The Prospect of Losing Benefits and the Work Decisions of Participants in Disability Programs

Abstract

Participants in public disability programs face the loss of benefits if they work and earn at substantial levels. Policy makers, researchers, and advocates have suggested that the prospect of benefit loss is, at least in part, an explanation for the low levels of earnings and work participation among disability program participants; however, research on the actual effects is very limited. The authors estimate the prospect of benefit loss effects on work participation and earnings of participants in four disability programs using a unique survey. The findings strongly suggest that the prospect of benefit loss decreases earnings and work participation among disability program participants. Unexpectedly, the authors found little variation in the size of the prospect of benefit loss effects across programs even though there is substantial cross-program variation in the actual benefit loss that occurs with work participation and increased earnings.

Keywords

health care Medicare/Medicaid policy eligibility

Persons participating in most public disability programs face the loss of benefits if they work and earn at substantial levels. The prospect of losing benefits affects participants’ work decisions including whether to work and, if working, how much to work. This study examines how the prospect of losing benefits affects the work decisions of participants in four disability programs with different benefit loss characteristics: Massachusetts Medicaid for persons with disabilities (MassHealth), Medicare, Supplemental Security Income (SSI), and Social Security Disability Insurance (DI).

Four characteristics define the benefit loss within disability programs. Each of the characteristics potentially affects participants’ work decisions. The first is the program earned income limit. Nearly all programs place a limit on how much participants may earn. These limits affect earnings if the limits are sufficiently low so that participants limit their work hours to keep their benefits. The second benefit loss characteristic is the benefit loss at the earnings limit. This is the loss that occurs when a person’s earning crosses over the earnings limit. Some programs (e.g., DI and Medicare) have very high loss as the full benefit is lost as earnings cross the limit. Other programs have lower loss at the earnings limit because benefits are gradually decreased with increasing earnings and only a small remaining benefit is lost as earnings cross the limit (e.g., SSI). The benefit loss at the earnings limit will decrease earnings if the loss is sufficiently large so participants limit their earnings to avoid the loss. The third characteristic is the benefit reductions at earnings levels below the earnings limit. Some programs, for example SSI, reduce benefit amounts for each dollar earned over the entire range of earnings below the earnings limit. These benefit reductions effectively lower participants’ wages and consequently affect whether people work and, among the working, how much they work. The fourth characteristic is the benefit reinstatement provisions. Once benefits are terminated, reinstatement of benefits requires a new application and new benefit waiting periods resulting in costs for applicants. The work decisions of participants will be affected by the costs and risks of reinstatement. Some participants may limit their earnings to avoid the costs and risks.

The earnings limit, the benefit loss at the earnings limit, the benefit reductions below the earnings limit, and the reinstatement provisions, in combination, define the benefit loss characteristics for a particular program. We expect the benefit loss characteristics of a disability program will affect persons’ work decisions. However persons’ understanding of program benefit loss characteristics is often not complete (Hennessey, 1997; MacDonald-Wilson, Rogers, Ellison, & Lyass, 2003), and it is the perceived program benefit loss characteristics, rather than the actual benefit loss characteristics, that affect work decisions. In our analysis, we examine the effects of perceived benefit loss.

Specifically, this study estimates the following two effects: (a) the effects of the prospect of benefit loss on the earnings of working disability benefit participants and (b) the effects of the prospect of benefit loss on the work participation of nonworking disability benefit participants. We estimate effects for the MassHealth, Medicare, SSI, and DI programs. In the remainder of the article, we refer to the prospect of benefit loss effects as the “benefit loss effects”. We examine the program characteristics to determine the expected effects, assuming participants have a perfect understanding of program benefit loss characteristics. To estimate the empirical effects, we use a unique survey of MassHealth participants. Survey respondents report whether they “do not work” or they “earn less” because of benefit loss for each program in which they are participating. The survey responses provide a direct estimate of the benefit loss effects by program.

Understanding the benefit loss effects is important. The work participation rate among disability program participants is very low: approximately 9% for the SSI and DI programs (Livermore, 2008). Also, among working DI and SSI participants, earnings are low, averaging only US$622 per month (Livermore, 2008). Policy makers, researchers, and advocates have suggested that benefit loss effects are, at least in part, a reason for the low levels of work participation and low earnings (D. C. Stapleton, O’Day, Livermore, & Imparato, 2006). Consequently, recent disability program initiatives have addressed benefit loss. The Balanced Budget Act of 1997 and the Ticket to Work and Work Incentives Act of 1999 (Ticket Act) gave states the authority to reduce the Medicaid benefit loss for working persons with disabilities through Medicaid Buy-In programs. Through Medicaid Buy-In, working persons with disabilities with income in excess of the standard Medicaid limits can buy in to Medicaid by paying premiums. As of the end of 2009, 42 states were operating Buy-In programs (Kehn, Croake, & Schimmel, 2010). The Ticket Act also reduced the benefit loss in the DI and SSI programs by expanding benefit reinstatement periods and reducing benefit reinstatement costs. For DI recipients who lose eligibility because of earnings, the Ticket Act extended the period of post-DI Medicare coverage. In addition, the Social Security Administration (SSA) is currently implementing a national demonstration, Benefit Offset National Demonstration (BOND), to empirically determine the work participation and earnings effects of DI program changes that reduce the benefit loss at the earnings limit, namely, a benefit offset of US$1 of benefits for US$2 of earnings above the substantial gainful activity (SGA) amount (D. Stapleton, Bell, Wittenburg, Sokol, & McInnis, 2010).

Prior Research

Although there have been numerous policy initiatives to reduce benefit loss, there is very little prior research on the effects of benefit loss characteristics on the earnings and work participation among disability program participants. Research is limited because it is very difficult to find sources of exogenous variation in benefit loss characteristics to identify the effects. Existing research is limited to estimates of the effects of only a single benefit loss characteristic, the earnings limit. There are no estimates for the effects of the size of the benefit loss at the earnings limit, benefit reductions at earnings levels below the earnings limit, or reinstatement provisions. In the earnings limit studies, researchers used the variation in earnings limits over time to identify the effects by examining the relationship between changes in earnings limits and changes in the earnings of persons with income in the neighborhood of the earnings limit. In their study, D. C. Stapleton and Tucker (2000) used this identification strategy with the addition of a comparison group to control for unobserved effects of time to estimate the effects of Medicaid earnings limits on the earnings of disabled SSI participants. D. C. Stapleton and Tucker estimated that a US$1,000 increase in the earnings limit increases earnings for those with earnings in the neighborhood of the earnings limit by between US$37 and US$482. However, the number affected by the earnings limit is small because only approximately 2% of disabled SSI participants have earnings in the neighborhood of the earnings limit. The General Accounting Office (GAO) examined the relationship between DI earnings limit changes and earnings changes for persons with earnings in the neighborhood of the earnings limit (GAO, 2002). The findings suggest the earnings limit decreases earnings for a small percentage of DI participants, less than 1%, but data limitations preclude a definitive finding.

In a related body of research, a number of researchers have estimated the overall DI program work participation effects consistently finding that the DI program reduces the work participation of beneficiaries (Chen & van der Klaauw, 2008). In the most direct estimate to date, Chen and van der Klaauw (2008) compared the work participation of rejected DI applicants to the work participation of accepted DI applicants. Using SSA administrative data linked to Survey of Income and Program Participation data, the authors compare beneficiaries and rejected applications and estimate that the work participation of DI beneficiaries would have been at most 20 percentage points higher in the absence of DI benefits. Chen and van der Klaauw also used a second approach to estimate the DI program effects on work participation. The approach exploits discontinuities in the disability determination process that occur at certain discrete ages (e.g., transition in age from 54 to 55). Using this method, the authors estimate that the DI program reduces work participation by between 6% and 12%.

Autor and Duggan (2007) raised the possibility that the negative effects of disability benefits on work participation among disability program participants may be caused by income effects rather than by benefit loss effects. That is, some people may choose not to work solely on the basis of the disability benefit income they receive regardless of the benefit loss they will incur if they work. If the income effects are predominant, policy changes to reduce benefit loss may not result in desired gains in work participation. To evaluate possible income effects, Autor and Duggan examined a disability program where there is no benefit loss with increased participant earnings, the U.S. Department of Veterans Affairs Disability Compensation Program (VDC). The authors use a 2001 liberalization of the VDC program that resulted in more Vietnam-era veterans being eligible and participating. They find that the liberalization substantially decreased work participation. Comparable research has not been conducted for the Medicaid, SSI, DI, or Medicare programs.

Thus, although there is strong evidence that disability programs reduce work participation, there is uncertainty about whether the effects are because of benefit loss effects, income effects, or both. In this article, we provide estimates of benefit loss effects that are separate from income effects. With the exception of D. C. Stapleton and Tucker (2000), these estimates are not available in prior research. The estimates provide an indication of whether changes in program rules to reduce benefit loss will increase work participation and earnings.

Disability Program Background

The benefit loss characteristics vary by program. In this section we describe the earnings limits, benefit loss at the earnings limit, benefit reduction at earnings below the limit, and reinstatement provisions for the following four programs: SSI, MassHealth, DI, and Medicare. The benefit loss characteristics are summarized in Table 1.

Table 1.

Benefit Loss Characteristics

Program	Earnings limit	Loss at earnings limit	Reduction below the earnings limit	Reinstatement provisions
MassHealth	None^a	Not applicable	None for gross income below 150% FPL. Increases in six steps from approximately 5.5% for persons with gross income of 150% FPL to 17.7% for persons with gross income of 1,000% FPL	Not applicable
SSI	US$1,453 per month^b	Small remaining benefit	50%	Possibility of expedited reinstatement without new application for up to 5 years following loss of benefits
DI	US$1,000 per month^c	Full benefit^d	None	Resumption of benefits without new application during the 36-month extended period of eligibility after trial work period ends. Also possibility of expedited reinstatement without new application or new waiting period for up to 5 years following loss of benefits.
Medicare	US$1,000 per month^e	Full benefit^f	None	Follows DI provisions. Previous Medicare waiting period months count toward current application waiting period if prior DI benefits were within 6 years.

Note: FPL = federal poverty level; SSI = supplemental security income; DI = social security disability insurance.

Applicable for persons working at least 40 hr per month.

2009 earnings limit for persons with no unearned income. For persons with unearned income, the earnings limit will be less.

Limit in 2010 after the trial work period and three grace period months, substantial gainful activity (SGA) amount.

After the trial work period and three grace period months, there is a full loss of benefits.

Limit in 2010 after the extended Medicare period, SGA amount.

After the extended Medicare period, there is a full loss of benefits.

SSI

The SSI program, administered by the SSA, provides cash assistance payments to low-income elderly persons and low-income nonelderly persons with severe disabilities. For our purposes, we use the term SSI participants to refer to nonelderly working-age adult SSI participants who qualify on the basis of their disability. SSI eligibility is contingent on a person’s disability, income, and assets. The disability standard is based on a SSA determination of earnings potential and the duration of disability. To meet the earnings potential requirement, an applicant must be earning less than the “substantial gainful activity” amount of US$1,000 per month in 2011 and either have a qualifying medical condition or be determined by SSA to have a limited functional capacity that precludes earnings in excess of SGA. To meet the duration requirement, the disability must be expected to last 12 months or result in death. To be eligible, an applicant’s net income, defined by Equation 1, must be less than a net income limit (see Note 1).

Net Income = (Unearned Income - US$20) + (Earned Income - US$65) / 2

The earned income limit is dependent on the person’s unearned income. As described by Equation 1, the earned income limit is the earned income amount, given the person’s unearned income, which results in a net income equal to the maximum SSI benefit. For persons with no unearned income, the earned income limit in 2010 is US$1,453 per month. The limit will be smaller for persons with unearned income. Also, the limit will be higher for persons with impairment-related work expenses (IRWE). Persons with an IRWE that enables work receive an earned income deduction in the amount of the IRWE when determining net income. Some persons’ net income is also reduced by their participation in the Plan to Achieve Self-Support (PASS) program. PASS participants set aside income to pay for expenses necessary to achieve a work goal. The set aside income is deducted from the net income as calculated by Equation 1. A nonmarried, working-age adult’s Federal SSI payment amount is the difference between the maximum payment amount, US$674 per month in 2010, and the person’s net income. In addition, the Massachusetts State Supplement program provides a supplemental payment of US$114.39 to SSI participants. SSI participants lose approximately US$1 of benefits for every US$2 of earnings over the full range of eligibility, at the earnings limit and below the earnings limit (see Equation 1). The SSI program does not have a benefit waiting period, and benefits may start in the month following the application month. The SSI program has reinstatement provisions that lower the cost of reinstatement for persons who recently terminated SSI because of an increase in earnings. Persons who lose SSI because of an increase in earned income may be reinstated, referred to as expedited reinstatement, without a new application or disability determination if they meet the following conditions: (a) they are earning less than SGA, (b) they are unable to earn at SGA levels because of their medical conditions, (c) their current impairment is the same or related to the impairment of the original disability determination, and (d) the reinstatement is within 5 years from the loss of benefits.

Massachusetts Medicaid, MassHealth

The Massachusetts Medicaid program, referred to as the MassHealth program, provides health insurance to low-income elderly, children, persons caring for children, and persons with severe disabilities. For our purposes, we use the term MassHealth participant to refer to working-age adults who qualify for Massachusetts Medicaid on the basis of disability. MassHealth coverage is contingent on disability, income, and asset requirements. The MassHealth and SSI programs are closely linked, with SSI participants being automatically enrolled in MassHealth. In addition, MassHealth has less restrictive disability and financial eligibility provisions to provide eligibility for persons who do not meet SSI requirements. The disability standard for MassHealth matches the disability standard for SSI with the exception of the requirement that applicants’ earnings are less than the SGA amount. Also, the MassHealth program is unusual among disability-based programs in that there is no earnings limit for participants who work at least 40 hr per month or for persons who meet a one-time spend-down. The component of the MassHealth program with a 40-hr-per-month work requirement and no earnings limit is referred to as the CommonHealth Working program and the component of the MassHealth program with a one-time spend-down is the CommonHealth Not-Working program. The CommonHealth program was initiated as a state-funded program in 1988 and became part of the Massachusetts Medicaid program under an 1115 Waiver in 1997. For the purposes of our analysis, because of the low monthly hours required, we assume the 40-hr requirement does not limit persons’ earnings, and thus, effectively, there is no MassHealth earnings limit. Because there is no earnings limit, there is also no benefit loss at the earnings limit and no cost of reinstatement. However, MassHealth does require cost sharing in the form of premiums for persons whose gross income is greater than 150% of the federal poverty level (FPL; see Note 2). The premiums result in an effective benefit loss, in the form of higher out-of-pocket premium costs, that is not constant but increases with gross income in six steps, ranging from a US$5 per 10% FPL increase in gross income for persons with gross income between 150% and 200% of FPL to a US$16 per 10% FPL increase in gross income exceeding 1,000% of FPL. For persons with no unearned income in 2009, this corresponds to approximately a US$0.05 increase in premium for each US$1 of increased earnings for a person with annual earnings of US$16,245 (net income of 150% FPL) and approximately a US$0.18 increase in premium for each US$1 of increased earnings for a person with annual earnings of US$108,300 (net income of 1,000% FPL).

Social Security Disability Insurance

The DI program, administered by the SSA, provides cash payments to persons who meet disability and work history requirements. The work history requirements vary by age, but generally, persons need 40 work credits to qualify. Credits are determined annually by dividing annual earned income by a year-specific standard, US$1,120 in 2010. A maximum of 4 credits can be earned per year. The DI disability standard is the same as the SSI disability standard described above. First-time participants must meet the disability standard for a period of 5 months before receiving cash payments. After a 9-month trial work period (TWP) and 3 grace period months when persons may work without an earnings limit, DI benefits are suspended when earnings exceed SGA in any month until the 36th month after the TWP completion, and terminated if they are above SGA in any month thereafter. Thus, the earnings limit for DI participants is the SGA amount once the TWP and grace period are exhausted. The full benefit is lost as earnings cross the earnings limit. Other than the requirement that persons are earning less than the SGA amount to meet the disability standard, there are no other income limits. Also, for persons with an IRWE, the IRWE is deducted from earned income prior to determining if earnings exceed the SGA amount. There is no benefit reduction for earnings below the earnings limit. Similar to the SSI program, the DI program also includes provisions to lessen the cost of reinstatements that occur soon after benefits are terminated because of earnings. The DI program includes a 36-month extended period of eligibility after the TWP ends, during which benefits are reinstated immediately without a new application for months when earnings are less than SGA. After the 36th month, the criteria for DI expedited reinstatement are identical to the criteria for the SSI program listed above. Persons who do not meet the criteria for the extended period of eligibility or the expedited reinstatement face the possible future costs of a new application including a new disability determination and a new 5-month benefit waiting period.

Medicare

Medicare provides hospital insurance (Part A), medical insurance (Part B), and, starting in 2006, drug coverage (Part D) to DI participants. There are no Part A premiums for DI participants; however, there are coinsurance and deductible cost-sharing requirements. Part B has a premium, US$96.40 per month for most beneficiaries in 2010, and deductibles. DI participants are entitled to Medicare starting 24 months after their DI entitlement date. For persons who terminate DI because of earnings above SGA and continue to work at these levels, Medicare continues for up to 93 months (extended Medicare). Thus, the earnings limit for Medicare participants is the SGA amount once the extended Medicare period is exhausted. There is no earnings limit during the extended Medicare period. Persons who lose their benefits because of earnings and continue to have a disabling impairment may buy Medicare coverage. There is no benefit reduction for earnings below the earnings limit. Because Medicare entitlement is closely linked to DI entitlement, the costs of Medicare reinstatement mirror the cost of DI reinstatement with the exception of the benefit waiting period. For persons who are reinstated to DI within 6 years of the previous DI benefit, prior Medicare waiting period months count toward the Medicare waiting period months for the reinstatement.

Comparison Across Programs

Comparing program benefit loss characteristics, there is considerable cross-program variation (see Table 1). MassHealth eligibility is not affected by earnings so there is no earnings limit, benefit loss at the earnings limit, or earnings-related reinstatement provisions. In contrast, the SSI, DI, and Medicare programs all have earnings limits, benefit loss at the earnings limit, and earnings-related reinstatement provisions. Comparing earnings limits for persons with low levels of unearned income, the SSI earnings limit are higher compared with the earned income limit of the DI and Medicare programs. Comparing benefit loss at the earnings limit, the US$1 benefit loss for US$2 of earnings for the SSI program is much lower than the full benefit loss of the DI and Medicare programs. The SSI program has a high (US$1 for US$2 of earnings) benefit reduction for earnings below the earnings limit. In contrast, there is no benefit reduction below the earnings limit for the DI, SSI, or low-income MassHealth participants. For MassHealth participants with income above 150% FPL, the benefit loss (in the form of premium payments) ranges from 5.5% to 17.7% of every dollar earned, much lower than the 50% of every dollar earned for SSI. There is also cross-program variation in reinstatement provisions. The SSI and DI programs have identical expedited reinstatement provisions but only the DI program has a 36-month period of extended eligibility. Medicare participation is tightly linked with DI participation; however, Medicare has a much longer period of extended eligibility, up to 93 months compared with 36 months for DI. For persons returning to DI within 6 years, the 24-month Medicare waiting period is reduced by counting previous waiting period months toward the current waiting period. We expect this variation across programs in benefit loss characteristics to result in cross-program variation in benefit loss effects, which we discuss in detail in the next section.

Expected Benefit Loss Effects

How would we expect these differences to affect the work decisions, whether to work and how much to work, of persons with disabilities? We examine the decision on whether to work first.

A person’s decision on whether to work is dependent on the wages available. In economic theory, the reservation wage is the lowest wage that will elicit work participation from a nonworking person (Heckman, 1974; Mitra, 2007). Each person has a unique reservation wage that depends on the person’s preferences and unearned income. Benefit loss characteristics do not affect a person’s reservation wage, but they do affect the effective wages available to a person. Benefit reductions that occur with increases in earnings are, in effect, a reduction in wages. The wages, adjusted for the benefit reduction, must exceed the reservation wage to elicit work participation. We would expect programs with greater benefit reductions as earnings increase to have larger negative effects on work participation compared with programs with smaller or no benefit reductions. Thus, we would expect a much larger negative work participation effect for the SSI program with the US$1 for US$2 benefit reduction compared with the DI or Medicare program where there is no benefit reduction until earnings exceed SGA amounts or the MassHealth program where premiums results in relatively small reductions in effective wages.

Work participation may also be affected by the earnings limit and the associated benefit loss as earnings cross the limit. The SGA earnings limit for DI and Medicare recipients may hinder employment for some persons. Constraining earnings to amounts less than the SGA may limit job opportunities. Furthermore, because of work-related expenses, for example, transportation, clothing, and child care, it may not be worthwhile for persons to work at the low earnings levels below the SGA. We expect the negative effects on work participation to be greater for the DI and Medicare programs compared with the SSI program and MassHealth program because of the lower earnings limits of the DI and Medicare programs.

How do benefit loss characteristics affect how much people work? The effects of the benefit reductions below the earnings limit (e.g., US$1 for US$2 SSI reduction) are theoretically ambiguous. Some people may work fewer hours because it is not worth their effort to work as many hours at the reduced effective wage. However, other people may work more hours because additional hours will be necessary to achieve the same income. Thus, among those who work, it is not clear whether the high benefit reduction of the SSI program will elicit more or fewer hours of work relative to the small benefit reduction (in the form of premiums) of the MassHealth program or the nonexistent benefit reduction of the DI and Medicare programs.

The earnings limit and the benefit loss at the earnings limit are expected to affect how much people work. The total loss of DI and Medicare benefits with earning exceeding the earnings limit (SGA amount) is expected to reduce work hours. Some working participants will likely constrain their work hours to ensure that their earned income remains below the SGA amount to avoid a total loss of benefits. We expect that this effect would only be predominant among persons with earnings in the neighborhood of the SGA amount. The effects of the earning limit and benefit loss at the earnings limit is expected to be smaller for the SSI program, relative to DI and Medicare, because the earnings limit is higher and the benefit loss at the earnings limit is smaller for SSI. Because there is no MassHealth earnings limit or benefit loss at the earnings limit, we do not expect any MassHealth effects on how much people work.

In summary, we expect the SSI program, because of the US$1 for US$2 benefit reduction, to have a larger negative work participation effect compared with the MassHealth, DI, and Medicare programs. It is possible that the SGA limit may reduce the work participation of some DI and Medicare participants. We do not expect that the MassHealth premiums will have any substantial effect on work participation.

Among working DI and Medicare participants, we expect the benefit loss at the earnings limit to have a negative effect on how much participants work; however, we expect these effects to be limited to persons with earnings in the neighborhood of the SGA amount. There is uncertainty about the effect of US$1 for US$2 benefit reduction on how much SSI participants work. Last, we do not expect any substantial MassHealth effects on how much people work.

Data and Method

The MassHealth Employment and Disability Survey (MHEDS)

We use a unique survey of MassHealth participants, the MHEDS, to estimate the benefit loss effects. The MHEDS is a 202-item survey conducted by the Center for Health Policy and Research at the University of Massachusetts Medical School and developed in 2005 under the Massachusetts Medicaid Infrastructure and Comprehensive Employment Opportunities (MI-CEO) grant from the Centers for Medicare and Medicaid Services. We conducted a dual-mode administration of the survey, which included two mailings of an English language version of the MHEDS and telephone interviews with members who did not respond by mail. The telephone interview version of the MHEDS was available in English and Spanish. The MHEDS sample consisted of a random selection of 2,939 MassHealth members from a population of approximately 175,000 working-age adults (19–64 years of age) who qualified for MassHealth on the basis of disability. The sample was stratified by age and MassHealth coverage group. The stratified random sample was drawn from MassHealth members using the Statistical Analysis System SurveySelect procedure. The stratification was necessary to ensure adequate representation of working MassHealth persons. The age stratification resulted in the following age distribution: 10% of the sample included members aged 19 to 25, 60% of the sample included members aged 26 to 49, and 30% included members aged 50 to 64. In addition to CommonHealth Working, the MassHealth coverage groups included two groups with low levels of work participation, MassHealth Standard and CommonHealth Nonworking.

All data were collected between November 2005 and January 2006, with a response rate of 53%. We compared respondents and nonrespondents. Using MassHealth administrative data, we found that respondents were significantly more likely to be female, older, and have higher family income than nonrespondents. Respondents and nonrespondents did not differ significantly on Medicare or private insurance enrollment nor on MassHealth expenditure.

The MHEDS contains a series of questions about how the prospect of benefit loss affects participants’ work decisions. Survey respondents report whether they “do not work” or they “earn less” because they might lose the respective benefit. The question was asked for each of the person’s respective benefits.

The respondent selection criteria for the question, the text of the question, and the effect measured by the question are presented in Table 2.

Table 2.

Selection Criteria, Survey Question, and Benefit Loss Effect

Respondent selection criteria	Question	Effect
Working MassHealth participants	QW1: Are you now earning less money than you think you could because you might lose your MassHealth coverage if you earn more?	Effect of MassHealth benefit loss on labor supply among working participants
Working MassHealth participants who report receipt of Medicare	QW2: Are you now earning less money than you think you could because you might lose your Medicare coverage if you earn more?	Effect of Medicare benefit loss on labor supply among working participants.
Working MassHealth participants who report receipt of disability cash benefits (SSI or DI)	QW3: Are you now earning less money than you think you could because it might affect your disability cash benefit if you earned more?	Effect of benefit (SSI or DI) loss on labor supply among working participants.
Nonworking MassHealth participants	QNW1: Are you not working now because you might lose your MassHealth coverage if you were working?	Effect of MassHealth benefit loss on work participation among nonworking participants.
Nonworking MassHealth participants who report receipt of Medicare	QNW2: Are you not working now because you might lose your Medicare coverage if you were working?	Effect of Medicare benefit loss on work participation among nonworking Medicare participants.
Nonworking MassHealth participants who report receipt of disability cash benefits (SSI or DI)	QNW3: Are you not working now because you do not want to lose your disability cash benefits?	Effect of benefit loss on work participation among nonworking (SSI or DI) participants.

Note: QW = question working; QNW = question nonworking; SSI = supplemental security income; DI = social security disability insurance.

Method

The responses to the MHEDS benefit loss questions provide a direct measure of the benefit loss effects by program. The benefit loss effect is the proportion of persons answering “yes” to the respective benefit loss question. Proportions were measured for eight groups corresponding to the MHEDS questions. The groups are (a) working MassHealth participants, (b) working Medicare participants, (c) working DI-only participants (no SSI), (d) working SSI-only participants (no DI), (e) nonworking MassHealth participants, (f) nonworking Medicare participants, (g) nonworking DI-only participants, and (h) nonworking SSI-only participants. The working and nonworking groups are mutually exclusive; however, because persons participate in multiple programs, persons may be included in more than one group within the working and nonworking categories.

To compare benefit loss effects between any two of the four programs, we compared the proportions answering yes to the benefit loss question for the respective programs. To statistically test if the proportions were equal for two programs, we used logistic regression and Generalized Estimating Equations (GEE). To do this, the data were pooled for the two programs being compared. The dependent variable in the logistic regression was the dichotomous answer to the benefit loss question for the respective program. The independent variable was a dichotomous variable representing the respective program. The independent variable coefficient estimate represents the difference in benefit loss effects between the programs. GEE was used to obtain coefficient estimate standard errors adjusted for the lack of independence between observations for persons participating in multiple programs. The adjusted standard errors were used to determine if the coefficient estimates were statistically significant.

Results

Summary Statistics

All survey respondents participate in the MassHealth program. Most respondents also participate in one or more of the DI, Medicare, and SSI programs. Survey respondents’ program participation, by work status, is described in Table 3.

Table 3.

Survey Respondent Program Participation Counts by Work Status Groups

Program participation	Working group (n = 481)^a	Not-working group (n = 877)^a
MassHealth only	148	61
MassHealth and DI	14	37
MassHealth and Medicare	60	105
MassHealth and SSI	9	80
MassHealth, DI and Medicare	184	406
MassHealth, SSI, and Medicare	56	140
MassHealth, SSI, and DI	0	6
MassHealth, DI, SSI, and Medicare	10	42

Source: MassHealth Employment and Disability Survey.

Note: DI = social security disability insurance; SSI = supplemental security income. See text for a description of MassHealth Employment and Disability Survey.

n is the number of persons in the respective group who provided responses for all of the survey questions corresponding to the DI, SSI, and Medicare program participation.

Summary statistics for the eight groups are presented in Table 4. The statistics are presented to provide a description of the sample and are not intended to provide an explanation for differences in benefit loss effects across groups. As expected, compared with nonworking persons, working persons are younger (p < .0001) and have better physical health (p < .0001), better mental health (p = .025), and higher education levels (p = .034; see Note 3). In addition, working persons have a lower proportion with a physical disability (p < .0001), and fewer working persons have a long-term illness (p < .0001) compared with nonworking persons. Comparing DI-only and SSI-only persons, the DI-only have higher levels of education (p < .0001).

Table 4.

Summary Statistics for Eight Groups: Means and Percentage of Sample

	Working group				Not-working group
Characteristic	Mass health (n = 467)^a	Medicare (n = 270)	DI-only (n = 177)	SSI-only (n = 59)	Mass health (n = 882)	Medicare (n = 623)	DI-only (n = 366)	SSI-only (n = 193)
Age (M)	40.0	40.2	42.3	35.6	47.9	47.9	49.0	47.2
Gender
Female	56.4	55.4	63.2	37.5	59.2	56.6	50.9	65.8
Male	43.6	44.6	36.8	62.5	40.8	43.4	49.1	34.2
Physical health
Excellent or very good	24.3	27.0	27.9	27.9	7.0	7.7	4.3	7.3
Good, fair, or poor	75.7	73.1	72.1	72.1	93.0	92.3	95.7	92.7
Mental health
Excellent or very good	25.7	14.8	11.0	32.1	17.8	18.1	22.2	18.3
Good, fair, or poor	74.3	85.2	89.0	67.9	82.2	81.9	77.8	81.7
Education level
Less than high school degree	23.3	25.1	20.0	36.9	33.9	27.1	19.8	43.7
High school degree	43.0	40.4	32.3	46.9	33.9	36.1	27.3	32.9
Some college	33.7	34.5	47.7	16.2	32.2	36.8	52.9	23.4
Physical disability
Yes	28.7	26.7	29.2	28.0	61.6	64.7	72.7	59.7
No	71.3	73.3	70.8	72.0	38.4	35.3	27.3	40.3
Psychiatric disability
Yes	62.1	63.4	77.0	47.9	71.5	70.6	74.5	72.8
No	37.9	36.6	23.0	52.1	28.5	29.4	25.5	27.2
Substance abuse
Yes	6.0	4.9	9.8	0.4	7.9	6.9	5.9	3.9
No	94.0	95.1	90.2	99.6	92.1	93.1	94.1	96.1
Developmental disability
Yes	14.2	12.8	10.3	26.9	9.4	6.2	7.0	11.5
No	85.8	87.2	89.7	73.1	90.6	93.8	93.0	88.5
Long-term illness
Yes	29.2	29.7	36.7	15.5	55.8	56.9	60.8	54.2
No	70.8	70.3	63.3	84.5	44.5	43.1	39.2	45.8

Source: MassHealth Employment and Disability Survey.

Note: See text for a description of MassHealth Employment and Disability Survey.

n is the number of persons in the respective group who provided responses for all of the survey questions corresponding to the summary statistics.

The Prospect of Losing Benefits Effects

The proportion of respondents answering “yes” to the eight questions and the 95% confidence intervals are presented in Table 5. All proportions are greater than 0 (p < .05). The proportions of working persons who report earning less because of the prospect of benefit loss are .260 for MassHealth, .314 for Medicare, .447 for DI-only, and .348 for SSI-only. We found the MassHealth effect on earnings to be smaller than the DI program effect (p = .07). We were not able to detect a difference in effect size for the other program comparisons.

Table 5.

Proportions Reporting Benefit Loss Effects for Eight Groups

Question	Group	Effect on	n ^a	Proportion “yes”	Lower 95% confidence interval	Upper 95% confidence interval
QW1	MassHealth	Earnings	474	.260	.173	.347
QW2	Medicare	Earnings	289	.314	.195	.433
QW3	DI-only	Earnings	187	.447	.272	.624
QW3	SSI-only	Earnings	64	.348	.162	.534
QNW1	MassHealth	Work	907	.072	.045	.098
QNW2	Medicare	Work	664	.093	.053	.133
QNW3	DI-only	Work	386	.048	.005	.090
QNW3	SSI-only	Work	209	.093	.046	.141

Source: MassHealth Employment and Disability Survey.

Note: QW = question working; QNW = question nonworking; DI = social security disability insurance; SSI = supplemental security income. See text for a description of MassHealth Employment and Disability Survey.

n is the number of persons in the respective groups who responded to the respective questions.

The proportions of nonworking persons who report not working because of the prospect of benefit loss are .072 for MassHealth, .093 for Medicare, .048 for DI-only, and .093 for SSI-only. We found the MassHealth effect on work participation to be smaller than the SSI effect (p < .01); however, the estimated difference in proportions is small, .021. We were not able to detect a difference between the MassHealth effect and the DI or Medicare effects. We were also not able to detect a difference in the effect size between the SSI program and either the DI or Medicare programs.

Discussion and Policy Implications

The results suggest that the elimination or reduction of benefit loss will increase the work participation among nonworking disability program participants. For example, eliminating benefit loss among DI participants would raise work participation rate from the current level of approximately 9% to a new rate of 13.8%—a very substantial 53% increase.

While these findings suggest the potential for a substantial increase in work participation, the findings also suggest that the prospect of benefit loss effects explain only a small portion of the work participation gap between disability program participants and persons not participating in disability programs. In comparison with the work participation rate of approximately 9% among DI and SSI participants, the overall work participation rate for adults, age 19 to 64, who are not SSI or DI participants is approximately 77% (see Note 4).

Unexpectedly, the results revealed a negative work participation effect for MassHealth even though MassHealth has no benefit loss other than the earnings loss associated with premium payments. Further research is needed to explain the sizable MassHealth effect. Possible explanations include (a) participant information on benefit loss that is imperfect and (b) an interdependence of benefit loss effects across programs for persons participating in multiple programs.

A high proportion of working participants report that the prospect of losing benefits decreases their earnings. The effects are consistently large across all four programs, ranging from approximately one quarter of MassHealth participants to nearly half of DI-only participants. These findings are consistent with the hypothesis that the BOND change to reduce the benefit loss for earnings above the SGA amount (US$1 for US$2) will increase the earnings among some working DI beneficiaries. This is also consistent with the findings of the Benefit Offset Pilot demonstration reducing benefit loss that show an increase in percentage of persons with earnings above SGA (D. Stapleton et al., 2010).

Contrary to our expectations, we found a large MassHealth benefit loss effect on how much MassHealth participants work, with approximately one quarter of MassHealth participants reporting earning less because of potential benefit loss. Because of the absence of a MassHealth earnings limit and the low MassHealth benefit reduction with increased earnings, we expected that the effect, if existent, would be very small. The large effect is surprising, and additional research is needed to explain the effect. Consistent with the above discussion on work participation effects, imperfect information and interdependence of effects across programs are possible explanations. Although the MassHealth effect is large, it is, as expected, smaller than the DI program benefit loss effects suggesting that program-specific benefit loss characteristics do matter. Also, although the Medicare program has a much longer period of extended eligibility compared with the DI program, we were not able to detect a difference between the Medicare and DI program benefit loss effects.

Limitations

The self-reported benefit loss effects may differ from the actual benefit loss effects because of self-reporting bias. Some persons may justify their not working or low earnings by reporting benefit loss as a reason. Other persons may accurately report benefit loss as a reason for not working or earning less but face additional barriers that would preclude work participation or higher earnings even if the benefit loss effects were eliminated.

Program participation status is contingent on self-report of SSI, DI, and Medicare participation. The self-reported benefit loss effects may differ from actual benefit loss effects because of inaccurate self-reports of program participation.

The data source used in this study does not provide detailed data on persons’ disabling condition or severity of disability, and we were not able to conduct separate analysis by subgroups. It is possible that results could, in part, be due to differences in disabling condition or severity of disability across programs.

Conclusion

Participants in public disability programs face the loss of benefits if they work and earn at substantial levels. Benefit loss is not a single program characteristic but a combination of multiple program characteristics, including earnings limits, benefit loss at earnings limits, benefit reductions at earnings below the earnings limit, and benefit reinstatement provisions. There is substantial variation in these characteristics across disability programs.

Benefit loss is expected to affect persons’ work decisions. Our findings suggest that there is a substantial benefit loss effect on work participation and earnings among disability program participants. Our estimates suggest that reducing benefit loss could change the work status of nearly 1 out of 10 disability program participants from nonworking to working and increase the earnings of approximately one third of working disability program participants. These findings are consistent with the hypothesis that the BOND change will increase the earnings among some working DI beneficiaries.

Surprisingly, we found relatively little variation in the size of the benefit loss effects across programs even though there is considerable cross-program variation in benefit loss characteristics. Notably, we found the MassHealth program effects to be sizable even though the actual MassHealth benefit loss that occurs with increased earnings is small compared with the benefit loss in the SSI, DI, and Medicare programs. The limited cross-program variation in the size of the effects is unexplained and an important area of future research. It is possible that the benefit loss effects are interdependent across programs and that the coordination of policies across programs will be necessary to reduce benefit loss effects among persons participating in multiple programs. It is also possible that the lack of participant understanding of benefit loss provisions (Hennessey, 1997; MacDonald-Wilson et al., 2003) may explain the limited cross-program variation. If this is the case, increased participant understanding of benefit loss, for example, through benefits counseling, in addition to reductions in benefit loss may be necessary to achieve increases in earnings and work participation.

Footnotes

Acknowledgements

The authors thank David Stapleton for very helpful suggestions on an earlier draft.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by the Massachusetts Medicaid Infrastructure and Comprehensive Employment Opportunities grant (CFDA No. 93.768), awarded by the Centers for Medicare and Medicaid Services to the University of Massachusetts Medical School. The opinions and conclusions in this paper are the sole responsibility of the authors, and do not represent the official views of the Commonwealth of Massachusetts, the Centers for Medicare and Medicaid Services, or the University of Massachusetts.

Notes

References

Autor

D. H.

Duggan

M. G.

(2007). Distinguishing income from substitution effects in disability insurance. AEA Papers and Proceedings, 97, 119–124. doi:10.1257/aer.97.2.119

Chen

van der Klaauw

(2008). The work disincentive effects of the disability insurance program in the 1990s. Journal of Econometrics, 142, 757–784. doi:10.1016/j.jeconom.2007.05.016

General Accounting Office. (2002). SSA disability: SGA levels appear to affect the work behavior of relatively few beneficiaries, but more data needed (GAO-02-224). Washington, DC: United States General Accounting Office.

Heckman

(1974). Shadow prices, market wages, and labor supply. Econometrica, 42, 679–694. doi:10.2307/1913937

Hennessey

J. C.

(1997). Factors affecting the work efforts of disabled-worker beneficiaries. Social Security Bulletin, 60(3), 3–20.

Kehn

Croake

Schimmel

(2010). A Government Performance and Results (GPRA) Report: The status of the Medicaid Infrastructure Grants Program as of 12/31/09. Retrieved from http://www.mathematica-mpr.com/publications/pdfs/disability/GPRA_09.pdf

Livermore

G. A.

(2008). Earnings and work expectations of Social Security Disability beneficiaries (Number 08-01). Retrieved from http://www.mathematica-mpr.com/publications/pdfs/earnings_disbrief1.pdf

MacDonald-Wilson

K. L.

Rogers

E. S.

Ellison

M. L.

Lyass

(2003). A study of the Social Security work incentives and their relation to perceived barriers to work among persons with psychiatric disability. Rehabilitation Psychology, 48, 301–309. doi:10.1037/0090-5550.48.4.301

Mitra

(2007). The reservation wages of Social Security Disability Insurance beneficiaries. Social Security Bulletin, 67, 89–111.

10.

Stapleton

Bell

Wittenburg

Sokol

McInnis

(2010). BOND Final Design Report. Cambridge, MA: Abt. Retrieved from http://www.ssa.gov/disabilityresearch/documents/BOND_Design%20Report_FINAL_Del2-2_12-17-10.pdf

11.

Stapleton

D. C.

O’Day

B. L.

Livermore

G. A.

Imparato

A. J.

(2006). Dismantling the poverty trap: Disability policy for the twenty-first century. Millbank Quarterly, 84, 701–732. doi:10.1111/j.1468-0009.2006.00465.x

12.

Stapleton

D. C.

Tucker

(2000). Will expanding health care coverage for people with disabilities increase their employment and earnings? Evidence from an analysis of the SSI work incentives program. In Salkever

D. S.

Sorkin

(Eds.), Research in human capital and development (pp. 133–180). Greenwich, CT: JAI.