did not provide MPR with individual level employment and earnings data for the SSDI-EP subgroup. This was an unfortunate omission as it did not allow MPR to perform significance tests of the differences between the SSDI-EP subgroup and the other subgroups on our behalf. Nonetheless, the scale of the differences observed in figures 1 through 5 is large and it is inconceivable that they are not statistically significant at the.05 level.

16 15 Figure 1: UI Employment Rates Q0-Q8 for SSDI-EP Participants and Eligible Subgroups UI Employment Rates, SSDI-EP Participants and Eligible Non-Participants 60% 50% 40% 30% SSDI-EP Eligible 05 Eligible 06 20% 10% 0% Q0 Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Relative Quarter Source: Wisconsin UI records Figure 2 allows comparison between the UI employment rates for the SSDI-EP subgroup and the two subgroups reflecting outcomes for all working age SSDI beneficiaries. Though the all beneficiary subgroups include persons that might be expected to be somewhat less likely than the pilot eligible subgroups to have UI reported employment, the subgroups employment rates are marginally higher than for the pilot eligible subgroups. 30 Nonetheless, the all SSDI subgroups have UI employment rates that hover about the 30% level, and stand in about the same relation to the pilot participant subgroup as those for the pilot eligible subgroups. 30 For example, one of the most salient differences between the all SSDI and pilot eligible subgroups is the much higher proportion of those with a primary disability classified as mental retardation in the former. The data in table 4 show the proportion to be more than twice as high as for those in the pilot eligible subgroups or, for that matter, the SSDI-EP participant subgroup. Though we lack data, we would have expected a relatively high proportion of those classified as having mental retardation to be employed in sheltered workshops or similar enclaves. Such employment is not reported to the Wisconsin UI system.

17 16 Figure 2: UI Employment Rates Q0-Q8 for SSDI-EP Participants and All SSDI Subgroups UI Employment Rates, SSDI-EP Participants and All SSDI 60% 50% 40% 30% SSDI-EP All SSDI 05 All SSDI 06 20% 10% 0% Q0 Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Relative Quarter Source: Wisconsin UI records It is notable that despite some significant differences in their baseline characteristics the pilot eligible and all SSDI subgroups exhibit very similar employment rate trends over Q0-Q8. Indeed all of these subgroups posted declines in employment rates over the analysis period of 3-4%. Though hardly a desirable result, the employment rate for the SSDI-EP subgroup was essentially the same in Q8 as it was in Q0. Figure 3 and figure 4 display the mean quarterly UI earnings trends for the five subgroups. Again, we emphasize comparisons between the SSDI-EP subgroup and the respective pairs of subgroups for those who are pilot eligible and for all beneficiaries. As displayed in figure 3, SSDI-EP participants in the Buy-in modestly increased their earnings over the Q0-Q8 period. The increase was from $931 to $1067, a gain of 14.6%. By contrast, mean earnings for the two pilot eligible subgroups were typically $500 or more lower. Further, the 2006 sample witnessed no earnings growth over the Q0-Q8 period while the 2005 subgroup experienced a 6.5% decline. Clearly, those in the SSDI-EP subgroup faired much better, but mainly because they started from a better position at Q0. Once again, the differences between the two pilot eligible subgroups are modest relative to the differences with the pilot participant subgroup.

18 17 Figure 3: Mean UI Quarterly Earnings Q0-Q8 for SSDI-EP Participants and Eligible Subgroups UI Mean Earnings per Quarter, SSDI-EP Participants and Eligible Non-Participants $1,200 $1,000 $800 $600 SSDI-EP Eligible 05 Eligible 06 $400 $200 $0 Q0 Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Relative Quarter Source: Wisconsin UI records Quarterly mean earnings are a bit higher for those in the all SSDI beneficiary subgroups than for the pilot eligible subgroups, though still far behind those for the SSDI-EP subgroup. Yet over the Q0-Q8 period these subgroups tend to lose ground to all the other subgroups. For example, Q0 mean earnings for the 2006 all SSDI subgroup were $110 higher than for the 2006 pilot eligible subgroup and almost $475 less than the SSDI-EP subgroup. By Q8, the advantage over the 2006 pilot eligible subgroup had decreased to $75, whilst the deficit relative to the SSDI-EP group had grown to $645. Moreover, both of the all SSDI groups suffered comparable declines in average earnings over their Q0-Q8 periods; -8.0% for the 2005 cohort, -7.7% for the 2006.

19 18 Figure 4: Mean UI Quarterly Earnings Q0-Q8 for SSDI-EP Participants and All SSDI Subgroups UI Mean Earnings per Quarter, SSDI-EP Participants and All SSDI $1,200 $1,000 $800 $600 SSDI-EP All SSDI 05 All SSDI 06 $400 $200 $0 Q0 Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Relative Qurater Source: Wisconsin UI records Of course average earnings reflect differences in employment rates. Given the roughly 20% difference in the employment rate trend lines between the SSDI-EP subgroup and the other four, it is reasonable to ask whether the observed differences are present for those who have UI employment in any given calendar quarter. The answer is that the differences in mean earnings that are exhibited in figures 3 and 4 persist, though the differences are somewhat narrowed. Figure 5 displays earnings trends for those employed in any calendar quarter for all five of our subgroups. Again, those in the SSDI-EP subgroup display the best performance. Mean earnings are $1953 per quarter at Q0 and $2257 at Q8. This represents a gain of $305 or 15.6%. It is also notable that mean quarterly earnings are approaching the three month equivalent of the SGA level, ending the Q0-Q8 period at 91% of that benchmark. 31 A look at figure 5 also confirms that the earnings trends for those who have UI reported employment are very similar across all of the other subgroups. Differences between any pair of these four subgroups pale in comparison to the difference between any of them and the SSDI-EP subgroup. For instance, in Q8 the 2005 all SSDI subgroup posted the highest level of mean earnings except for the SSDI-EP subgroup. Still, the mean value for the 2005 all SSDI subgroup was only 75% of that for those in the SSDI- 31 The quarterly equivalent of the SGA level is $2490 in August 2005 constant dollars.

20 19 EP subgroup with non-zero earnings in that quarter. Though this represents a considerable narrowing of the earnings differential (the mean earnings across the full 2005 all SSDI subgroup at Q8 was only 45% of those of the pilot participants), it indicates that an appreciable part of the disparity was caused by some factor or factors other than differences in the employment rate. 32 Figure 5: Mean UI Quarterly Earnings Q0-Q8 for Those with UI Reported Employment, All Subgroups UI Mean Earnings per Quarter for Those Employed, All Groups $2,500 $2,000 $1,500 $1,000 SSDI-EP Eligible 05 Eligible 06 All SSDI 05 All SSDI 06 $500 $0 Q0 Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Relative Quarter Source: Wisconsin UI records Note: Calculated by dividing mean quarterly earnings by the employment rate for that quarter Conclusion When we submitted our proposal to MIG-RATS to obtain data from the MPR Integrated Data Set one of our main goals was to learn how closely those jointly participating in the Wisconsin Medicaid Buy-in and the Wisconsin SSDI-EP Employment resembled other Buy-in participants, specifically (1) those who were eligible to participate in the SSDI benefit offset pilot but did not do so and (2) those who would be eligible for an offset should one ever be added to the law. We conceptualized this second group as working age SSDI beneficiaries participating in the Buy-in. 33 An 32 Differences in mean earnings can also reflect variation in hours worked and/or hourly wage rates. Unfortunately, Wisconsin UI data does not include these kinds of information. 33 It is our understanding that the Benefit Offset National Demonstration, scheduled to begin in 2011, will utilize primary study groups that include all adult beneficiaries. It is also our